The product data rules AI shopping surfaces actually publish
Last reviewed
Short answer
OpenAI's product feed specification sets a hard maximum of 150 characters on product titles and instructs merchants to avoid all-caps and use plain text only. Required fields cover title, description, price, availability and image. Most catalog problems are failures against these documented basics, not anything exotic.
The published limits
OpenAI's commerce documentation includes a product feed spec that states the constraints directly. The title field is required, capped at 150 characters, and carries the instruction to avoid all-caps. Fields are plain text. This is unusually concrete for a space full of speculation, and it is worth optimising against precisely because it is written down.
Note what the spec does not say. It contains no instruction about keyword stuffing. Avoiding stuffed titles is sound practice for shoppers and for classic search, but do not let anyone tell you OpenAI's spec forbids it. It does not mention it at all.
| Field | Requirement | Why it matters |
|---|---|---|
| Title | Required, max 150 characters, avoid all-caps | The main string an assistant matches a shopper question against |
| Description | Required, plain text | The text an assistant can quote when explaining the product |
| Price and availability | Required | Assistants filter on both before recommending |
| Image | Required | Product cards render from it |
Where real catalogs fail
In practice the failures are boring. A one or two word title with no attributes. An empty description. A title that never names the brand, so nothing attributes the product to you. A product type set in Shopify but missing from the title. Promotional filler like "BUY NOW" that burns characters and tells a shopper nothing.
None of these are subtle. They are the kind of thing you can grade deterministically, which is exactly why a readiness score can be free and instant rather than requiring an AI call per product.
How to check your own catalog
You can do this by hand. Export your products, check each title against the length and completeness rules, look for empty descriptions, and note anything that never names the brand. On a catalog of 30 products that is an afternoon. On 500 it is not realistic.
Entitled automates exactly that check. It grades every product against these rules, tells you which rule failed and why, and sorts by which fixes move your score most. The score is free forever. See it running on a sample catalog.
Frequently asked questions
What is the maximum product title length for AI shopping?
OpenAI's product feed spec sets a maximum of 150 characters for the title field and instructs merchants to avoid all-caps. That is considerably longer than the roughly 60 characters Google shows in a search result, so a title tuned only for Google is usually leaving room unused.
Does OpenAI's spec ban keyword stuffing?
No. The spec addresses length, all-caps and plain text formatting. It does not mention keyword stuffing. Stuffed titles are still bad practice for shoppers and for classic search, but the prohibition is not in the spec.
Do these rules apply to Perplexity and other assistants too?
Each surface publishes its own requirements. Perplexity runs a separate merchant program with its own catalog submission. The overlap is large though: complete, accurate, attribute-rich product data satisfies all of them, which is why it is worth fixing once at the source.
Sources
Keep reading
Score your own catalog
Entitled grades every product against the rules on this page, explains each failure, and fixes them with AI you approve. The score is free forever.