Claude Shopping Visibility: What Merchants Can and Cannot Control

What you can control

Public product evidence

Publish exact identity, attributes, limitations, price, stock, policies, and supporting documentation on stable canonical URLs.

Access policy

Anthropic documents separate crawler identities for search, user-directed retrieval, and model-development purposes. A merchant can set intentional robots rules for public and private paths.

Consistency

Keep visible content, Product JSON-LD, variants, and feeds aligned. Claude may use public sources even when it does not consume a merchant feed.

Your own application

If a retailer builds with the Claude API, it can control the catalog source, retrieval rules, prompts, tools, and user experience within that application.

What you cannot control

• whether public Claude retrieves a particular page
• which sources it selects
• the final recommendation
• answer wording
• personalization and platform experiments
• the timing of search refresh
• every user’s location or mode

No robots rule or llms.txt file can force a citation.

What to measure

Use fixed buyer questions and record actual:

• brand and product mentions
• canonical-domain citations
• cited third parties
• fact accuracy
• named competitors
• provider failures

Keep crawler-access tests separate. A 200 response to a spoofable user-agent string is not proof of a Claude visit.

The merchant action

Choose five products that require comparison. Publish the decisive attributes and limitations, link supporting first-party documents, and verify stable URLs. Then run dated Claude web-search observations.

If a product is absent, investigate the evidence and access path. Do not claim suppression or promise placement without evidence.

Read the full Claude and Anthropic crawler guide for the technical controls.