Product Data That AI Agents Can Actually Read

We have audited thousands of e-commerce stores. The pattern is always the same: merchants spend months perfecting their website design, then wonder why AI shopping agents skip right past them. The reason is almost never the product. It is the data. AI agents do not see your beautiful storefront. They see structured data, feeds, and metadata. If that data is incomplete, inconsistent, or missing, you are invisible.

What AI Agents Actually See

When an AI agent visits your store, it does not browse like a human. It does not admire your hero banner, scroll through lifestyle photography, or read your brand story. It parses machine-readable data: JSON-LD structured data, meta tags, product feed submissions, and raw page text. If the information is not in one of those formats, it effectively does not exist.

The complication is that each major AI shopping platform consumes different data sources and weights them differently. Optimizing for one platform while ignoring the others leaves money on the table. Here is what each platform actually reads:

Google AI Mode

Google AI Mode synthesizes product comparisons directly in conversational search results. It pulls data from three sources: your Google Merchant Center feed (highest priority), on-page structured data (JSON-LD Product schema), and crawled page content. Google's Shopping Graph merges all three into a unified product entity. If your Merchant Center feed says one price and your on-page schema says another, Google flags the discrepancy and may suppress the listing entirely.

ChatGPT Shopping

ChatGPT Shopping reads two sources. OAI-SearchBot crawls your product pages, and merchants in OpenAI's merchant program can also submit a product feed as JSONL, CSV or TSV. The feed spec requires nine fields: item_id, title, description, url, brand, seller_name, image_url, availability and price. gtin is optional, but it is how ChatGPT matches your product to the same item sold elsewhere. OpenAI asks merchants to update the feed whenever a price or availability changes. Without a feed your products can still surface through search, which is why crawlable pages and clean Product schema still matter here.

Perplexity

Perplexity crawls pages directly and also accepts submitted feeds. Its source attribution model shows users exactly which merchants provided the data, creating a transparency layer that other platforms lack. Perplexity places heavy weight on on-page structured data and well-formatted product pages. Clean JSON-LD markup gives you a significant advantage here because Perplexity can extract and cite your product data with confidence.

What small stores are usually missing

Product data enrichment is not enterprise Master Data Management. You do not need a six-figure PIM platform or a dedicated data ops team. For a store with 50 to 500 products, enrichment means one thing: filling in the gaps that AI agents need to recommend your products.

When we run Agent Ready audits, we consistently find the same gaps across small and mid-size stores. These are the data holes that prevent AI agents from confidently recommending your products:

  • Missing GTINs — The most common gap we see. Without a GTIN, AI agents cannot verify your product or match it against other retailers. Our GTIN guide covers where to find them.
  • Thin descriptions — One sentence instead of use-case context. AI agents need to understand what the product is, who it is for, and when someone would use it.
  • Absent attributes — No material, weight, dimensions, or color specified. These are the structured data points agents use to compare products.
  • No review data — Or reviews locked inside JavaScript widgets that AI crawlers cannot read.
  • Wrong or shallow categories — Everything filed under "Products" or "Shop" instead of mapped to the deepest relevant taxonomy node.
  • Low-quality images — A single photo, low resolution, no white background. Multimodal AI agents analyze images as part of product evaluation.

Think of it this way: your product data has holes. Every hole is a question an AI agent cannot answer. Every unanswered question makes the agent less likely to recommend your product. Enrichment fills those holes.

Is Your Product Data AI-Readable?

Run a free audit to see exactly how AI agents read your store. We check structured data, product feeds, identifiers, and 20+ other signals that determine whether AI shopping agents trust and recommend your products.

Get Your Free Score →

What changed in AI shopping, every two weeks

Agent Ready HQ on Substack covers ChatGPT Shopping, Google AI Mode, and the UCP and ACP protocols in plain terms for store owners. Each issue ends with one fix you can make to your store that week.

Sign up on Substack

Free. Opens Substack in a new tab.