Great products.
Make them discoverable.
See what search and AI can read.
Find the gaps. Get your next useful fix.
Your URL and HTML are sent to the scanner for analysis. When enabled, the consistency review sends extracted product facts and limited visible excerpts to OpenAI.
Google
Bing
ChatGPT
Perplexity
ClaudeEveryday Tote
fieldwork.example/products/everyday-tote
Choose a provider to explore findings.
Scores reflect observed page checks. They do not predict rankings or recommendations.Start with these findings.
This blocks the web crawl route. It does not establish whether a product appears through a catalog or merchant feed.
Fix neededOAI-SearchBot permission25 pts
OAI-SearchBot: disallowed for /products/everyday-tote by robots.txt.
Next step. Allow OAI-SearchBot on the intended product URL. Verify actual crawler requests against OpenAI’s published IP ranges in your CDN logs.
Source guidanceFix neededSeller of this offer3 pts
The selected Offer does not identify its seller.
Next step. Add the merchant supplying this Offer. Brand, platform, and seller are not interchangeable.
Source guidanceAll ChatGPT evidence 17 checks
Fix neededOAI-SearchBot permission25 pts
OAI-SearchBot: disallowed for /products/everyday-tote by robots.txt.
Next step. Allow OAI-SearchBot on the intended product URL. Verify actual crawler requests against OpenAI’s published IP ranges in your CDN logs.
Source guidanceObservedPage indexing restrictions10 pts
No generic or OAI-SearchBot noindex/none directive found in the fetched page or headers. Actual index inclusion remains unverified.
Next step. Review generic or OAI-SearchBot noindex/none directives if discovery is intended. This is a conservative web-search check; it does not inspect merchant feeds.
Source guidanceObservedExtractable product name6 pts
Everyday Tote
Next step. Publish a clear product name in readable page content and Product.name. Product schema is our extraction method, not a special ChatGPT requirement.
Source guidanceObservedStable item identifier5 pts
FW-TOTE-OLV
Next step. Use a real, stable item identifier. Confirm uniqueness across your catalog and variants; a page scan cannot establish it.
Source guidanceObservedBrand identity3 pts
Fieldwork
Next step. Supply the actual brand when applicable. Do not add a placeholder just to improve the score.
Source guidanceFix neededSeller of this offer3 pts
The selected Offer does not identify its seller.
Next step. Add the merchant supplying this Offer. Brand, platform, and seller are not interchangeable.
Source guidanceObservedProduct and offer selection5 pts
One Product and at most one direct offer detected; variant completeness is not verified.
Next step. Scan a URL for the selected variant and give each item stable identifiers, specific images, price, and stock. Review multiple sellers separately.
Source guidanceObservedExtractable product description6 pts
A durable everyday tote made from recycled canvas.
Next step. Provide factual features, use cases, dimensions, and compatibility where relevant. This check verifies presence, not copy quality.
Source guidanceObservedProduct image URL5 pts
https://fieldwork.example/images/tote.jpg — URL syntax only; image bytes and crawler access were not fetched.
Next step. Use a publicly accessible product image showing the selected variant. Verify the image response and accuracy manually.
Source guidanceObservedName in initial page text4 pts
Checks whether the extracted name also appears outside scripts, metadata, and obvious hidden elements. CSS and JavaScript are not rendered.
Next step. Render the product name in the initial page body. Review the match if the schema uses different wording.
Source guidanceObservedDescription in initial page text4 pts
Checks a literal normalized text match between Product.description and initial body content. Equivalent wording can produce a warning.
Next step. Keep the structured description aligned with the product copy shoppers can read.
Source guidanceObservedSpecific offer price10 pts
Offer · price 48. Price accuracy over time is unverified.
Next step. Use the actual nonnegative price of the selected purchasable item. An AggregateOffer range does not establish one item’s payable price.
Source guidanceObservedOffer currency4 pts
USD — format checked; feed currency support unverified.
Next step. Use the actual ISO 4217 currency alongside the price and confirm support for your feed’s market.
Source guidanceObservedExplicit stock state5 pts
InStock
Next step. Publish the current availability. Out-of-stock items receive the same clarity credit as in-stock items; never misstate inventory.
Source guidanceObservedStable canonical page5 pts
https://fieldwork.example/products/everyday-tote
Next step. Choose one canonical URL. Different variant canonicals can be intentional; confirm which product page should be indexed.
Source guidanceManual reviewReview evidenceUnscored
Rating 4.8 / 5 · count 124 · minimum 1
Next step. Reviews are optional. Verify genuine product feedback, the rating scale, and matching counts. Schema alone does not prove authenticity.
Source guidanceManual reviewShipping and returnsUnscored
Policy availability and fulfillment accuracy require review outside this page-data score.
Next step. Publish applicable costs, delivery expectations, and return terms. Confirm feed policy fields with your integration; digital items may not ship.
Source guidanceWhat this assessment covers Method & limits
100% of weighted page checks assessed.
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.
Verify real crawler access and accepted, current feed records. The page-to-feed field map is a set of candidates, never proof of ingestion.
65 shared product-data points + OAI-SearchBot 25 + conservative indexing check 10. These are our audit weights, not a provider’s ranking formula.
One catalog. Different discovery paths.
Official guidance, on demand. Reviewed 2026-09-14.
GoogleGooglebot
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.
Bingbingbot
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.
ChatGPTOAI-SearchBot
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.
PerplexityPerplexityBot
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.
ClaudeClaude-SearchBot
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.
Make the store usable by agents
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.
Understand it. Fix it. Scan again.
A clear path from product page to practical fixes.
Paste a product URL
Any public product page. Any storefront.
See what’s missing
Product data, markup, and crawler access.
Make the next fix
Open a finding for evidence and a next step.
How is the broad SEO score calculated?
Checks are weighted by their technical importance. Passed checks earn full credit, warnings half, and failed checks zero. Informational, unavailable, and inapplicable checks are excluded. SEO coverage shows how many scored checks could be assessed. The provider views use fixed 100-point rubrics, with unknowns left open. Each finding shows its weight. Scores are our heuristic, not a Google or Bing metric.
No ratings? No penalty. We only validate ratings when markup is present. Training crawler preferences and llms.txt never affect the score.
What can’t this scan tell me?
- Initial HTML only; JavaScript is not executed. Client-rendered product data may be missed.
- This is a technical readiness heuristic, not a ranking, indexing, rich-result, or AI citation guarantee.
- Crawler rules describe robots.txt permissions; actual crawler access, firewall rules, and search index status are not verified.
- One Product and Offer are scored, preferring a URL match. Linked remote schemas, Microdata itemref, and advanced RDFa relationships are not resolved.
- Reviews require manual authenticity checks. Shipping/returns are presence checks; this is not a complete schema validator.
- OpenAI merchant approval, feed ingestion, freshness, supported markets, and search eligibility cannot be verified from a page scan.
Merchant Center feeds, Search Console data, Core Web Vitals, and actual AI citations need separate tools.
Do AI bots need special schema?
Google’s AI features use the same search foundations, with no special AI schema requirement. ChatGPT, Perplexity, and Claude have separate search crawler controls. Training permissions are a separate choice.