Permission
Is the crawler allowed?
AI Search Readiness Report
Free, no signup. Five foundations—permission, extractability, renderability, operability, and commerce readiness—feed a four-stage explanation of whether an AI system can retrieve, use, cite, and accurately represent one page. These stay independent rather than becoming one invented AI score.
Five separate tests · registry 2026-07-29
Crawler access is only the first layer
Extractability
Are useful facts present in raw HTML?
Renderability
Does JavaScript successfully expose them?
Operability
Can an agent identify and use semantic controls?
Commerce readiness
Do page, schema, feed and checkout facts agree?
HTML remains primary. llms.txt, UCP, and other protocol files are optional distribution layers and cannot compensate for inaccessible or unextractable HTML.
Validate a protocol response
Upload verified-bot log evidence
Expected columns: crawler, timestamp, url, status, verification. A user-agent string alone is never treated as verified identity.
No log evidence uploaded.
Method note: stage weights are provisional and visible. An unavailable check never becomes a pass or a zero; it lowers the stage’s confidence instead. This observation describes the named consumer product at the recorded query time. A model API with search is an execution surface, not a proxy for the provider’s consumer product. Results can correlate, but they are not interchangeable.
General preserves the base readiness run. Named profiles add required page-signal checks; unavailable schema guidance and weights are shown as not evaluated.
Add observed answer evidence and retrieval-path inputs
When supplied, this also enables one bounded retrieval-off brand observation. It is not a query of ChatGPT, Gemini, or another consumer search product.
Checks run from our server; we fetch the URL you enter and don't keep the results. Anonymous run-level outcome counters may be used for aggregate research; URLs, domains, IPs, and identifiers are never included, and no statistic is released below 100 runs.
Sample report Honest output shape
Suppose the public files and page HTML are fetchable, while rendered-page evidence and the optional AI-probe budget are unavailable. A faithful report does not fill those gaps with zeros:
Retrieve · scored from the crawler evidence that was evaluated · reduced confidence
Use · not evaluated where rendering or additional inputs are required
Cite · evaluated modules shown individually; unavailable modules excluded
Believe · not evaluated when the current run has no supporting module evidence
This is a static explanation of state handling, not a captured score for a named site. Live numeric results depend on the entered URL and the modules available during that run.
How to use it
- Enter the complete URL of one public page and choose its profile.
- Complete the anti-abuse check and select Run readiness report.
- Start with the four-stage summary, then expand each stage to see evaluated and unavailable modules.
- Fix explicit access or content findings first. Re-run after publishing, while allowing for provider-specific cache lag.
+ saves the current site or page. Use ☆ beside any saved site, page, or list to favorite it. Recent check history appears below.
Create a named list
Target filled from your local choices.
Site passport Local context for this saved site
Local data
Saved targets, named lists, and recent check summaries remain only in this browser.
Consumer-product context
No consumer AI product is queried by this report. These are dated qualitative priors, not simulated results or universal per-product scores.
ChatGPT consumer search: not evaluated — no product-specific query occurred. Contextual prior: citable source, quotability, information gain, and corroboration; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Gemini app: not evaluated — no product-specific query occurred. Contextual prior: parametric brand footprint; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Google AI Overviews: not evaluated — no product-specific query occurred. Contextual prior: classic SEO retrieval and ranking; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Google AI Mode: not evaluated — no product-specific query occurred. Contextual prior: extractable passage coverage across fan-out queries; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Perplexity consumer search: not evaluated — no product-specific query occurred. Contextual prior: freshness, indexing speed, and citable passage volume; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Claude consumer search: not evaluated — no product-specific query occurred. Contextual prior: broad authority and Brave-indexed presence; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Microsoft Copilot: not evaluated — no product-specific query occurred. Contextual prior: Bing indexation; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Grok consumer search: not evaluated — no product-specific query occurred. Contextual prior: parametric brand footprint and current community presence; provenance: Source unavailable — editorial prior without a cited primary or study source, as of 2026-07-11, low confidence.
Answer-caching lag: AI answers may continue to reflect cached or previously retrieved material for days to weeks after a fix. Recheck over time before treating an unchanged answer as proof that the fix failed.
Rate this tool
What the results mean
- Retrieve — whether an answer system can discover and fetch the page.
- Use — whether the available page content is usable by the evaluated modules.
- Cite — signals related to selecting and attributing a page as a source.
- Believe — signals related to identity, support, and accurate representation.
- Confidence — how much of the provisional stage weight was actually evaluated.
- Not evaluated — unavailable evidence, never an inferred pass or error.
How it works
A protected run receives a bounded request budget, compares browser and GPTBot responses, fetches robots.txt and public agent-discovery files through SSRF-guarded endpoints, models retrieval chunks from the captured HTML, and checks the exact URL against recent Common Crawl indexes. It also evaluates answer structure, citations, entity facts, and parseable freshness evidence. If you supply a brand, one named Workers AI observation is added with model and query-time attribution. Each evaluated numeric module contributes its published provisional weight; unavailable evidence remains reason-coded not evaluated.
Features
- Four-stage funnel instead of a universal composite score.
- Visible per-module weights, findings, and fix links.
- Per-crawler/model breakdown when a module has that evidence.
- Reason-specific unavailable states for fetch, render, AI, timeout, and input limits.
- Protected, bounded network runs and shareable URL state.
Limitations
Weights are provisional. The chunk, freshness, and entity-fact checks are deterministic approximations over captured HTML, while the optional brand check is one retrieval-off Workers AI observation. A fetch from this service can differ from a provider’s geography, identity, cache, or index. The report does not query commercial answer products, observe their private retrieval systems, or guarantee retrieval, use, attribution, or factual treatment.
Frequently asked questions
Why are there four stages instead of one AI readiness score?
Retrieval, use, citationThree distinct states of AI visibility: retrieved (an AI fetched your page as source material), mentioned (your brand appears in the answer text), and cited (your URL is linked as a source). They don't always happen together, and each is measured with a different tool., and accurate representation can each report errors independently. Keeping them separate prevents a strong content signal from hiding an access error or an unavailable module.
What does not evaluated mean?
The module had no usable evidence because it needs another input, exceeded a budget, timed out, failed to fetch, or does not apply. It is excluded from the stage score and lowers confidence rather than becoming zero.
Does a strong report guarantee AI citations?
No. The report checks observable page and access signals. It cannot see every provider’s index, ranking, generated answer, cache, or source-selection system.
Why can a recent fix take time to appear?
Answer systems may cache fetched pages, retrieval indexes, or generated responses. Re-crawl and regeneration timing varies by provider, so this report can verify the current public page without proving that a provider has refreshed its copy.
What modules does the current public run evaluate?
The public run checks crawlerA crawler — also called a spider or bot — is an automated program that fetches web pages, extracts their links, and queues new URLs to visit. Search engines use crawlers to discover and download content for their index. and edge access, recent Common Crawl presence, public agent-discovery signals, modeled chunk usability, answer and citation evidence, freshness, entity facts, and—when you provide a brand—a bounded retrieval-off model observation. An unavailable dependency remains not evaluated.
Feature requests for Ai Search Readiness Report
Upvote what you want most. New ideas can be submitted from the floating Feedback menu; requests appear here once approved, and the most-wanted rise to the top.
You won't be emailed about that request anymore.
Loading…
➕ Request a feature
New requests are reviewed before they appear here.
Where this tool helps
Common use cases
Check retrieval readiness first
Confirm whether AI crawlers can reach the page before investing in downstream citation or representation work.
Expose unavailable evidence
See which future-use, citation, and representation checks could not be evaluated instead of receiving a misleading zero.
Plan work by readiness stage
Use visible stage weights to separate access blockers from later content and authority opportunities.
Create an honest stakeholder baseline
Share a stage-by-stage report without collapsing different AI surfaces into one universal score.
Watch the full workflow
AI Search Readiness Report walkthrough
Read the transcript
AI Search Readiness Report
A single A-I readiness score would hide missing evidence and unrelated failure modes. I’ll show you appropriate use cases, how to configure a page run, understand the four-stage report and confidence, add observed evidence, prioritize findings, respect limitations and cache lag, and choose the next tests.
Step 1
The report keeps Retrieve, Use, Cite, and Believe independent. Five underlying foundations cover permission, extractability, renderability, operability, and commerce readiness. This helps locate a specific bottleneck without allowing strong content to hide blocked access or missing evidence.
Step 2
Use it before publishing, during an A-I visibility investigation, after template or crawler-policy changes, or when documenting evidence for one important page. Compare representative templates separately. It is a readiness diagnostic, not a guarantee that any consumer product retrieves, ranks, cites, or believes the page.
Step 3
Enter one complete public U-R-L, choose the closest page profile, complete the anti-abuse check, and run the report. Profiles add page-specific requirements while General preserves the base checks. This walkthrough does not trigger protected network calls; it explains the honest report shape.
Step 4
Optionally supply a brand, an answer exactly as observed, the complete visible citation list, provider trace, cited passages, expected raw-H-T-M-L phrase, or separately captured rendered text. Each input enables bounded evidence modules; leaving it blank should produce not evaluated, never an invented failure or pass.
Step 5
The static sample shows state handling rather than a fabricated site score. Retrieve can be scored from available crawler evidence while confidence is reduced. Use, Cite, or Believe modules can remain not evaluated when rendering, inputs, or dependencies are unavailable. Gaps remain visible.
Step 6
Retrieve covers discovery and fetching. Use covers usable available content. Cite covers source-selection and attribution signals. Believe covers identity, support, and accurate representation. Confidence measures how much provisional module weight was evaluated, so a high score with low confidence is incomplete evidence.
Step 7
Not evaluated can mean missing input, failed fetch, exhausted render or A-I budget, timeout, unavailable module, or non-applicability. It is excluded from the numeric stage score and lowers confidence. Never convert it to zero, pass, or clean when comparing pages or reporting completion.
Step 8
A protected run can compare browser and G-P-T-Bot access, public files, recent Common Crawl presence, modeled chunks, answer structure, citations, freshness, entity facts, and optional brand observations. Fix direct access and content findings first, using each module’s reason and verification step.
Step 9
Features include a four-stage funnel, visible provisional weights, per-module findings and links, crawler or model breakdowns when observed, reason-specific unavailable states, page profiles, bounded protected runs, shareable state, operational evidence panels, and downloadable findings for ticket workflows.
Step 10
A successful rerun verifies the current public page from this service’s perspective. Providers may use different regions, identities, indexes, caches, retrieval schedules, and generated-answer caches. A recent fix can be real yet absent from a consumer answer until recrawl and regeneration occur.
Step 11
Weights are provisional. Chunk, freshness, and entity checks are deterministic approximations over captured H-T-M-L, and the optional brand module is one retrieval-off observation. The report does not query commercial answer products, observe private retrieval systems, or guarantee use, citation, or factual treatment.
Step 12
Save the checked U-R-L, profile, time, acquisition notes, each module state, weight, confidence, finding, input, and verification method. Repair the highest-impact evaluated gaps, publish, and rerun. Then separately measure crawl logs, observed answers, complete citations, provider traces, and changes over time.
Fix observable gaps—then collect stronger evidence.
Repair explicit access, rendering, content, citation, freshness, and entity gaps that the evidence supports. Re-run after publication, preserve not-evaluated states, and separately monitor logs, provider outputs, citations, and cache refreshes before claiming retrieval, attribution, or accurate representation.
Common issues & how to fix them
- Error AI crawler access is blocked Fix: Remove or narrow the matching robots.txt rule for AI crawlers you intentionally allow to access this content.
- Error Important content depends on rendering Fix: Put meaningful title, heading, and primary copy in the server-delivered HTML.
- Warning Content is not answer-ready Fix: Lead with a direct answer and add visible author and freshness evidence where relevant.
- Warning Entity evidence is weak Fix: Name the primary entities consistently, reconcile them to stable identifiers, and support them with explicit facts.
- Warning Citation evidence is insufficient Fix: Place a direct primary source beside each factual claim and recheck aging evidence.