Great products.
Make them discoverable.

Search and generative AI share SEO foundations. Product snippets and merchant listings have different markup rules. Merchant Center provides another product-data route.

Access. Googlebot permission, indexing, and snippet controls matter. Google also documents a Search Console generative AI inclusion setting; a page scan cannot read that setting.

Commerce. A zero price can support product snippets, but merchant listing markup requires a positive price and a direct Offer. A structured-data pass is not Merchant Center acceptance.

Verify. Verify the inherited Search Console AI control, actual indexing, Merchant Center diagnostics, and page/feed agreement. Account settings are unverified here.

Measure. Use Search Console’s Generative AI performance report and product enhancement reports, then measure store conversions separately.

Bing’s index supports search, Copilot, and grounding. Clear entities, accurate markup, stable URLs, useful content, and timely updates support those experiences.

Access. noindex affects indexing. noarchive prevents Chat/Copilot linking; nocache limits use to URL, title, and snippet. When both occur, Bing documents nocache precedence. These controls do not all remove ordinary search results.

Commerce. Microsoft Merchant Center and its product catalog are separate from a public page scan. IndexNow helps notify participating engines of changes; it does not guarantee indexing.

Verify. Verify Bing Webmaster Tools, Merchant Center item status, real Bingbot requests, and IndexNow delivery. A missing visible IndexNow script does not establish missing integration.

Measure. Use Bing Webmaster Tools AI Performance for citation observations, plus referral and conversion data. A citation count is not a revenue or ranking guarantee.

Web search and approved merchant product feeds are separate discovery routes. Readable product facts help the scanner assess preparation; the recommendation formula is not public.

Access. OAI-SearchBot controls search crawling. GPTBot is a separate training preference. ChatGPT-User is user-initiated retrieval, not a search-inclusion gate.

Commerce. Review feed approval, exact item/variant identity, ingestion, freshness, and supported markets. Shopify Catalog can deliver eligible products to enabled agentic storefronts even when web crawlers are blocked. A page scan does not verify that delivery.

Verify. Verify real crawler access and accepted, current feed records. The page-to-feed field map is a set of candidates, never proof of ingestion.

Measure. Track accepted feed records, referred visits, and conversions. Record date, market, and conversation context when sampling shopping answers.

PerplexityBot gathers pages for search results and links. It is not a foundation-model training crawler. Perplexity also offers a merchant program.

Access. Allow PerplexityBot and verify its published IP ranges at the firewall. Perplexity-User handles user requests and generally ignores robots.txt; its robots rule is not scored as a search gate.

Commerce. Merchant onboarding, product data, and checkout are a separate path. We do not assume OpenAI’s feed fields are Perplexity’s ingestion contract.

Verify. Confirm your actual merchant integration and current specification with Perplexity. Check crawler logs, item freshness, and public product images.

Measure. Use available merchant reporting and store referral/conversion data. Sample real shopping questions without treating one answer as a universal ranking.

Claude-SearchBot supports web search. Claude-User fetches pages for user requests, with separate robots controls. ClaudeBot is a separate training crawler.

Access. A Claude-SearchBot block affects search indexing; a Claude-User block affects user-requested retrieval. Both are checked separately here. Blocking ClaudeBot earns no penalty.

Commerce. Anthropic documents a blueprint for commerce agents connected to a merchant’s own catalog tools. That is different from submitting a universal public Claude shopping feed; no such feed contract was established in the reviewed documentation.

Verify. Verify both search and user-fetch access. For a Claude-powered store agent, ground product claims in catalog tools and validate writes and checkout in the application.

Measure. Measure actual referred visits and conversions. For an embedded commerce agent, evaluate catalog accuracy, successful tasks, and confirmed checkout outcomes.

Search discovery and completing a shopping task are different. Browser agents can use HTML, the accessibility tree, and screenshots. Give buttons clear names, label variant controls, preserve selected state, and keep price and stock visible.

For catalog tools, return stable product IDs and current facts. Prepare edits and checkout as explicit actions with reviewable outcomes. Agent protocols and checkout integrations are not universal SEO prerequisites.