Free audit
Run a free website audit for AI search readiness.
RankX AI samples your key pages and scores how ready they are to be found, read and cited by ChatGPT, Claude, Gemini, Grok, Perplexity and Microsoft Copilot.
Free. No account. RankX AI samples your homepage plus up to five key pages and scores them against the published rubric, reading the raw HTML exactly as AI crawlers do. The score and top findings are open; the full report unlocks with an email. Results cache for 24 hours.
What it checks
The audit runs the published rubric on sampled key pages.
Six areas, 23 published checks, every weight public. 6 sampled pages assess up to 20 checks, including 4 site-level ones no single-page grader can answer.
01 · Reach
Can AI systems fetch the page at all?
5 checks, weight 92 of the catalogue. Fail here and every other check measures a page nobody retrieves.
02 · Read
Can a machine parse what it fetched?
3 checks, weight 23 of the catalogue. A machine that cannot parse you paraphrases you, or skips you.
03 · Know
Can an engine resolve who you are?
3 checks, weight 39 of the catalogue. An unresolved name is a string, and strings do not get recommended.
04 · Answer
Does the site answer buying questions?
5 checks, weight 52 of the catalogue. The buying question gets answered either by your pages or about them.
05 · Trust
Do credibility signals exist?
2 checks, weight 15 of the catalogue. Unattributed claims lose to the same claims with a name attached.
06 · Freshness
Does the page state when it was true?
5 checks, weight 31 of the catalogue. A page that never says when it was true is easy to distrust.
The full check-by-check rubric, with every weight and pass condition, is published on the AI Readiness Score page. The audit and the single-page score share it, so they can never disagree about what matters.
Why it matters
What lets AI search find, read and cite a site.
Each area of the rubric is a measured behaviour of real retrieval systems, not a best practice. Three measurements carry most of the weight.
GPTBot, ClaudeBot and PerplexityBot execute no JavaScript
Analysis of over 500 million crawler fetches found GPTBot, ClaudeBot and PerplexityBot download JavaScript and execute none of it. Words that exist only after a framework renders are words those systems never see, which is why the audit reads every sampled page as raw HTML. A site that renders client-side is invisible however good its content is.
Retrieval quotes the top of documents
An analysis of 18,012 verified ChatGPT citations found 44% of cited passages in the first third of the page, and a controlled split test measured question-format headings lifting organic sessions by 12%. The audit checks that sampled pages open with the answer. An answer built up to is an answer never quoted.
Buying answers come from buying pages
AI engines break a question into sub-queries, and the commercial ones land on pricing, comparison and integrations pages. A measured test also showed no engine reads a price that exists only in markup, so the audit checks those pages exist and state their facts in visible text. A missing pricing page hands the cost answer to a third party.
What you get
Anyone sees the score; the full report costs an email.
Most free graders gate everything and email a PDF. This one shows results first, says what the gate buys, and shares the report by link.
Open on the page, no form
The overall score, the six area readings, the list of sampled pages and the top findings render for everyone the moment the scan finishes. The scan itself is the proof the audit works; gating the proof would defeat it. If the open half is not useful, do not trust the locked half.
Unlocked by name and email
The full report holds every finding with what was found on each page, the page-by-page breakdown and the fix list in value order. It unlocks on the page immediately, and the link arrives by email so the report survives a cleared browser. Then three short emails about what it found, and no more.
What happens to the address
The address receives the report link, then a three-email series about that report, and nothing else unless the separate newsletter box is ticked, which it never is by default. Both consents are stored with their time, source and IP, separately, as UK law expects. A consent record that cannot be evidenced is a liability, yours and ours.
Go deeper
Sibling tools go deeper on single pages, files and answers.
The audit samples; these focus. Each is free, accountless and built on the same measurements.
Questions
What to know before you run it.
Including what the audit deliberately does not measure.
What does the free site audit check?
The audit samples your homepage and up to 5 more key pages, then scores up to 20 of the 23 checks in the published RankX AI readiness rubric: crawler access, raw-HTML readability, entity markup, answer structure, dates, and whether pricing, comparison and integrations pages exist. Checks the scan cannot run on your site, and the three that need a full crawl or off-site search, are reported as not assessed, never silently passed.
Do I need an account or a card?
No. The scan runs with no signup, and the score, the six area readings and the top findings are open on the page. The full finding list and the page-by-page breakdown unlock with a name and an email, and the report link is emailed so you can reopen or share it.
How is this different from the AI Readiness Score?
The AI Readiness Score grades one URL at a time. The site audit runs the same rubric across a sample of your key pages and adds four site-level checks a single page cannot answer: whether pricing, comparison and integrations pages exist, and whether prices survive in readable text.
How is this different from RankX AI's Website Audit?
The free audit samples up to 6 key pages against the 23-check AI readiness rubric, with no account. RankX AI's Website Audit crawls up to 2,000 pages a scan against 48 technical, content and user experience checks, scores the whole crawl and turns every fix into a tracked task. To grade one URL, use the AI Readiness Score.
Which pages does the audit scan?
Up to 6: the homepage, the pricing page whenever one exists, and the highest-value pages found in your sitemap, favouring features, about, comparison and integrations pages. Without a sitemap it samples from your homepage links instead, and the report says which route it took.
Does it measure how often AI assistants name my brand?
The audit measures whether the machinery of retrieval can reach, read, resolve and quote your site, which is the part you control on-page. How often ChatGPT, Claude, Gemini, Grok, Perplexity and Microsoft Copilot actually name you is what the RankX AI product tracks, because that takes repeated sampling over time rather than one scan.
Why does it read raw HTML without running JavaScript?
Because AI crawlers do. Measurement across hundreds of millions of crawler fetches shows GPTBot, ClaudeBot and PerplexityBot download JavaScript files and execute none of them, so the audit judges exactly the bytes they judge. Text that appears only after a framework renders does not exist for those systems.
What is a good score?
100 is achievable on this rubric. Anything failing outright is worth fixing before anything else is worth debating: a pass earns its full weight, a partial pass earns half, and the score divides by assessed weight only, so a check the audit could not run never punishes you.
Can I share the report?
Yes. Every report lives at its own link, and the emailed copy carries the same link. Report pages are kept out of search indexes, so a shared report is reachable by whoever holds the link and findable by nobody else.
Will you email me marketing?
Your report link arrives first, then three short emails over the following week explaining what the report found and how to act on it. That series is about your report and nothing else, it stops on its own after the third, and every one of them unsubscribes in a single click. The newsletter is a separate box that is never pre-ticked, and both consents are recorded separately with their time and source, which is how UK law expects them kept.
How often can I re-run the audit?
Results for a domain are kept for 24 hours, so a repeat run inside that window returns the same report instantly. After that, run it again. Hourly limits per visitor and per domain keep the tool honest to operate.
Why did my audit report no score?
Fewer than two checks could be assessed, which usually means the homepage could not be fetched from our probe: a firewall, a bot wall or a broken certificate. The audit reports no score rather than zero, because zero would say your site failed when the truth is it could not be measured.
Start here
See where you show up in AI answers today.
Add your site and RankX AI suggests prompts, tracks your keywords and audits your pages, with first results minutes after setup.
7-day free trial. No credit card required. Cancel anytime.