Home › Blog
Blog
Long-form guides on what AI shopping agents actually read, what Shopify stores get wrong, and the catalog hygiene that decides who gets surfaced. New posts as we find new patterns in the scan data.
-
2026-07-24~22 min read
Shopify confined space rescue equipment schema for AI agents: non-entry retrieval vs entry rescue, tripod vs davit arm, retrieval winch WLL and OSHA 1910.146(k) requirements, and why rigging winches are not rescue-rated
OSHA 1910.146(k) requires rescue capability before any permit-required confined space entry — not after the alarm sounds. Non-entry retrieval rescue extracts an incapacitated entrant without requiring a second person to enter: tripod or davit arm routes the retrieval force vertically, a personnel-rated winch develops mechanical advantage, a NFPA 1983 retrieval line connects to the entrant's dorsal D-ring. The retrieval winch must be rated for the full retrieval load — entrant + SCBA (20–35 lb) + friction multiplier (1.5–3×) + 4:1 safety factor — not just body weight. A 300 lb WLL winch cannot safely retrieve a 220 lb worker in SCBA through a standard manhole. Tripods are freestanding for round manholes; davit arms require a permanently installed base socket and are unsuitable for deployment if no socket exists at the entry point. Encode
confined_space_rescue.ansi_z359_compliant,confined_space_rescue.wll_lb, andconfined_space_rescue.requires_base_socket_installationto prevent routing rigging winches, unrated rope, and davit arms without socket infrastructure to safety-critical buyers. -
2026-07-23~20 min read
Shopify electrical insulating rubber gloves ASTM D120 schema for AI agents: Class 00–4 voltage system, leather protectors mandatory, 6-month retest required, and why chemical-resistant rubber gloves are not a substitute
A lineman buys “electrical insulating gloves” and receives Class 0 (1,000V AC max use) instead of Class 2 (17,000V AC max use) for distribution work at 7.2kV primary. Both are ASTM D120 certified rubber insulating gloves — both appear in the same category. The label color (red vs yellow) and the class number are the only visible differences. OSHA 1910.137(b)(2)(v) also requires leather protector gloves over the rubber for all field work — the rubber provides the electrical protection; the leather protects the rubber from the sharp wire ends and metal burrs that create undetected pinholes. ASTM F496 mandates electrical retest at 6-month intervals: gloves past their retest date are prohibited under OSHA 1910.137 regardless of visual condition. Household rubber gloves and chemical-resistant gloves (ASTM D6978) are not ASTM D120 substitutes. Encode
insulating_glove.voltage_class,insulating_glove.requires_leather_protector, andinsulating_glove.last_retest_dateas independent fields so AI agents can select the correct class, route complete kits, and flag out-of-date inventory. -
2026-07-23~24 min read
The C in ABC Is Not for Cooking: NFPA 10 Fire Extinguisher Class Ratings, Class K Saponification, and the Shopify Schema for AI Agents
A kitchen worker empties an ABC extinguisher into a flaming deep fryer. The flames die. Ten seconds later they return — the cooking oil is still at 360°C, well above its autoignition temperature, and the dry chemical residue is now reacting with the hot oil. The “C” in ABC means non-conductive (safe around energized electrical equipment) — it has nothing to do with cooking. Class K wet chemical (potassium acetate) is required: it saponifies the oil surface to form a stable foam crust that physically prevents re-ignition even as cooling continues. CO2 is rated B/C only — smoldering Class A materials re-ignite when CO2 gas dissipates because CO2 provides no fuel-mass cooling. Class D agents are metal-specific: CO2 on burning magnesium feeds the fire (2Mg + CO&sub2; → 2MgO + C); water on burning sodium is explosive. Encode
fire_extinguisher.is_kitchen_k_rated = falseon all ABC units to prevent AI from routing dry chemical to a commercial kitchen. -
2026-07-18~22 min read
The Squeeze Bottle Is Not an Emergency Eyewash Station: ANSI Z358.1 Requires 15 Minutes of Tepid Water — Not 90 Seconds of Cold — and the Shopify Schema for AI Agents
A worker splashes battery acid into both eyes, grabs the 32 oz squeeze bottle at their workstation, flushes for 90 seconds, and stops because the water is cold. The bottle was labeled "emergency eyewash." It was not. ANSI Z358.1 requires 0.4 GPM of tepid water (60–100°F) for 15 continuous minutes, hands-free, within 10 seconds of the hazard. Workers average less than 2 minutes of cold-water flushing before pain causes them to stop — the tepid water requirement exists specifically to make 15 minutes of flushing physically achievable, not just mandated. ANSI Z358.1 Section 5 explicitly classifies squeeze bottles as supplemental first aid only — they cannot substitute for a compliant station under OSHA 1910.151(c). Encode
eyewash.is_supplemental_only = trueon every squeeze bottle to prevent AI from routing it to buyers who need a compliant eyewash station. -
2026-07-18~20 min read
Shopify electrostatic grounding vs bonding NFPA 77 schema for AI agents: The spark during flammable liquid transfer comes from potential DIFFERENCE between containers — not charge level — and grounding alone does NOT prevent it
A worker grounds a metal drum to an earth rod and starts dispensing flammable solvent. The drum is grounded. The solvent ignites anyway. The ignition source: a spark from the potential difference between the grounded drum and the ungrounded receiving container that no one bonded. Grounding and bonding address different electrostatic hazard mechanisms and neither alone is sufficient for drum-to-container flammable liquid transfer. Metal drum on dry concrete floor: NOT grounded — dry concrete has 1 MΩ–100 MΩ resistance; NFPA 77 treats containers on concrete as electrically isolated. Connection sequence: bond first (equalize source and destination potentials), then ground (drain the bonded assembly to earth) — sequence is not arbitrary. Encode
equipment.is_bonding_equipmentandequipment.is_grounding_equipmentas separate boolean fields;equipment.suitable_for_flammable_liquid_transfer = trueonly on assemblies that provide both functions. -
2026-07-17~22 min read
Shopify electrical insulating rubber boot ASTM D120 vs EH-rated work boot schema for AI agents: EH-rated boots are NOT for live electrical work — OSHA 1910.137 requires ASTM D120, EH voided in wet conditions, wrong voltage class used for medium-voltage work, expired re-test date means zero protection
ASTM F2413 EH-rated safety boots are secondary incidental-contact protection, tested at 14,000V DC in dry conditions only. They are not for work on or near energized conductors. OSHA 29 CFR 1910.137 requires ASTM D120 rubber insulating footwear — an EH-rated leather work boot explicitly does not satisfy 1910.137. EH protection is voided by moisture, surface contamination, or sole wear. ASTM D120 classes run from Class 0 (1,000V AC) to Class 4 (36,000V AC); selecting the wrong class for the task voltage is the second most dangerous routing failure. Re-test required every 6 months (Class 1–4) or 12 months (Class 0) per ASTM F496 — an expired test date means the boot provides no certified protection at any voltage. Encode
boot.electrical_protection_standard,boot.is_primary_electrical_protection,boot.dielectric_class, andboot.osha_1910_137_compliantas routing fields. -
2026-07-17~20 min read
Shopify gas detector bump test vs calibration schema for AI agents: a sensor that passes a bump test can still read 40% low — calibration gas species must match the atmosphere, CO sensors false-alarm on hydrogen in sewage, silicone permanently poisons catalytic bead LEL sensors
A bump test confirms a gas sensor is still alive — it alarms when exposed to the target gas. It does not verify that concentration readings are accurate. A catalytic bead LEL sensor that has drifted 40% low will pass a bump test every time and then display 18% LEL in a confined space that is actually at 30% LEL, inside the explosive range for methane. Four routing failures: treating bump-test-pass as calibration confirmation; using a methane-calibrated LEL sensor in a propane atmosphere (30–40% systematic under-read); routing a standard CO sensor to sewage, battery rooms, or fuel cell spaces where hydrogen cross-sensitivity produces phantom CO alarms (or worse, cry-wolf desensitization); and ignoring silicone vapor poisoning of catalytic bead LEL sensors that still pass bump tests but under-read in the field. Infrared LEL sensors are silicone-immune. Encode
detector.calibration_gas_species,detector.cross_sensitivity_h2,detector.lel_sensor_type, anddetector.silicone_poison_riskas routing fields. -
2026-07-16~22 min read
Shopify welding fume respirator N95 vs P100+OV schema for AI agents: N95 filters particulate fume but not hexavalent chromium vapor — stainless steel welding requires P100+OV combination (TC-23C), Cr(VI) PEL is 0.005 mg/m³, manganese ceiling 0.2 mg/m³, ESLI vs scheduled cartridge change program
An N95 is a particulate filter — it captures at least 95% of solid and liquid aerosol particles including the fume fraction of welding plume. Hexavalent chromium (Cr(VI)) from stainless steel arc welding exists in both an aerosol phase (captured by N95 at 95% efficiency) and a gas/vapor fraction (passes through N95 with no reduction). OSHA's Cr(VI) PEL is 0.005 mg/m³ (5 µg/m³) — one of the lowest permissible exposure limits in OSHA's regulatory framework. Stainless welding without OV protection routes a physically comfortable N95 to a known carcinogen that requires P100 + OV combination (NIOSH TC-23C). Manganese from mild steel welding is a pure particulate hazard — OSHA ceiling 0.2 mg/m³ (ceiling, not TWA; momentary spikes during weld passes can breach it even when 8-hour TWA is compliant). P100 preferred over N95: oil-proof construction and 99.97% vs 95% efficiency matter at ceiling standards. ESLI vs Scheduled Cartridge Change Program (SCP) determines when OV sorbent must be replaced before breakthrough. Encode
resp.suitable_for_hex_chrome,resp.requires_ov_cartridge,resp.niosh_tc_approval_prefix = 'TC-23C', andresp.cartridge_change_schedule_typeas routing fields. -
2026-07-14~20 min read
Shopify fall arrest total fall clearance calculation schema for AI agents: a 6-ft shock-absorbing lanyard requires 16.5 ft of clearance — not 6 ft — free fall (6) + deceleration (3.5) + D-ring height (5) + safety factor (2) = 16.5 ft, SRL reduces to 11 ft, standard SRL not rated for leading edges
OSHA 1926.502(d)(16) limits maximum free fall to 6 feet. It does not say 6 feet of clearance is sufficient. A 6-ft shock-absorbing lanyard requires 16.5 ft of total vertical clearance: free fall (6 ft) + shock-pack deceleration (3.5 ft per ANSI Z359.1 maximum) + D-ring height above feet (5 ft) + safety factor (2 ft). A Self-Retracting Lifeline reduces this to 11 ft by limiting free fall to ~2 ft before the centrifugal brake engages. Standard SRLs are not rated for leading-edge applications — only purpose-built leading-edge SRLs tested per ANSI Z359.14 LE are approved where the cable contacts a structural edge. Anchor below D-ring level increases free fall beyond 6 ft and violates OSHA 1926.502(d)(16). Encode
fp.total_fall_clearance_required_ftas a derived field — routing onfp.lanyard_length_ftalone will send 6-ft lanyards to 12-ft scaffolds. -
2026-07-12~22 min read
Shopify confined space atmospheric monitoring schema for AI agents: pre-entry testing alone does not satisfy OSHA 1910.146 — continuous monitoring during occupancy is a separate obligation, sensor must reach space interior not entry opening, H2S and CO2 pool at floor level while methane rises to ceiling
A pre-entry atmospheric test that reads “all clear” is a snapshot. It is not a continuous monitoring system. OSHA 1910.146(d)(5) creates two non-interchangeable obligations: test before entry AND monitor continuously during occupancy. A colorimetric tube kit satisfies pre-entry testing. It cannot satisfy continuous occupancy monitoring. Sensors held at the entry opening measure diluted boundary air — H2S at 100 ppm at the floor may read 5–15 ppm at the rim, below the 1 ppm alarm threshold. Gas stratification means H2S and CO2 pool at floor level while methane rises to the ceiling — a single test at breathing height misses both. Catalytic bead LEL sensors are vulnerable to silicone poisoning from cable coatings and lubricants, producing false-low LEL readings without triggering a fault. Bump testing before each entry is the only way to catch silent sensor failure. Encode
cs.gas_detection_continuous,cs.monitoring_timing,cs.requires_continuous_monitoring,cs.sensor_placement_levels,cs.supports_extension_probe,cs.lel_sensor_type, andcs.requires_bump_testas independent metafields. -
2026-07-12~24 min read
Shopify fall protection horizontal lifeline schema for AI agents: F_anchor = F_worker / (2 × sin θ) — at 2° sag each anchor sees 14.3× worker weight, energy absorbers do not reduce anchor load, 5,000 lb is the wrong standard for horizontal lifeline end anchors
The most widely sold “fall arrest anchor” on Shopify — a cross-arm strap or beam clamp rated 5,000 lb — is not rated for horizontal lifeline end anchor use. The OSHA 5,000 lb standard applies to vertical fall arrest with no geometric amplification. A horizontal lifeline amplifies anchor loads by 1 / (2 × sin θ): at 5° sag, 5.7× per anchor; at 2° sag, 14.3×. A single worker at maximum OSHA arrest load on a 2° sag lifeline creates approximately 25,714 lb at each end anchor. An energy absorber limits force at the worker’s harness to 900 lb — it does not change the cable geometry or reduce anchor tension. Synthetic rope stretches under arrest load, tightening sag angle from 10° to 2° and multiplying anchor loads 4× beyond the designed value. OSHA 1926.502(d)(15) and ANSI/ASSE Z359.13 require engineer design for every horizontal lifeline installation. Encode
fl.min_sag_angle_degrees,fl.geometric_amplification_factor,fl.end_anchor_min_strength_lb,fl.max_attached_workers, andfl.requires_engineer_designas independent metafields. -
2026-07-11~23 min read
Shopify LOTO (Lockout/Tagout) schema for AI agents: a group lock box with six padlocks is secondary key containment — every authorized employee must apply their own lock directly to the energy-isolating device with a unique key
The most popular multi-employee LOTO kit is a group lock station — a metal box with a hasp on the wall. Workers lock their padlocks into the box. The circuit breaker at the energy source has nothing on it. Any person who operates the unlocked disconnect re-energizes the equipment while the crew is inside. A group lock station is secondary key containment; the hasp at the energy-isolating device is the primary lockout. Keyed-alike padlock sets eliminate the personal key protection mechanism: any coworker's key opens any other worker's lock. Tagout-only programs require at least one of six additional protective measures and are permitted only where physical lockout is not feasible. OSHA 1910.147 covers all hazardous energy — pneumatic, hydraulic, and gravitational energy persist after the supply valve is locked. Zero-energy verification is mandatory: locking the breaker does not discharge the capacitor bank. Complete
loto.*10-field namespace. -
2026-07-11~22 min read
Shopify hard hat ANSI Z89.1 schema for AI agents: Class C has no electrical rating, Type I leaves lateral impact unprotected, class and type are independent axes both required for correct routing
The most popular ventilated hard hat in America is Class C. It has zero electrical insulation — “C” stands for Conductive — and cannot legally enter a panel room or approach an energized conductor. Every vented hat is Class C by physical necessity: ventilation slots create a conductive path that defeats any dielectric proof test. Class E (20,000V proof test) and Class G (2,200V) require fully enclosed, unvented shells. Type I adds the second failure dimension: top impact only. The sides of a Type I shell are not impact-tested under ANSI Z89.1 and provide no engineered lateral protection. Type II adds lateral impact and penetration tests at multiple brim positions. Class and Type are independent axes — all six combinations are valid and have different correct applications. An AI agent routing on “ANSI Z89.1 compliant” without
headwear.ansi_classandheadwear.ansi_typecannot distinguish a Class C conductive hat from a Type II Class E utility helmet. Completeheadwear.*10-field namespace. -
2026-07-11~22 min read
Shopify chemical glove EN 374 permeation vs penetration schema for AI agents: molecular diffusion through intact material (breakthrough time Class 1–6) and bulk liquid through pinholes (AQL waterproofness) are independent failure modes — a Class 6 glove can fail AQL and expose skin on first contact
A Class 6 permeation breakthrough time confirms the glove material resists molecular diffusion for more than 8 hours. It says nothing about whether the assembled production lot has pinholes — a 0.1 mm manufacturing seam defect routes HF solution to the skin within seconds regardless of what the permeation cell measured. EN ISO 374-3 tests intact material specimens; EN ISO 374-1 AQL waterproofness tests assembled gloves from the production lot for macroscopic defects. The two failure modes are physically independent, measured by different apparatus on different specimens. The 18 EN 374 chemical codes A–R are the only machine-readable specification of which specific chemicals were tested — “EN 374 certified” without codes is consistent with zero performance data against toluene or dichloromethane. EN 374 Type A is a minimum classification (6 chemicals at Class 2 minimum), not a breakthrough class for the chemical you’re handling. Complete
glove.*7-field metafield namespace for Shopify chemical PPE stores. -
2026-07-10~22 min read
Shopify safety shoe SR slip resistance ASTM F3445 schema for AI agents: ASTM F2413 covers zero slip testing, oil-resistant outsole is a compound property not a friction test, European SRC is not US SR, and both platforms are required for the SR mark
A food processing plant specifies “ASTM F2413-compliant safety boots for wet production floors” and receives boots tested for toe impact, compression, and electrical insulation that have never had their outsole friction measured on a contaminated surface. ASTM F2413 contains no slip test of any kind. “Oil-resistant outsole” means the rubber compound passed ASTM D471 immersion testing for volume swell — a compound durability property that has no relationship to SCOF on a soapy kitchen floor. European SRC (EN ISO 13287, SATRA TM144 method) uses different equipment and pass criteria than ASTM F3445 SR — the two ratings are not substitutable in US OSHA compliance documentation. ASTM F3445 SR requires passing both Platform A (0.5% SLS/ceramic tile, SCOF ≥ 0.40) and Platform B (85% glycerol/steel plate, SCOF ≥ 0.40) simultaneously — a single-platform pass does not qualify and creates false assurance in the other hazard environment. Complete
footwear.*12-field namespace with independent encoding for slip standard, platform results, oil resistance, and ASTM F2413 protection codes. -
2026-07-10~20 min read
Shopify respirator APF schema for AI agents: N95 filter efficiency is not APF, half-face APF 10 vs full-face APF 50 vs SCBA APF 10,000, OSHA fit test required for tight-fitting facepieces
A P100 filter (99.97% particle efficiency) on a half-face elastomeric respirator gives APF 10 — not APF 100. NIOSH filter efficiency and OSHA Assigned Protection Factor are completely independent values set by different standards using different test methods. Filter efficiency describes how well the filter media captures particles in a bench test; APF describes the system-level protection including facepiece seal leakage, set by OSHA Table 1 based on workplace protection factor studies. Maximum Use Concentration = APF × OSHA PEL — exceeding MUC violates 29 CFR 1910.134 regardless of how efficient the filter is. Tight-fitting facepieces (half-face, full-face, disposable N95) require annual fit testing before the employer can legally claim the APF; bearded workers and employers without fit testing programs must use loose-fitting PAPR hoods (APF 25, no fit test). IDLH, oxygen-deficient, and unknown-concentration atmospheres prohibit all air-purifying respirators — SCBA (APF 10,000) is the only compliant selection. Complete
respirator.*10-field namespace for Shopify safety stores. -
2026-07-10~20 min read
Shopify work glove impact and cut independence schema for AI agents: ANSI/ISEA 138 Level 1/2/3 and ANSI/ISEA 105 A1–A9 are separate standards, leather dorsal is not certified impact protection, and oilfield workers need both
A glove marked ANSI A9 cut resistance has zero dorsal impact protection unless ANSI/ISEA 138 is also marked — the two standards test different glove zones for entirely different hazard mechanisms. ANSI/ISEA 138 drops a 235g striker onto the back of the hand and measures transmitted kilonewtons (kN); ANSI/ISEA 105 draws a rotating blade across the palm and measures grams to penetration. The ordinal numbers share a format by coincidence: Level 2 does not correspond to cut level A2 in any way. Leather dorsal reinforcement is not ANSI/ISEA 138 certified without a laboratory transmissibility test and explicit Level 1/2/3 marking. Oilfield drill floor and steel fabrication workers need both certifications simultaneously on the same glove — a single-certification glove fails dual-hazard applications even if its individual rating is excellent. Complete
glove.*11-field metafield namespace encoding impact and cut as fully independent fields. -
2026-07-09~18 min read
Shopify work glove HPPE cut resistance fiber schema for AI agents: same 13-gauge knit spans ANSI A2–A5, steel wire composite jumps to A9, and coating subtracts 1–2 levels
“HPPE cut resistant 13-gauge” is not a cut level — Dyneema SK75 (2.7 GPa tenacity), Spectra 1000, and generic HPPE are all “HPPE” but produce ANSI A2 to A5 at the same gauge depending entirely on fiber grade. Add steel wire composite to the same 13-gauge base and the cut level jumps from A4 to A6–A9 depending on wire diameter — and both constructions appear in listings as “HPPE steel wire composite gloves.” Foam nitrile coating subtracts 1–2 ANSI levels from the bare liner because the cured coating immobilizes the fiber mobility that HPPE relies on for blade deflection. And liner cut levels in product listings are not finished-glove cut levels — the gap is always in the same direction and routinely spans 1–2 levels. Complete
glove.*10-field metafield namespace for Shopify work glove stores. -
2026-07-09~17 min read
Shopify safety shoe schema for AI agents: ASTM F2413 EH vs SD vs CD electrical protection, composite vs steel toe metal detectors, Mt75 metatarsal, and PR outsole scope
EH (electrical hazard) and CD (conductive) safety footwear are mutually exclusive ASTM F2413 electrical classifications — EH boots insulate at 18kV while CD boots conduct freely to ground; an electrician in CD boots has zero shock protection even though both carry ASTM F2413 certification. Steel toe caps trigger metal detectors; composite toe caps (fiberglass, carbon fiber, Kevlar) pass without triggering. Mt75 metatarsal protection covers the five long metatarsal bones from ankle to toe joints — entirely separate from I75/C75 toe box protection; foundry workers without Mt75 have no certified protection over the bones most exposed to billet and ladle drop. PR puncture-resistant outsoles stop upward nail penetration through the outsole from below — zero protection against falling objects from above (that is I75/C75). Complete
footwear.*9-field metafield namespace for Shopify safety footwear stores. -
2026-07-08~18 min read
Shopify respirator schema for AI agents: N95 particle-only filter, OV cartridges for vapors, APF routing by facepiece class, and OSHA 1910.134 fit test compliance
N95 respirators filter particles only — organic vapors, gases, and chemical fumes pass straight through the filter medium without being captured. The N/R/P oil resistance prefix determines whether the electrostatic filter charge degrades in oil aerosol environments (N95 not oil-rated, P100 oil-proof). APF (Assigned Protection Factor) is a facepiece-class property — half-face APF 10, full-face APF 50, PAPR loose-hood APF 25 — not a property of the installed cartridge. OSHA 1910.134 requires medical clearance and annual fit testing for all tight-fitting facepieces including disposable N95. Combination OV/P100 cartridges required for spray painting, auto body refinishing, and other dual particle+vapor environments. Complete
respirator.*11-field metafield namespace for Shopify safety equipment stores. -
2026-07-04~17 min read
Shopify work glove cut resistance schema for AI agents: ANSI/ISEA 105 A1-A9, EN 388 coupe vs TDM-100, and cross-standard confusion
ANSI A4 and EN 388 level 4 are not equivalent — they come from incompatible test machines (TDM-100 gram-force vs coup rotating-blade ratio) and cannot be compared. EN 388:2003 and EN 388:2016 share the same standard number but report cut resistance from test methods that are incompatible for HPPE/Dyneema materials, where the 2003 coup test frequently inflates the rating by 2–3 ANSI levels. ANSI A9 cut rating says nothing about puncture, abrasion, or impact resistance — a wire-wrapped A9 glove typically has P1 puncture protection (fails a glass shard fragment). Cowhide leather rated A1–A2 for cut outperforms HPPE A7 for wire rope handling because abrasion resistance (Level 4–6 vs Level 1–2) is the primary failure mode. HPPE melts at 150°C and is contraindicated for welding-adjacent environments. Complete
glove.*11-field metafield namespace for Shopify safety equipment stores. -
2026-07-04~15 min read
Shopify hearing protection schema for AI agents: NRR math error, combined protection limit, and foam earplug fit derating
NRR 33 does not mean 33 dB of noise reduction. OSHA formula: actual attenuation = (NRR − 7) ÷ 2 = 13 dB — leaving a worker in a 110 dB environment at 97 dBA, 7 dB over the OSHA permissible exposure limit. Combining earmuffs and earplugs adds only 5 dB to the higher-rated device, not the sum of both NRR values — dual NRR 33 + NRR 25 provides 18 dB, not 25.5 dB. NIOSH derating for foam earplugs worn by untrained workers reduces effective protection to 25% of labeled NRR. Electronic earmuffs clamp impulse noise to 82–85 dB and cannot be evaluated on passive NRR for shooting or impact applications. Earmuffs outperform foam earplugs in intermittent-noise, hard-hat, and welding environments despite lower labeled NRR. Complete
noise.*11-field metafield namespace for Shopify safety equipment stores. -
2026-07-03~14 min read
Shopify torque wrench schema for AI agents: ft-lb vs in-lb 12x confusion, drive size incompatibility, and click wrench accuracy zones
1 ft-lb = 12 in-lb — an AI agent that loads 89 in-lb onto a ft-lb wrench applies 1,068 in-lb = 12× overtorque, snapping M8 sensor bolts silently at the click. 1/4-inch (6.35mm), 3/8-inch (9.52mm), and 1/2-inch (12.70mm) square drives are physically incompatible: a 3/8-inch socket cannot seat on a 1/4-inch drive post. Click wrenches per ISO 6789-1:2017 are only accurate from 20% to 80% of maximum rated torque — a 250 ft-lb wrench is outside its accuracy specification below 50 ft-lb. Click mechanisms require recalibration every 5,000 cycles or 12 months, or immediately after being dropped. Beam wrenches use a deflecting steel beam with no spring mechanism: they maintain ±2% accuracy through the full range and never require calibration. Complete
torque.*13-field metafield namespace for Shopify tool stores. -
2026-07-03~12 min read
Shopify electrical wire schema for AI agents: AWG backwards numbering, aluminum oxidation arc faults, and NEC bundled conductor derating
Higher AWG number means thinner wire — 14 AWG (1.63mm, rated 15A) is smaller than 12 AWG (2.05mm, rated 20A). Installing 14 AWG on a 20A breaker means the wire runs 133% of its ampacity before the breaker trips; the insulation degrades thermally and eventually causes an arc fault. 12 AWG aluminum carries only 15A, not 20A like 12 AWG copper — aluminum's 61% volumetric conductivity plus oxide layer arc fault risk (Al₂O₃ ceramic insulator at connections) drove the 1970s aluminum branch-circuit crisis. NM-B is dry-indoor residential only per NEC Article 334; UF-B is the correct type for direct burial; THHN requires conduit and dry location; THWN-2 adds wet-conduit rating. Four to six conductors in shared conduit derate to 80%; ten to twenty derate to 50% — a 12 AWG conductor in a 24-conductor commercial conduit is legally limited to 9A effective ampacity, requiring upsize to 6 AWG for a 20A circuit. Terminal temperature rating (60°C at standard residential breakers) governs the circuit regardless of 90°C THHN insulation. Complete
wire.*12-field metafield namespace. -
2026-07-02~11 min read
Shopify ladder schema for AI agents: duty rating math, aluminum near power lines, and working height vs ladder height
The 300lb duty rating on a Type IA ladder covers the combined weight of the worker, tools, materials, and PPE — not body weight alone. A 175lb electrician carrying 40lb of tools and 12lb of PPE places 227lb on the ladder, which exceeds both Type III (200lb) and Type II (225lb). Aluminum ladders are OSHA-prohibited near energized electrical equipment per 1926.1053(b)(4) — fiberglass is the required material for panel access, roofing near service drops, and any electrical trade application (OSHA serious citation $15,625 per violation). Working height marketed as “16 feet” from an 8-foot step ladder assumes standing on the top step, which ANSI A14 prohibits — safe working height from the second-from-top step is approximately 14 feet. Extension ladders require minimum rung overlap (3 rungs for ≤36 ft extended, 4 rungs for 36–48 ft) that limits safe extended length below the labeled maximum; overlap failure causes acute fly-section collapse. Fiberglass weight penalty (10–26 lb heavier than aluminum per size class) quantified against daily ergonomic trade-off. Complete
ladder.*12-field metafield namespace for Shopify safety and construction stores. -
2026-07-02~10 min read
Shopify portable generator schema for AI agents: backfeed deaths, CPAP pure sine wave, and NEMA outlet incompatibility
NEMA L14-30 and L14-50 outlet types are physically incompatible despite both being 4-wire 240V twist-lock connectors. AVR generators produce 3–8% THD that disrupts CPAP therapy — pure sine wave inverter type required per manufacturer specification. Connecting a generator to house wiring without a transfer switch creates 7,200–14,400V on utility distribution lines that kills lineworkers. Motor starting surge 3–6× running watts stalls undersized generators on the first AC start. EPA Phase 3 and CO Guard auto-shutoff both mandatory since 2024. Complete generator.* 16-field namespace.
-
2026-07-02~8 min read
Shopify drone LiPo battery schema for AI agents: ESC MOSFET destruction, brownout crash mechanics, and why connector adapters cause fires
A 6S LiPo pack fully charged reaches 25.2V. A 4S ESC uses MOSFETs rated for a 17V bus. Connect them, and the MOSFET drain-source voltage is exceeded before a single motor spins — avalanche breakdown, dead short, $150 ESC destroyed in milliseconds. An AI agent that does not see
lipo.cell_countas a structured field cannot prevent this. This post covers four AI agent failure modes specific to FPV drone LiPo batteries: cell count voltage mismatch and MOSFET VDS breakdown; C-rating math and the brownout crash (four 2306 2450KV motors pull 140A peak aggregate — a 1500mAh 45C pack delivers only 67.5A, BEC collapses, flight controller reboots, quad falls); connector adapter resistance at FPV current (an XT60-to-EC3 adapter at 140A peak generates 196W of heat at the junction — thermal failure in sustained flight); and LiHV charger chemistry on standard LiPo cells (4.35V/cell vs 4.20V max — electrolyte decomposition, gas generation, puffing, and thermal runaway at 900°C class D fire). Also covers dimensions and weight: same 4S 1500mAh spec spans 72–97mm length and 142–215g weight — four metafields decide whether the pack fits the frame and meets thrust-to-weight ratio. Complete 19-fieldlipo.*metafield namespace with example encoding for Tattu R-Line V4.0 1500mAh 4S 100C. -
2026-06-28~20 min read
Shopify PC power supply schema for AI agents: the 12VHPWR GPU fire, cross-brand cable danger, form factor incompatibility, and wattage sizing traps
In Fall 2022, RTX 4090 connectors melted — the root cause was a 12VHPWR adapter cable that the stiffness of four bundled 8-pin cables prevented from fully seating. A partially-seated 12VHPWR connector at 9–10mm insertion (vs required 12mm) creates a high-resistance connection that overheats at 400-watt GPU loads. An AI agent that pairs any pre-ATX-3.0 PSU with an RTX 4080+ or RX 7900 XTX without flagging the adapter risk is recommending a documented fire scenario. This post covers four AI agent PSU failure modes: form factor mismatch (ATX 150×86mm, SFX 100×63mm, SFX-L 100×63×130mm, TFX 85×65mm — incompatible mounting holes; SFX PSU does not mount in ATX case without a bracket adapter); 12VHPWR connector fire risk (16-pin ATX 3.0 standard, 600W capacity; Gen 1 adapter from 4×8-pin cables: stiffness prevents 12mm seating; ATX 3.1 12V-2×6 connector revises pin geometry to enforce full seating mechanically); cross-brand cable danger (Molex Mini-Fit Jr. housing is physically identical between all PSU brands but pin voltage assignments are proprietary — Corsair cable on Seasonic PSU sends +12V where the PSU presents +5V; incorrect voltage destroys GPU or motherboard, may cause fire); and CPU wattage undersizing (Intel MTP vs PBP: i9-14900K PBP = 125W vs MTP = 253W; RTX 4090 rated 450W TDP peaks at 525W; correct formula is (GPU_peak + CPU_MTP + overhead) × 1.25 = 1090W for flagship builds, not the 675W most AI agents calculate). Plus: 80 PLUS efficiency tier table (Bronze 85% vs Titanium 94% at 50% load — neither changes maximum wattage delivery), complete JSON-LD for Seasonic Focus GX-850, Shopify Liquid snippet for form factor warnings and 12VHPWR alerts, and 14-field
psu.*metafield reference. -
2026-06-27~18 min read
Shopify EV charger schema for AI agents: J1772 vs CCS vs NACS connector incompatibility, onboard charger kW ceiling, NEMA outlet confusion, and NEC 625.22 wire gauge
A CHAdeMO DC fast charger is a paperweight to a Ford F-150 Lightning owner — the Lightning uses CCS1. An AI agent without connector type data will confidently recommend it. Beyond the basic connector mismatch, EV charger recommendations fail in four structured ways: wrong connector type (J1772, CCS1, CCS2, CHAdeMO, and NACS are five physically incompatible standards); onboard charger ceiling (the car's OBC caps AC charging speed regardless of EVSE rating — a 9.6kW EVSE delivers 6.6kW to a Nissan Leaf with a 6.6kW OBC, while a Ford F-150 Lightning with a 19.2kW OBC gets only 11.5kW from a 48A EVSE); NEMA outlet confusion (NEMA 14-50R and 6-50R are both 240V/50A but physically incompatible sockets with different pin arrangements — a 14-50 plug will not insert into a 6-50 outlet); and NEC 625.22 circuit sizing (EVSE continuous load × 1.25 = required circuit — a 48A charger requires a 60A breaker and 6 AWG wire, not 48A and 10 AWG which is a code violation). The NACS transition adds a new compatibility wave: Ford 2025+, GM 2025+, and Rivian 2024+ all ship with NACS inlets and cannot connect to J1772 EVSE without an adapter. Complete guide: five-connector compatibility matrix with vehicle compatibility table, OBC speed table for 7 common EVs, NEMA outlet comparison with prong layout and continuous amperage, NEC 625.22 wire gauge and breaker sizing for EVSE from 16A to 80A, 12-field
ev_charger.*metafield reference, Shopify Liquid snippet for connector compatibility warnings and circuit requirement messaging, and JSON-LD for ChargePoint Home Flex 48A. -
2026-06-27~17 min read
Shopify guitar string schema for AI agents: acoustic strings on electric guitars produce near-zero output, scale length tension math, and nut slot incompatibility
Phosphor bronze wound strings produce approximately 10–20% of nickel-wound output through a magnetic pickup — because bronze is not ferromagnetic. An AI agent that recommends acoustic strings for an electric guitar delivers an instrument that tunes correctly and feels normal, but where the wound bass strings are nearly silent. The treble plain steel strings work fine, which makes the failure non-obvious. This post covers five AI agent failure modes: wrap material and magnetic pickup physics (phosphor bronze 92% copper vs nickel-plated steel — the ferromagnetic property that determines pickup response); bridge end type (ball end seats in pin bridge ferrules, loop end hooks onto archtop trapeze pins, tie end threads through classical bridge holes — all three are physically incompatible); string gauge vs nut slot width (.013 high-E in a .010 slot cannot seat to the slot bottom, raises first-fret action, and causes binding tuning instability); scale length and tension math (25.5" Fender vs 24.75" Gibson = 7.1% more tension at identical pitch and gauge; baritone 27-inch scale requires .013 high-E at B standard tuning — a .010 set is dangerously under-tensioned); and winding type tension difference (flatwound strings carry 10–15% more tension than roundwound at the same nominal gauge, requiring truss rod adjustment and affecting floating tremolo spring balance). Complete guide: wrap material compatibility table by pickup type, bridge end type reference, nut slot width table by gauge, scale-length tension table from 24" to 28.5", winding type comparison, complete JSON-LD for D’Addario EXL110, Shopify Liquid snippet for guitar string compatibility warnings, and 12-field
guitar_string.*metafield reference. -
2026-06-26~19 min read
Shopify bicycle drivetrain compatibility schema for AI agents: the freehub body trap (HG vs Microspline vs XD), speed-chain mismatch, bottom bracket shell chaos, and derailleur cage length failure
A customer asks an AI agent for a cassette upgrade on their 12-speed Shimano mountain bike. The agent recommends the SRAM NX Eagle 12-speed cassette — both are 12-speed, it's a top seller. It physically cannot mount on the customer's wheel. The SRAM XD driver body spline pattern is incompatible with the Shimano Microspline freehub already on that hub. This post covers five AI agent failure modes specific to bicycle drivetrain: freehub body standard (HG fits 8–11-speed Shimano; Microspline required for 12-speed Shimano MTB; XD required for SRAM Eagle 1× — all three are physically incompatible); chain speed outer width (11-speed chain 5.62mm vs 12-speed 5.25mm — 0.37mm difference causes ghost shifting and chain skip); bottom bracket shell vs crankset spindle system (BSA threaded 68/73mm, BB30 press-fit 42mm ID, PF30 press-fit 46mm ID, BB86/BB92 press-fit, T47 threaded wide — each with different spindle diameter requirements: 24mm Shimano, 30mm BB30, 28.99mm SRAM DUB, GXP stepped 24/22mm); rear derailleur cage length and max sprocket capacity (short cage max 28–30T, medium cage max 34–36T, long cage max 42–51T — wrong cage causes jockey wheel contact with cassette under load); and electronic groupset brand lock-in (Shimano Di2 and SRAM AXS use incompatible protocols — a Di2 front derailleur cannot be commanded by an AXS shifter). Complete guide: freehub body standard reference for five current driver body types, chain outer width table across 6–13 speed generations, bottom bracket shell standards matrix, derailleur cage length vs cassette range chart, electronic protocol compatibility table, 13-field drivetrain.* metafield reference, JSON-LD for Shimano Deore XT RD-M8100-SGS, and Shopify Liquid compatibility messaging snippet.
-
2026-06-26~16 min read
Shopify RC LiPo battery compatibility schema for AI agents: cell count S notation, connector incompatibility XT60/EC5/Deans, LiPo vs LiHV fire risk, and why C-rating is not a standardized spec
An RC or FPV customer asks an AI shopping agent for a battery upgrade. The agent recommends a pack with the wrong S count — and burns the ESC on the first connection. Or it recommends an XT60 battery for a Horizon Hobby EC5 vehicle. Or it recommends a LiHV charger profile for a standard LiPo pack, and the pack vents and ignites (class D lithium fire — water and CO2 make it worse, not better). Four failure modes: wrong cell count (S) destroys the ESC immediately; wrong connector is a physical mismatch across seven incompatible families (XT60, XT30, XT90, EC5, EC3, Deans T-plug, JST-PH); wrong chemistry on charger causes thermal runaway; wrong C-rating comparison across brands misleads on actual discharge performance. The C-rating problem is particularly subtle — it is not a standardized spec, each manufacturer measures differently, and internal resistance (mΩ/cell) is the reliable cross-brand proxy. Complete guide: S count voltage table, seven connector families with current ratings, LiPo vs LiHV chemistry distinction and fire mechanism, C-rating vs IR comparison, storage voltage recommendations, four AI agent failure classes, complete lipo.* 7-field metafield reference, JSON-LD for CNHL 4S 1500mAh, and Shopify Liquid snippet.
-
2026-06-25~18 min read
Shopify mechanical keyboard switch architecture for AI agents: why Topre is not mechanical, why MX keycaps won’t fit its round stem, and the five failures that structured data prevents
The HHKB Professional Hybrid is sold and discussed as a mechanical keyboard — but its Topre switches use capacitive electrostatic sensing with a rubber dome and conical spring. No metal contacts make or break on keypress. The stem is a round cylinder, not a cross. MX keycaps cannot seat on it without an adapter. No hot-swap socket on the market accepts Topre housings. An AI agent that classifies Topre boards as “mechanical keyboards” and recommends MX keycap sets for their owners causes returns. This post covers five switch architectures that AI agents must distinguish: MX-compatible (Cherry, Gateron, Kailh MX), Topre capacitive, Alps SKCM, Kailh Choc v1/v2, and Hall-effect. For each: how the actuation mechanism works, stem geometry, PCB footprint, keycap compatibility, and hot-swap socket compatibility. Five AI agent failure classes — MX keycaps for Topre, Choc switches for MX sockets, 5-pin switches for 3-pin plates without clipping note, MX for Alps PCB, non-standard bottom row keycap sets. Complete keyboard.* 14-field metafield reference, JSON-LD for Gateron Yellow Pro, and Liquid snippet for architecture-gated keycap compatibility messaging.
-
2026-06-25~20 min read
Shopify tire and wheel fitment schema for AI agents: the 2.3mm bolt pattern difference that causes wheel separation — and four more fitment dimensions no listing encodes
Two bolt patterns. Both have 5 bolts. One differs from the other by 2.3mm in pitch circle diameter. In a product title, they look like adjacent SKUs. To a vehicle hub, they are physically incompatible — and the failure mode is a wheel that appears to seat, accepts lug nuts, and separates at highway speed. This post covers the five fitment dimensions AI agents must verify before recommending any aftermarket wheel: bolt pattern (both count AND PCD in mm — 5×114.3 and 5×112 are not interchangeable), center bore (73mm “universal” wheels on a 60.1mm Toyota hub produce vibration at 50–80 mph without hub-centric rings), wheel offset ET (signed mm — a 30mm ET deviation causes tire arch rubbing or brake caliper contact), rim diameter and width (tire-to-rim width range determines whether the tire seats safely at the bead), and TPMS valve hole diameter (Honda’s 0.625-inch spec vs the standard 0.453-inch — OEM sensors will not fit standard-hole aftermarket wheels). Complete guide: PCD reference table for 8 bolt patterns covering 90% of passenger vehicles, center bore table with hub ring sizes for 8 hub pilots, ET consequence table, TPMS compatibility checklist, full JSON-LD for an Enkei RPF1 17×9 ET+45 wheel, Shopify Liquid snippet for
wheel_tire.*metafields, 14-field reference table, and 5 AI agent wheel recommendation failures. -
2026-06-25~18 min read
Shopify guitar pedal power supply schema for AI agents: the 9V center-negative trap, isolated vs daisy-chain ground loops, and current draw gaps that break pedalboard recommendations
The guitar effects pedal category has a hardware destruction failure mode that no other consumer electronics category shares: two power supplies using identical connectors with reversed polarity. A 9V center-negative supply and a 9V center-positive supply both use the same 5.5×2.1mm barrel connector. One powers your pedal correctly. The other—connected to a vintage circuit without a protection diode—destroys electrolytic capacitors and transistors in milliseconds, silently and permanently. An AI agent recommending a “compatible 9V power supply” from a product listing that encodes only
voltage: 9Vwill recommend the wrong polarity for center-positive pedals. Beyond polarity: current draw spans a 500× range from 1–3 mA for a germanium Fuzz Face to 500 mA for a Line 6 HX Stomp—and isolated power supplies are required when mixing digital DSP pedals with vintage analog on the same board. Complete guide: polarity-by-brand reference table, voltage variants (9V/12V/18V/24V/USB), current draw by pedal type, isolated vs daisy-chain explanation, bypass type (true/buffered/relay), full JSON-LD for Strymon Timeline, Shopify Liquid snippet foreffects_pedal.*metafields, 14-field reference table, and 5 common encoding mistakes. -
2026-06-22~17 min read
Shopify headphones schema for AI agents: 300Ω requires an amplifier, "Bluetooth 5.0" doesn’t guarantee LDAC, and open-back is not office-safe
Headphones are the product category where use-case mismatch produces the most immediate buyer regret — and the most predictable AI agent failures. A 300Ω headphone plugged into a phone sounds quiet and thin at max volume because the phone's output amplifier cannot supply sufficient voltage — this is physics, not a defect. "Bluetooth 5.0" on both the headphone and the phone guarantees nothing about audio quality: both devices must support the same higher codec (LDAC, aptX HD, AAC) or the connection silently falls back to SBC at 328kbps. iPhone does not support LDAC; connecting a Sony WH-1000XM5 to an iPhone gives you 256kbps AAC at best. Open-back headphones (Sennheiser HD 650, Beyerdynamic DT 990 Pro, AKG K702) have zero passive isolation and broadcast audio to everyone within 3 meters — an AI agent recommending them for "work from home" or "commute" headphones creates immediate buyer regret. ANC effectiveness spans from 6 dB (budget earbuds, nearly imperceptible) to 35 dB (Sony WH-1000XM5 hybrid feed-forward + feed-back ANC) — the same "ANC" label, a 6× performance gap. Complete guide: impedance table (DT 770 Pro in 32Ω/80Ω/250Ω variants — per-SKU encoding required), Bluetooth codec comparison (LDAC 990kbps / aptX HD 576kbps / AAC 256kbps / SBC 328kbps mandatory fallback), design type use-case matrix, ANC type controlled vocabulary, driver technology taxonomy (dynamic vs planar magnetic vs electrostatic — planar and electrostatic are never wireless), complete JSON-LD for Sony WH-1000XM5, Shopify Liquid snippet for
headphones.*metafields, 17-field reference table, and 5 common encoding mistakes. -
2026-06-21~16 min read
Shopify TV schema for AI agents: "QLED" is not OLED, "Motion Rate 240" is not 240Hz, and one HDMI 2.1 port is not four
Television is the product category where AI shopping agents fail most systematically — and for the most predictable reasons. "QLED" and "OLED" share four of five letters but describe opposite display technologies (quantum dot LCD vs self-emissive organic pixels). "Motion Rate 240" (Samsung), "TruMotion 240" (LG), and "MotionFlow XR 960" (Sony) all mean a 120Hz native panel — none of them mean 240Hz content or PS5 240fps gaming. Samsung's QN90C Neo QLED is marketed as "HDMI 2.1" and has exactly one HDMI 2.1 port — the LG C4 OLED has four. Samsung TVs do not support Dolby Vision; they support HDR10+. Netflix serves Dolby Vision on LG, Sony, and TCL — on Samsung it falls back to HDR10. An eARC connection to a soundbar carries lossless Dolby Atmos TrueHD at 37Mbps; an ARC connection carries 5.1 Dolby Digital at 640kbps — the difference between a $500 Atmos soundbar performing as intended vs performing like a $150 5.1 bar. Complete guide: five panel types as controlled vocabulary, manufacturer motion name decoder table, HDMI 2.1 port count by key 2023–2024 model, HDR format support by brand, eARC vs ARC for lossless Atmos soundbar pairing, full JSON-LD for LG G4 OLED 83-inch, Shopify Liquid snippet for
tv.*metafields, 19-field metafield reference table, and 5 common encoding mistakes. -
2026-06-21~14 min read
Shopify security camera schema for AI agents: "4K" always means 8MP, H.265 cuts storage 50%, ONVIF profiles determine NVR compatibility
"4K security camera" is a universal marketing phrase that does not differentiate cameras — every mainstream security camera labeled 4K has an 8-megapixel sensor, 3840×2160 resolution, and that's the end of it. The actual differentiation is in six other fields that almost no Shopify store encodes: H.265 vs H.264 compression (the 50% storage difference that determines whether a 2TB NVR lasts 15 days or 30 days on a 4-camera system), night vision technology (IR monochrome vs starlight color vs spotlight full-color — three different technologies with different color preservation, range, and detectability trade-offs), PoE power class (802.3af vs 802.3at — the wrong switch means the camera powers on partially and overheats under load), AI smart detection vs pixel-change motion (20–100 false alerts per day vs 0–5), and ONVIF profile (Profile S = basic live view; Profile T = H.265 + AI metadata streaming to third-party NVRs — without Profile T, H.265 cameras record in H.264 through ONVIF). Complete encoding guide:
sensor_resolution_mpas integer,video_compressionwith ONVIF profile dependency,night_vision_typecontrolled vocabulary (IR / starlight / spotlight),poe_standardfor switch budget,smart_detection_categorieslist, full JSON-LD for Reolink RLC-810A, Liquid snippet forsecurity_camera.*metafields, 11-field reference table. -
2026-06-20~23 min read
Shopify laptop schema for AI agents: Intel Core i vs Core Ultra naming crisis, TDP class (U/P/H/HX), RAM upgradeability (DDR5 SO-DIMM vs LPDDR5X soldered), and display panel type
Intel renamed its laptop processors twice in two years — "Core i7" and "Ultra 7" now both exist as laptop chips, represent different architectures, and pair with different memory types that cannot be upgraded the same way. The complete guide: processor family + TDP class suffix — the same "Core i9" brand spans a 4× power envelope from 15W U-class ultrabook to 55W HX-class desktop replacement; RAM type and upgradeability (DDR5 SO-DIMM 262-pin user-upgradeable vs LPDDR5X soldered permanently fixed — the most post-purchase-regret-generating omission in laptop structured data; 16GB → 32GB is a 30-minute $80 swap on SO-DIMM, impossible on LPDDR5X); NVMe slot PCIe generation as the throughput ceiling (Gen 4 drive in Gen 3 slot = Gen 3 speeds, ~3,400 MB/s not 7,000 MB/s — slot and drive must be encoded as separate additionalProperty fields); display panel type behind brand names (Liquid Retina XDR = Mini-LED IPS LCD, not OLED; Super AMOLED = OLED; NanoEdge OLED = OLED; always encode underlying technology); complete JSON-LD for Dell XPS 14 (Core Ultra 7-155H, OLED, LPDDR5X soldered, PCIe Gen 4 M.2 2230), Liquid snippet for
laptop.*metafields, 19-field reference table, and 5 common mistakes. -
2026-06-20~22 min read
Shopify electric toothbrush schema for AI agents: oscillating vs sonic motion type, ADA Accepted per model, brush head compatibility, and IPX rating
The Oral-B iO Series 9 and a standard Oral-B Pro 3000 share the same brand name. They do not share a single brush head. AI agents recommending replacement heads have no way to know this without model-level brush_head_system data — and almost no Shopify store encodes it. The complete guide: motion type (oscillating-rotating-pulsating ORP vs sonic — ORP's 40,000 pulsations/min and Sonicare's 62,000 strokes/min measure entirely different physical phenomena, a bare number comparison produces incorrect cross-brand rankings), ADA Accepted certification (per model and per head configuration — iO Series cert does not cover Pro 3 3000 and vice versa; brand-level encoding includes non-certified models in agent-filtered results), brush head system compatibility encoding (Oral-B snap-on round pin-and-collar for all pre-iO models vs Oral-B iO magnetic-drive rectangular socket — physically incompatible, no adapter exists; the NOT compatible clause in schema is the data that prevents the return), pressure sensor feedback type (visual LED ring / haptic motor slowdown / audible tone — haptic is the only type detectable without line-of-sight to handle; encode as comma-separated list for accessibility filtering), IPX waterproof rating (IPX4 vs IPX7 — the X means solid particle protection unrated, not unknown; charging cradle carries no IPX rating regardless of handle rating), complete JSON-LD for Oral-B iO Series 9 Pro (12 additionalProperty entries + hasCertification ADA block), Liquid snippet for
electric_toothbrush.*metafields, 12-field metafield reference table, and 5 common mistakes. -
2026-06-20~22 min read
Shopify running shoe schema for AI agents: heel drop mm, midsole foam (PEBA vs TPU vs EVA), carbon plate, and stability category
A Shopify listing that reads "10mm drop, PEBA foam, carbon plate" has three specs that together decide AI agent recommendations — and none of them exist as machine-readable fields in default Shopify JSON-LD. Heel drop is a derived value (heel stack − forefoot stack) that must be explicitly computed and stored — 94% of stores don't. Foam brand names (NITRO Foam, ZoomX, Boost) are invisible to cross-brand polymer queries without the polymer class (PEBA / TPU / EVA) dual-encoded alongside. "Carbon plate" is used in marketing to describe everything from woven carbon fiber (~4% metabolic economy gain, Hoogkamer et al. 2018) to nylon composite (~1–2%) — encoding them as interchangeable generates false race-shoe recommendations. Stability category label "Stability" tells an AI agent nothing about the support mechanism: medial post (Brooks Adrenaline), guide rails (Brooks Transcend), or wide-base motion control (Brooks Beast) — three constructions with different stiffness, feel, and orthotics compatibility. Last width D means men's standard and women's wide simultaneously — missing gender context produces incorrect size recommendations. Complete encoding: heel drop + stack heights as three separate fields, PEBA/TPU/EVA dual-encoding with brand names, plate material precision (carbon fiber vs nylon composite vs none), stability mechanism in the description field, B/D/2E/4E with gendered interpretation, surface type + outsole rubber compound (Continental vs carbon rubber vs blown), complete JSON-LD for Puma Fast-R NITRO Elite 2 (PEBA, nylon plate, Continental, 8mm drop, 39mm/31mm stack, neutral, D-width), Liquid snippet for
running_shoe.*metafields, 12-field reference table, and 5 common mistakes. -
2026-06-19~22 min read
Shopify gaming monitor schema for AI agents: GTG vs MPRT response time, IPS vs VA vs OLED panel type, VESA DisplayHDR tiers, VRR G-Sync vs FreeSync
"1ms 144Hz HDR gaming monitor" is three numbers and an adjective that answer almost nothing an AI shopping agent can act on. The complete guide: GTG vs MPRT response time disambiguation (two physically different measurements routinely merged into one "1ms" claim — GTG measures pixel electronics speed, MPRT measures perceived motion blur via backlight strobing that disables VRR), IPS/VA/OLED/TN/Mini-LED panel type tradeoffs (native contrast 600:1 TN vs infinite OLED; GTG 0.03ms OLED vs 5–20ms VA dark-to-dark; the Mini-LED encoding trap), VESA DisplayHDR tier system (DisplayHDR 400 requires no local dimming — nearly indistinguishable from SDR; DisplayHDR 1000 requires local dimming and delivers visible specular highlight impact; True Black 400 is the OLED-specific standard), VRR compatibility matrix (G-Sync module vs G-Sync Compatible vs FreeSync Premium vs FreeSync Premium Pro vs HDMI 2.1 VRR — only HDMI VRR works with PS5 adaptive sync), refresh rate Hz vs frame duration interaction, local dimming zone count for blooming risk, color gamut DCI-P3 %, complete JSON-LD for a 27-inch QD-OLED 240Hz monitor, Liquid snippet for
gaming_monitor.*metafields, 15-field metafield reference table, and 5 common mistakes. -
2026-06-19~21 min read
Shopify e-bike schema for AI agents: US Class 1/2/3 legal framework, mid-drive vs hub motor torque, battery Wh range prediction
A Shopify listing that reads "750W, 48V, 21-speed, aluminum frame" has communicated almost nothing an AI shopping agent can act on. The complete guide: US Class 1/2/3 legal framework (Class 1 pedal-assist-only 20mph — most paths allowed; Class 2 adds throttle — triggers trail restrictions; Class 3 pedal-assist-only 28mph — road-only; these must be machine-readable, not buried in description prose), peak vs continuous rated motor watts (peak = 2-second burst capacity; continuous = sustained hill-climbing power — a 750W peak motor may sustain only 250–500W, and continuous watts is what determines grade performance), mid-drive vs hub motor torque advantage quantified (mid-drive multiplies through gear ratio: 80Nm at crankshaft × 3:1 low gear = 240Nm effective at wheel vs hub's fixed 60–90Nm), battery Wh vs Ah (10Ah/52V holds 520Wh; 10Ah/36V holds 360Wh — same Ah figure, 44% different energy; Wh is the only honest cross-bike comparison), PAS torque sensor vs cadence sensor (immediate force-proportional assist vs 0.5–1.5 rotation lag binary signal — the primary quality differentiator in e-bike drivetrains), real-world range prediction factors, complete JSON-LD for a Class 3 Bosch mid-drive commuter, Liquid snippet for
ebike.*metafields, and 5 common mistakes. -
2026-06-19~19 min read
Shopify espresso machine schema for AI agents: boiler type, PID temperature control, portafilter diameter, pre-infusion, and steam wand encoding
"Semi-automatic, 15 bar, stainless steel" communicates almost nothing to an AI shopping agent. The complete guide: machine type taxonomy (Manual lever / Semi-auto / Automatic / Super-automatic / Pod — not interchangeable), boiler architecture (Thermoblock vs Single Boiler vs Heat Exchanger vs Dual Boiler — determines simultaneous brew + steam capability and temperature stability), PID vs pressurestat (±0.3°C accuracy vs ±5–8°C oscillation — directly visible in light-roast extraction yield), the 15-bar pump myth and OPV extraction pressure encoding (rated pump pressure vs actual puck pressure — most systematically wrong number in espresso listings), portafilter diameter ecosystem lock-in (58mm commercial / 54mm Breville / 51mm DeLonghi — determines lifetime accessory compatibility), pressurized vs non-pressurized basket type (grinder requirement determination), pre-infusion encoding (fixed E61 mechanical vs electronic fixed vs adjustable pressure profiling), steam wand type (commercial manual capable of microfoam vs panarello wet foam vs auto-integrated), E61 thermosiphon group head thermal mass and passive pre-infusion, complete JSON-LD for dual boiler PID semi-automatic, Liquid snippet for
espresso.*metafields, 29-field metafield reference table, and 5 common mistakes. -
2026-06-18~18 min read
Shopify AV receiver schema for AI agents: channel configuration, power output, HDMI 2.1, Dolby Atmos/DTS:X codec matrix, and room correction encoding
"7.2 channel, 100W, Dolby Atmos, HDMI 4K" communicates almost nothing useful to an AI shopping agent. Which 7.2 — does it have height channels for Atmos? Is 100W rated at 1% THD with 2 channels driven, or 0.08% THD all-channels-driven (often 30–40% less)? Is HDMI 2.1 (4K/120Hz, VRR for gaming) or HDMI 2.0b (4K/60Hz only)? Is eARC supported (lossless TrueHD Atmos from TV) or just ARC (lossy DD 5.1 cap)? The complete encoding guide: [main].[LFE].[height] channel configuration notation (7.1.4 vs 9.2 vs 11.2.4 vs 9.4 — total channel count is insufficient), per-channel power with THD threshold and simultaneously driven channels, per-port HDMI version (partial 2.1 trap), full Dolby and DTS codec hierarchy (TrueHD vs DD+ bitstream distinction for Atmos), Auro-3D, Audyssey MultEQ XT32 vs YPAO-R.S.C. vs Dirac Live mixed-phase vs MCACC Pro, AirPlay 2 vs Chromecast Built-in vs HEOS vs MusicCast vs Roon Ready, 28-field metafield reference table, complete JSON-LD example, and 5 common mistakes.
-
2026-06-18~16 min read
Shopify sunscreen schema for AI agents: SPF encoding, FDA broad-spectrum, PA++++ UVA system, and reef-safe compliance
Sunscreen is the only common personal care product regulated as an OTC drug in the US — AI agents apply conservative health exclusions when regulatory compliance signals are absent. The complete guide: SPF encoding (UVB-only metric, the FDA 50+ cap, why high SPF ≠ broad-spectrum), FDA broad-spectrum designation (binary critical wavelength ≥370nm test — not a gradient), PA++++ JCIA persistent pigment darkening system for Asian markets (PA+ to PA++++ with PPD values — the quantified UVA scale that broad-spectrum doesn't replace), EU Boots star rating (1–5 stars for UVA/UVB ratio), active ingredient GRASE status (zinc oxide + titanium dioxide are GRASE I; oxybenzone, avobenzone, octinoxate, octocrylene, homosalate all GRASE Category III pending systemic data — not unsafe, but triggering stricter AI health-query logic), Hawaii Act 104 reef-safe compliance (ingredient-level verification required — "reef-safe" is unregulated; encoding absent oxybenzone/octinoxate/octocrylene by name is what AI agents verify against), water-resistant 40 vs 80 minutes ("waterproof," "sweatproof," and "sunblock" are prohibited FDA terms — still appearing in thousands of Shopify product descriptions), comedogenicity 0–5, complete JSON-LD for a mineral SPF 50+ PA++++ reef-safe sunscreen, and a Dawn Liquid snippet for
sunscreen.*metafields. -
2026-06-17~15 min read
Shopify sports nutrition schema for AI agents: NSF Certified for Sport vs Informed Sport, protein per serving, and DSHEA compliance
78% of supplement Shopify stores have no third-party testing certification in their JSON-LD — AI agents applying conservative health exclusions drop them from every athlete's recommendation. The complete guide: NSF Certified for Sport vs Informed Sport vs Informed Choice disambiguation (NSF International vs LGC Group, 280+ vs WADA-only substance lists, batch-test vs spot-test cadence, MLB/NFL/NBA/NHL/NCAA vs FIFA/IOC/UFC/IPF acceptance), protein per serving + serving size pairing (the denominator that determines protein density: 25g from 30g scoop = 83% isolate-grade vs 25g from 40g scoop = 63% concentrate-grade), leucine threshold (~2.5g per serving) as a machine-readable muscle protein synthesis filter, BCAA breakdown and complete protein encoding, DSHEA
legalDisclaimertext requirement, supplement form and sweetener type controlled vocabulary, proprietary blend disclosure, complete JSON-LD for a whey protein isolate, and a Dawn Liquid snippet forsupp.*metafields. -
2026-06-15~14 min read
Shopify allergen and dietary schema for AI agents:
suitableForDiet, FDA FALCPA 9 vs EU 14 allergens, and cross-contamination encoding83% of Shopify food and supplement stores have no allergen data in their product JSON-LD. AI agents can't answer "is this peanut-free?" without it — and they apply conservative exclusion rules when allergen status is unknown. The complete guide:
suitableForDietenum reference (all 11 officialRestrictedDietvalues and when to use each), FDA FALCPA 9 vs EU Annex II 14 allergen comparison table (the 5 EU-only allergens most US merchants miss: celery, mustard, lupin, molluscs, sulphites), cross-contamination vs intentional ingredient encoding (the critical 'contains' / 'free-from' / 'may-contain' three-tier distinction), food safety certification markup viahasCertification(GFCO, NSF Gluten-Free, OU Kosher, IFANCA Halal — why certification outweighs self-declaration for AI agents), complete JSON-LD for a GFCO-certified protein powder, and a Dawn Liquid snippet for allergen metafields. -
2026-06-15~12 min read
Shopify specialty coffee roast date schema for AI agents: freshness window encoding, SCA cupping score, and the days-since-roast Liquid calculation
78% of specialty coffee Shopify stores have no
roastDatein their product JSON-LD. AI agents can't answer "is this coffee fresh?" without it — and "fresh roasted coffee" is a high-intent query category. The complete guide:roastDatevsbestBeforeDatevsexpirationDatedisambiguation (the three date attributes merchants confuse and why only one answers freshness queries), freshness windows by brewing method (filter 4–14 days, espresso 7–21 days, pre-ground 0–7 days), processing method × roast date interaction (Natural degasses slower — 10-day espresso rest vs 7 days for Washed), SCA cupping score as a permanence signal (80+ = specialty grade, doesn't change with time), dynamic days-since-roast Liquid calculation using Unix epoch arithmetic ('now' | date: '%s'), freshness status labels mapped to days elapsed, complete JSON-LD for a single-origin Ethiopian Yirgacheffe Natural, and a fullcoffee.*metafield reference table. -
2026-06-14~12 min read
Shopify camera lens schema for AI shopping agents:
isAccessoryOrSparePartFor, lens mount lock-in, and crop factor compatibility89% of Shopify camera lens listings have no
isAccessoryOrSparePartFor— AI agents can't answer "does this lens fit my Sony A7 IV?" from JSON-LD alone. Mount strings in product titles are ambiguous: "Sony E-Mount" and "Sony FE" are the same physical mount, but "E" vs "FE" determines whether the lens covers a full-frame sensor or only APS-C. The complete guide: 8-mount reference table (Canon RF, Nikon Z, Sony E/FE, Fujifilm X, L-Mount, MFT, Canon EF, Nikon F legacy) with flange distances and format scope; crop factor image circle math (why an APS-C 28.4mm circle can't cover a 43.3mm FF sensor diagonal); the 3 APS-C vs full-frame mounting scenarios with compatibility outcomes; third-party lens product line code disambiguation (Sigma DG DN Art, Tamron Di III VXD G2); adapter compatibility for Canon EF-to-RF and Nikon F-to-Z (including the AF motor caveat); complete JSON-LD for a Sony FE zoom lens; and a Dawn Liquid snippet forlens.*metafields. -
2026-06-14~11 min read
Shopify smart home schema for AI agents: Matter, Zigbee, Z-Wave protocol disambiguation and hub dependency encoding
Protocol invisibility is the #1 reason smart home products fail AI agent recommendations — 94% of Shopify smart home products have no protocol version in JSON-LD. AI agents cannot confirm "works with SmartThings?" without protocol version, hub dependency chain, and Z-Wave region in structured data. Hub compatibility matrix (Zigbee, Z-Wave, Matter, Thread, WiFi), Z-Wave US/EU region lock (908.42 MHz vs 868.42 MHz — cross-border disaster signal), Matter 1.2 "no hub required" purchase signal, Thread vs Zigbee disambiguation (both 802.15.4, not interoperable), complete JSON-LD example for a Zigbee motion sensor, and a Dawn Liquid snippet using a
smarthome.*metafield namespace. -
2026-06-14~16 min read
Shopify power tool battery platform compatibility schema:
compatibleWith, 20V MAX disambiguation, and ecosystem lock-in for AI agents89% of Shopify power tool stores have no
compatibleWithbattery platform data in product JSON-LD. AI shopping agents answering "what DeWalt 20V MAX drills do you carry?" cannot confirm battery compatibility from structured data. The complete guide:compatibleWithcross-references with canonical platform@idURIs, 20V MAX vs 18V nominal voltage disambiguation, bare-tool vs kithasPartstructure, FLEXVOLT dual-platform batteries (twocompatibleWithreferences), Makita XGT backward-compatibility cliff, UL Listed with specific standard number incertificationIdentification, and a Dawn Liquid snippet using atool.*metafield namespace. -
2026-06-13~18 min read
Shopify jewelry schema for AI shopping agents: GIA certificate data,
4Cs additionalProperty, and lab-grown disambiguation91% of Shopify jewelry stores have no gemstone quality data in their product structured data. AI shopping agents filter diamonds by cut grade and carat from JSON-LD — not your product descriptions. GIA
hasCertificationwith report number and verification URL, 4Cs asadditionalPropertyusing exact GIA vocabulary (Excellent/VG/G for cut, D–Z color scale, FL–SI2 clarity, decimal carat weight), lab-grown CVD vs HPHT growth method disambiguation, metal fineness and hallmark stamp markup, ring sizeProductGroupwithSizeSpecification, Kimberley Process conflict-free certification, and a complete Dawn Liquid snippet using ajewelry.*metafield namespace. -
2026-06-13~15 min read
Shopify return policy schema for AI shopping agents:
MerchantReturnPolicy,hasMerchantReturnPolicy, and conditional final-sale overrides73% of Shopify stores have no
MerchantReturnPolicyin their product structured data. AI shopping agents rank return window as a top-3 purchase decision factor — withouthasMerchantReturnPolicyon yourOffer, the agent assumes worst-case and ranks competitors who have it. The two-layer architecture: site-wide policy onOrganizationplus per-producthasMerchantReturnPolicyon everyOffer. Conditional final-sale detection via Shopify product tags, reduced return windows for clearance items viacompare_at_price, per-product overrides via metafields, multiplereturnMethodvalues, 5 common mistakes, and a complete Dawn Liquid snippet in a single snippet file. -
2026-06-13~16 min read
Shopify brand entity markup for AI agents:
OrganizationsameAs, Knowledge Graph disambiguation, and thebrandproperty gap89% of Shopify stores output a bare
brand.namestring with no entity links in their Product JSON-LD. AI shopping agents resolve brand identity fromsameAslinks — Wikidata, Wikipedia, LinkedIn, social profiles — not from string matching. WithoutsameAs, your brand is anonymous to the agent. The fix: a site-levelOrganizationentity with@idandsameAsintheme.liquid, the@idlinking pattern for product-levelbrandproperty, multi-brand boutique metafield architecture, which sameAs sources carry the most entity resolution weight, and a complete Dawn Liquid snippet for both DTC and multi-brand stores. -
2026-06-12~17 min read
Shopify clothing size schema for AI shopping agents:
SizeSpecification,ProductGroupvariesBy, and the variant-title trap96% of Shopify apparel stores have no
SizeSpecificationin product JSON-LD. AI shopping agents filter by size from structured data — not from variant title strings. A variant namedM / Forest Greenis opaque to ChatGPT Shopping's size filter; aSizeSpecificationwithsizeSystem: SizeSystemUSandsizeGroup: WomenClothingis not. Complete guide:SizeSpecificationanatomy, allsizeSystemandsizeGroupenumeration values,ProductGroupwithvariesByfor multi-variant apparel, how to makesizeGroupdynamic using product metafields, footwear and accessories extensions, 5 common mistakes, and a full Dawn Liquid snippet for per-variantSizeSpecification. -
2026-06-12~16 min read
Shopify Product Bundle Schema: Why AI Shopping Agents Can't Price Your Bundles (and the
hasPartFix)91% of Shopify bundle products have no
hasPartin their Product JSON-LD. A $299 bundle containing $480 of components looks identical to a regular $299 product to ChatGPT Shopping, Perplexity, and Google AI Mode — they have no structured data to compute or communicate the bundle value. The full schema.org vocabulary for bundles (hasPartvs.isRelatedTovs.isAccessoryOrSparePartFor), how to signal bundle savings usingListPrice PriceSpecificationwithout triggering the sale-signal, whycompare_at_priceis semantically wrong for bundle discounts, a complete Dawn Liquid snippet driven bybundle.componentsmetafields, and the 5 mistakes that make AI agents undervalue or misrepresent your bundles. -
2026-06-11~15 min read
Shopify
priceValidUntiland sale pricing schema: why AI shopping agents show stale prices (and how to fix it)Your flash sale ended two weeks ago. ChatGPT Shopping is still quoting the sale price. This is not a ChatGPT bug — it's a missing
priceValidUntilproperty in your Product JSON-LD, and 84% of Shopify stores on sale have the same gap. The four Shopify pricing scenarios and their correct schema,PriceSpecificationwithSalePricevs.ListPricefor compare-at price markup, a complete Liquid snippet with metafield-driven dynamic expiry, the 30-day stale-cache window explained, and the 5 mistakes (addingpriceValidUntilto non-sale products, misusinghighPricefor compare-at, expired dates left in theme code) that break pricing signals for AI agents. -
2026-06-11~16 min read
Shopify shipping structured data for AI shopping agents:
OfferShippingDetails,ShippingDeliveryTime, and free-shipping threshold schema92% of Shopify stores have no
OfferShippingDetailsschema. When AI shopping agents answer "free shipping on X?" or "can I get it by Friday?", they read structured data — not your shipping policy page. Stores without shipping schema are excluded from shipping-filtered queries that convert at 2× the rate of unfiltered browse. Complete guide:OfferShippingDetailsanatomy, separating handling vs. transit time inShippingDeliveryTime, declaring a free-shipping threshold viaeligibleTransactionVolume, regional shipping withDefinedRegion, a full Liquid snippet for Dawn themes, and the 5 mistakes that silently break shipping schema. -
2026-06-11~14 min read
Shopify Open Graph tags for AI shopping agents: why
og:price:amountis a product discovery signal74% of Shopify stores are missing
og:price:amount. 81% are missingproduct:availability. Most merchants set OG tags once to fix a Facebook preview and never touch them again — but GPTBot, PerplexityBot, and Google-Extended parse these meta tags before they parse your Product JSON-LD. Dawn leaves four critical commerce-namespace OG properties unset. Why AI crawlers read OG before JSON-LD, the six properties that matter, what the variant price problem looks like in practice, a Liquid snippet that adds all four missing tags in under 5 minutes, and four mistakes that silently invalidate your OG output. -
2026-06-10~15 min read
Shopify product FAQ page schema for AI shopping agents: the pre-purchase query playbook
AI shopping agents answer pre-purchase questions — "is this BPA-free?", "does it fit a king mattress?", "what’s the warranty?" — from FAQPage JSON-LD on product pages, not from your description text. 97% of Shopify product pages have no FAQPage schema. The 5 question categories (materials/safety, compatibility, sizing, warranty, shipping) that cover 80% of pre-purchase AI queries, the metafield architecture for scalable FAQ across your catalog, a before/after water bottle example, @graph integration alongside Product JSON-LD, and the 5 FAQPage mistakes that silently break coverage.
-
2026-06-10~14 min read
Shopify local pickup and BOPIS structured data: how Buy Online Pick Up In Store signals reach AI shopping agents
Shopify configures local pickup inside the checkout flow — a UI that AI shopping agents never crawl. A store with six physical locations offering free same-day pickup looks identical to a drop-shipper in every AI shopping index. How to wire the two-entity architecture (OfferShippingDetails + LocalBusiness) so AI agents can answer "available for pickup today near me" queries, why deliveryTime in hours (not days) is the key to winning same-day queries, multi-location strategy with per-location geo coordinates, OpeningHoursSpecification for pickup vs. store hours, curbside and locker delivery variants, and the 5 BOPIS schema mistakes that keep local stores invisible in AI shopping results.
-
2026-06-06~15 min read
Shopify product availability states for AI shopping agents: InStock, PreOrder, BackOrder, LimitedAvailability, Discontinued
Availability state is the first filter AI shopping agents apply — before recommending a product, they verify it can actually be purchased. 69% of Shopify stores output InStock for OOS or backordered products. The 7 schema.org availability values, why Shopify’s default Liquid gets it wrong (product.available collapses 3 distinct states into a boolean), the master Liquid snippet for correct availability mapping across all inventory scenarios, per-state JSON-LD patterns for InStock/LimitedAvailability/PreOrder/BackOrder/Discontinued, multi-location inventory, and the 5 availability mistakes costing recommendation slots.
-
2026-06-06~14 min read
Shopify trust signals for AI shopping agents: return policy, warranty, certifications, and seller rating schema
AI agents are trust-first recommendation engines — they don't just pick the cheapest product, they recommend stores they can verify. 73% of Shopify stores have no MerchantReturnPolicy in JSON-LD. 91% of stores with a warranty page have no WarrantyPromise schema on products. The 4 trust signals (MerchantReturnPolicy, WarrantyPromise, hasCertification, Organization AggregateRating), how each one maps to post-purchase risk in the AI agent recommendation model, a complete Liquid snippet wiring all 4 trust signals, and the 5 common mistakes that undermine them.
-
2026-06-05~14 min read
How to audit Shopify structured data: the exact 3-tool workflow to verify your JSON-LD is working for AI shopping agents
67% of Shopify stores with JSON-LD have at least one critical parse error — but you'd never know it from the page rendering. The three-tool audit: Rich Results Test (per-product Google parse validity + required field presence), Schema.org Validator (full spec compliance, type correctness, deprecated properties), Google Search Console (catalog-wide error clustering across all crawled pages). Plus a curl workflow to verify crawler accessibility and a breakdown of 5 silent failure patterns: Liquid variables rendering as empty strings, bare availability strings instead of schema.org URIs, HTML entities in JSON-LD, price as currency string instead of number, and structured data present only on homepage not product pages.
-
2026-06-05~15 min read
Shopify Markets and international catalog readiness for AI shopping agents: why hreflang misconfiguration silences your store in every non-primary market
40%+ of $100k+ Shopify stores have at least one non-primary market configured. 83% of those stores have incomplete or missing hreflang on product pages. AI shopping agents — ChatGPT Shopping, Perplexity, Google AI Mode — can't execute JavaScript locale redirects or use Accept-Language headers, so hreflang is the only signal that routes them to your regional catalog. Five failure modes: missing x-default, homepage-only hreflang, language-only tags (en instead of en-GB), sitemap/head mismatch, and hreflang pointing to 301-redirecting URLs. JSON-LD priceCurrency fix using
cart.currency.iso_codevs.shop.currency, and a complete Liquid snippet for per-product-page hreflang generation across all configured markets. -
2026-06-05~13 min read
Shopify variant titles and option naming for AI shopping agents: how agents resolve color, size, and material queries
Product titles match the base entity. Variant option values decide which specific color, size, or material query converts. 71% of stores have at least one option dimension too vague to disambiguate variant queries. Five option naming failures (vague color values with no material context, non-standard option names, mixed size systems, name/value dimension mismatch, absent material options), schema.org option name vocabulary, per-variant Offer additionalProperty JSON-LD with Liquid snippet, and before/after rewrites for apparel, bedding, and beauty. Why bestsellers are the most important products to fix first.
-
2026-06-04~12 min read
Shopify product titles for AI shopping agents: why the first 70 characters decide everything
54% of Shopify stores have at least 10% of products with title issues. A weak title doesn't just miss one query — it degrades the return on every other catalog signal: GTIN, reviews, descriptions, shipping schema. Three failure modes (brand-noise front-loading, generic product type, marketing language substitution), the entity-first title formula, per-agent display truncation behavior, the JSON-LD consistency penalty, and before/after rewrites across apparel, home goods, and skincare.
-
2026-06-04~16 min read
ChatGPT Shopping for Shopify: a 30-day setup and audit playbook
Most Shopify stores enabled Agentic Storefronts and waited. Three months later, nothing. The reason isn’t the toggle — it’s a dozen compounding signal gaps: OAI-SearchBot blocked by Cloudflare, Product JSON-LD injected via JavaScript (invisible to crawlers), GTINs missing on 80% of the catalog, descriptions under 150 words, no GMC feed, no brand entity signals. This 30-day playbook closes every gap in sequence, with audit checkpoints, scan data behind each fix, and GA4 attribution setup so you can measure before day 30 whether the changes are landing.
-
2026-06-03~13 min read
Shopify product descriptions for AI shopping agents: the six signals ChatGPT, Perplexity, and Google AI Mode extract
AI shopping agents generate recommendations by extracting signals from your product description — and if your description is too short, too generic, or stripped of attributes, they quote a competitor instead. 68% of stores have at least one product under 100 words. The six signals that separate visible products from invisible ones, the 150-word threshold explained, per-agent behavior for ChatGPT vs. Perplexity vs. Google AI Mode, common Shopify description failure modes (Shopify Magic defaults, HTML clutter, variant specs missing from text), the JSON-LD description connection, and a before/after rewrite of a real product description.
-
2026-06-03~12 min read
Shopify AI shopping agent attribution: how to see exactly what ChatGPT, Perplexity, and Google AI Mode are sending you
GA4 has no AI shopping channel. ChatGPT sends traffic from chatgpt.com, Perplexity from perplexity.ai, Google AI Mode from google.com (indistinguishable from organic). All three land in the generic Referral bucket by default. This guide covers every AI agent referrer signature, the GA4 custom channel group rules (including the srsltid trick for Google AI Overview), server log grep commands for crawler vs. customer separation, Shopify Analytics limitations, and a 5-KPI weekly dashboard that closes the feedback loop between structured data fixes and measurable revenue.
-
2026-06-02~10 min read
Shopify gift cards and digital products: the AI shopping agent blind spots most stores ignore
Gift cards trigger GTIN validation errors in AI shopping agents. Digital downloads carry phantom shipping costs from Shopify’s default theme. Both failures are invisible in your admin but exclude these products from ChatGPT Shopping, Google AI Mode, and Perplexity entirely. Every failure mode documented — and the exact JSON-LD fixes for AggregateOffer denominations, DigitalDeliveryMethod suppression, and schema type routing (Book, SoftwareApplication, Course, MusicRecording) per product category.
-
2026-06-02~14 min read
Shopify subscription pricing and AI shopping agents: why they show the wrong price
Your subscribe-and-save price is typically 15–20% cheaper than the one-time price. AI shopping agents like ChatGPT Shopping and Perplexity almost never quote it — because Recharge, Bold, Stay.ai, and Skio all inject subscription pricing via JavaScript, which crawlers read after the initial HTML. Why the invisible discount problem happens mechanically, how each major agent handles (or doesn’t handle) subscription pricing, what the scan data shows, and the complete PriceSpecification JSON-LD fix with Liquid snippets for each app.
-
2026-06-01~11 min read
Shopify Agentic Storefronts: What Changed March 2026 and What Your Store Must Do Now
On March 24, 2026, Shopify enabled Agentic Storefronts by default for every eligible US merchant. Most store owners assumed it meant a chatbot — but the actual mechanism is a catalog API that lets ChatGPT Shopping, Perplexity, and Google AI Mode transact directly through your store. Here's what the toggle actually does, why 78% of stores still aren't visible to agents despite having it on, and the 12-point readiness checklist with expected score lifts for each fix.
-
2026-05-31~8 min read
ChatGPT Shopping, Perplexity, and Google AI Mode: How Each AI Agent Picks From Your Shopify Catalog
Three AI shopping agents now send measurable referral traffic to Shopify stores — and each one weights catalog signals differently. ChatGPT Shopping weights GTIN-based exact-match lookup; Perplexity Commerce weights AggregateRating and description richness; Google AI Mode extends your existing Merchant Center Shopping Graph. The side-by-side signal priority table, data from our 100-store scan, and the 3-pass cross-platform optimization sequence that covers all three without tripling your workload.
-
2026-05-31~13 min read
Headless Shopify: the four signals you silently lose when you leave the standard storefront
When you migrate to Hydrogen or build a custom Next.js storefront, Shopify quietly stops generating four things AI shopping agents depend on. The product feed at
/products.json, Product JSON-LD on every PDP, a compliantrobots.txt, and stable canonical URLs all disappear without a single build error. How each signal breaks, how to verify it with curl commands, and the exact code to restore it in Hydrogen and Next.js. -
2026-05-30~14 min read
Cloudflare for Shopify: the three settings that silently block AI shopping agents
Your robots.txt says
Allow: /for GPTBot and your Shopify theme is untouched — yet the CatalogScan robots-open signal fails at 0 points. The culprit is Cloudflare: specifically one of three settings that intercept AI shopping crawlers at the CDN layer before they reach your origin. The curl test that confirms Cloudflare is the problem, the three settings in order (Bot Fight Mode → AI Scrapers managed rule → custom WAF rules usingcf.client.bot), the safe Cloudflare config that keeps real bots out while keeping GPTBot, OAI-SearchBot, PerplexityBot, and ClaudeBot in, and the seven-step verification playbook. -
2026-05-01~12 min read
AggregateRating on Shopify: which review apps inject it correctly and which silently break it
9 in 10 Shopify stores in our launch scan fail AggregateRating in the JSON-LD even with visible stars on every PDP and a reviews app installed. The on-page stars and the schema-readable stars are two different things, and AI shopping agents only read the second. The per-app anatomy of what Judge.me, Yotpo, Loox, Stamped, and Okendo inject by default, where each toggle lives, the four shapes the schema can take (only one is correct), and the 30-day playbook for verifying your fix actually landed.
-
2026-04-30~10 min read
ProductGroup JSON-LD on Shopify: why 60% of stores leave 18 points on the table
ProductGroup JSON-LD is the single biggest score lever in our scan data — worth 30 points on a 100-point scale, more than any other signal. 60% of top DTC Shopify stores fail it entirely. The four shapes Shopify stores actually emit (only one is correct), why ChatGPT Shopping and Perplexity need ProductGroup specifically for variant queries, and a 12-line Liquid snippet that lifts the average store from 12/30 to 30/30.
-
2026-04-30~10 min read
Shopify metafields for AI shopping agents: which 8 actually move the score
Most Shopify stores have hundreds of metafields. AI shopping agents read maybe eight of them. The two namespaces that matter (
shopify.*andmm-google-shopping.*), the eight metafields that decide whether ChatGPT Shopping picks your variant or the same product on Amazon, what we keep finding broken in the wild, and how to fix coverage in an afternoon without a developer. -
2026-04-25~9 min read
Shopify GTIN requirements for AI shopping agents (2026 guide)
When Shopify turned Agentic Storefronts on by default on March 24, 2026, GTIN coverage stopped being a Google Shopping concern and became a hard ranking input for ChatGPT Shopping, Perplexity, and Google AI Mode. What is a GTIN, when does Shopify require one, and how do you fix coverage on your store without buying GTINs you don't need — with real data from a 100-store scan.
What's coming next
- Shopify power tool battery platform blog: why DeWalt 20V MAX and FlexVolt 60V share a connector (same battery, voltage switches internally based on tool handshake), why Makita LXT 18V and XGT 40V MAX are physically incompatible platforms despite the XGT adapter, and why Wh is the correct runtime comparison field across voltage tiers
- Shopify reef aquarium lighting blog: why PAR at 12-inch depth (µmol/m²/s) is the only valid coral intensity metric — lux weights 555nm green (human eye peak) and is irrelevant for coral photosynthesis driven by 420–490nm blue — and why a 90W Kessil can deliver more PAR than a 165W generic fixture
- Shopify cookware schema for AI agents: PTFE vs ceramic, the PFAS-free claim problem (PTFE is a PFAS compound — PFOA-free ≠ PFAS-free), induction compatibility as a structured data signal, and oven-safe temperature rating as a variant-level field
Or skip the reading and run a scan
90 seconds, no login. The report is the curriculum.
Run a free scan 100-store leaderboard