Shopify product description richness for AI shopping agents

Product descriptions are the primary text signal AI shopping agents use to match a user's free-text query — "running shoe for someone allergic to wool," "cookware safe for gas stoves," "vitamin C serum for sensitive skin" — to your catalog. The signal is direct and measurable: the median word count of body_html across your sampled products. Below 40 words median, the agent has so little signal that it can only match on the title and falls back to your competitors with longer copy. Above 80 words, your products surface for queries that mention attributes the title doesn't. The 8 ranking-spread points sit at "did you write a paragraph or a sentence per product."

Last updated 2026-04-30 · Ranking-spread signal · 8 pts

What it is

Every Shopify product has a body_html field — the rich-text body of the product description as edited in the admin's "Description" textarea. Stored as HTML; rendered into the PDP body and into the Product JSON-LD's description field. CatalogScan scores the stripped-text word count of this field — what an AI agent actually reads after stripping markup.

"Cotton crewneck tee."

"Soft, breathable, made to last. The everyday tee you've been looking for."

"180 GSM heavyweight cotton crewneck tee, ring-spun for softness. Single-stitched at the hem and sleeve cuffs. Fits true to size with a regular cut through the chest and torso..."

Why AI shopping agents care

• Long-tail query coverage. Short descriptions only match on the product title. Rich descriptions match on materials, fit, use cases, who it's for, what it's compatible with — every attribute mentioned in the body becomes a potential query match. A 100-word description matches roughly 5x more long-tail queries than a 20-word description, and long-tail is where AI shopping is concentrated (specific natural-language queries, not "buy shoes").
• Embedding density. AI rankers use vector embeddings of your description text. Short text has thin embeddings — small distance between any two products of the same category. Rich text gives each product a distinctive embedding that places it correctly in the semantic neighborhood. The gap between "this is a tee" and "this is a 180 GSM heavyweight ring-spun tee with single-stitched hem" is the difference between matching one query class and matching ten.
• Trust priming. AI rankers note description depth as a "well-tended catalog" signal. A store with thin descriptions reads as low-effort or recently-launched; a store with consistent 80-150 word descriptions reads as established, likely to deliver, likely to have accurate inventory. The signal compounds with alt-text coverage and review counts.
• Citation quality. When ChatGPT cites your product in an answer, it pulls a sentence from your description verbatim. With rich descriptions, the agent has a real sentence to quote ("180 GSM heavyweight cotton, single-stitched hem"). With thin descriptions, it has to paraphrase or omit — and an agent that can't quote you tends to surface a different store that gave it something to say.

How to test it on your store

One curl + jq + awk gives you the median across the first 30 products:

curl -s https://yourstore.com/products.json?limit=30 \
  | jq -r '.products[].body_html' \
  | sed 's/<[^>]*>//g; s/&[a-z]*;/ /g' \
  | awk '{print NF}' \
  | sort -n \
  | awk '{a[NR]=$1} END {print "median:", a[int(NR/2)]}'

One number out: the median word count. Anything ≥80 = full credit; 40-79 = half; under 40 = zero. The free CatalogScan scan runs the same calculation across your live feed and reports the median, the 25th percentile (so you can see the worst offenders), and a flagged list of products under 40 words.

Recipe: what 80 useful words look like

The point isn't word count for its own sake — keyword stuffing is a downweight on every modern ranker. The point is 80 words of distinct attributes:

1. Material/specs (15-25 words). Fabric weight, composition, key components. "180 GSM 100% organic ring-spun cotton, double-needle stitched seams, pre-shrunk."
2. Fit / dimensions / sizing reference (15-20 words). "Regular fit through the chest and torso. Model is 5'10" and wears size M."
3. Use case (15-20 words). Who it's for, when to wear/use it, what it's compatible with. "Designed for daily wear and casual layering. Pairs with our chinos and joggers."
4. Differentiator (10-20 words). What makes this specific product different from category competitors. "Hand-loomed in Portugal at a family-run mill we've worked with for six years."
5. Care / specs / warranty (10-20 words). Wash care, returns, materials sourcing certifications. "Machine wash cold; tumble dry low. Two-year stitching warranty."

Five blocks × 15 words each = 75 words. Add a sentence about provenance or sustainability and you're at 80-95. Each block answers a different class of query.

How to fix it

5 mistakes we keep finding