What Claude Looks for on a Product Page
A founder-voice guide to the on-page signals Claude reads for ecommerce shopping queries, where it differs from ChatGPT and Perplexity, and the fastest fixes.
If you sell a premium product on Shopify and Claude has been recommending competitors when buyers ask about your category, the gap is not Claude-specific. It is the on-page work that the broader AEO playbook covers, plus one Claude-specific signal: source diversity. Claude weighs being mentioned across multiple distinct sources more heavily than most assistants do. This guide covers what Claude reads on a product page, where it differs from ChatGPT and Perplexity, and the fastest fixes.
What does Claude look for on a product page?
Claude pulls product information from four signals: Schema.org Product, Offer, and AggregateRating in JSON-LD in the page head, a 2 to 3 sentence direct answer in the first 150 words of the body, third-party review consensus across multiple independent sites, and any tool surface the brand exposes through MCP. Claude differs from ChatGPT and Perplexity on one weighting: source diversity. A brand cited by Trustpilot, a category review site, and Reddit gets stronger signal than a brand cited 50 times on a single source. Brands that ship clean schema and a direct answer move to retrievable in days; brands with diverse third-party presence move to consistently cited within weeks.
The retrieve-vs-cite asymmetry that the 15% citation study documented for ChatGPT applies to Claude with similar mechanics. The page either has the answer in extractable form or it does not.
How Claude differs from ChatGPT and Perplexity
The on-page signals overlap heavily across all three assistants. The differences are in the discovery and tool layers.
Merchant feeds. ChatGPT Shopping has the OpenAI Commerce file-upload spec, which gives the brand a structural advantage in retrieval. Claude does not have an equivalent public feed standard. Brands that have invested heavily in the OpenAI Commerce feed should not expect the same lift on Claude; Claude relies on retrieving live pages and any public structured data.
Tool use. Claude supports tool calls through the Model Context Protocol. Brands that ship an MCP server let Claude make live calls during a buyer's conversation: list products, fetch one product, submit a checkout intent. The MCP server setup guide covers the minimum viable build. ChatGPT also supports MCP through ChatGPT Apps, but the integration paths differ.
Citations. Claude has a native citations feature that lets the model attribute claims to specific source documents. For ecommerce, this means Claude is biased toward sources where it can point at a specific page or paragraph. Pages with clear, extractable structure get cited more reliably than pages with a wall of marketing copy.
Source diversity. Claude weighs cross-source consensus more than the other assistants. Five Trustpilot reviews plus a Reddit thread plus a category review article carries more weight than 500 in-house reviews on the brand's own site. Brands that have invested in third-party presence rather than in-house volume tend to do well on Claude.
The four on-page signals in plain terms
Signal one: Schema.org Product, Offer, AggregateRating. Server-rendered JSON-LD in the document head, not injected by JavaScript. Same fields the broader AEO playbook calls for. The product schema markup guide covers exactly which fields to ship.
Signal two: Direct answer in the first 150 words. The opening 2 to 3 sentences of the body should answer the buyer's question in plain language. The model extracts these sentences when composing a citation.
Signal three: Third-party review consensus, weighted toward source diversity. Trustpilot, Reddit, G2 for B2B, and category review sites. Claude trusts the brand more when independent voices describe it the same way across multiple distinct sources. A single high-volume review source is weaker signal than three moderate-volume sources.
Signal four: MCP tool surface (optional but powerful). A small HTTP server that exposes products and checkout to MCP-compatible clients. Brands shipping this give Claude live data access during the conversation, which lifts citation reliability for time-sensitive buyer questions ("is this in stock," "ships to Norway by Friday").
What does not move Claude citation
A few patterns brands try that produce little lift specifically on Claude:
- Backlink campaigns. Domain authority is a Google ranking signal. Claude does not weigh inbound links for product queries the way Google does. The fix is on-page mechanics and third-party review presence.
- Long-form blog content with no third-party signal. Blog posts can get retrieved on long-tail queries, but for the core "best in category" question Claude still pulls from PDPs and review aggregators. Fix the PDP first.
- High-volume in-house reviews. A brand with 5,000 reviews on its own site gets less Claude signal than a brand with 100 reviews on Trustpilot plus a Reddit thread plus a Wirecutter mention. Source diversity outranks volume.
- OpenAI Commerce feed alone. The feed helps with ChatGPT Shopping discovery; it does little for Claude. Brands optimizing only for the feed miss Claude entirely.
How to ship a Claude-friendly PDP in 4 days
If your brand is invisible on Claude and you want to fix it without a full rebuild, run this 4-day plan.
Day one: Ship clean Schema.org Product markup on every PDP. Validate every PDP against Google's Rich Results Test. Confirm Product, Offer, and AggregateRating come back without errors. Fix the template, not individual pages.
Day two: Rewrite the first 150 words of every PDP. Open with a 2 to 3 sentence direct answer to the buyer's question. Move the brand origin story below the spec table. Move the long marketing copy below that. Claude extracts from the top of the body.
Day three: Audit third-party review presence with diversity in mind. Check Trustpilot, Reddit, and any 2 or 3 category review sites. The goal is presence on at least 3 distinct sources, not volume on any one. Add a post-purchase email asking for a Trustpilot review; engage on Reddit threads as the brand once a month with helpful answers.
Day four: Stand up a minimum MCP server. One tool: list_products with filters. Test from Claude Desktop. The build fits in an afternoon for a backend developer comfortable with the Shopify Storefront API. The MCP server guide covers the full setup.
These four days move a brand from invisible to consistently retrieved on Claude. Citation rate ramps up over the following weeks as third-party signal compounds.
How Claude shopping queries actually look
A buyer asks Claude "what is the best leather backpack for daily commute under $400." Claude does not return a Google-style results list. It composes a paragraph that names 2 to 4 brands with a sentence each, citing the sources behind each claim.
The brands that appear share a profile:
- A PDP with clean Product, Offer, AggregateRating in JSON-LD.
- A 2 to 3 sentence direct answer in the first 150 words.
- Mentions across at least 2 to 3 distinct third-party sources.
- An MCP server (optional, but lifts the live-data accuracy).
Brands missing two or more of those rarely appear in the response. The structural floor is the same as ChatGPT and Perplexity; the citation weighting is what separates Claude.
Where to start
If your brand is invisible on Claude, the first three days of the plan above fix the floor. The AEO playbook for ecommerce covers the broader signal set across ChatGPT, Claude, and Perplexity. How ChatGPT Shopping picks products covers the ChatGPT-specific mechanics. What Perplexity Shopping looks for covers Perplexity.
If your brand has the basics shipped and Claude still picks competitors, the gap is usually source diversity. The post-purchase email asking for a third-party review and the monthly Reddit engagement are the highest-impact moves.
The buyer is asking the assistant. The model fetches the PDP, reads the structured data, weighs the third-party signal, and composes the answer. The brand whose page is extractable across multiple sources is the brand cited.