This free AI readiness checker reads your live page and scores how ready it is to be quoted by ChatGPT, Perplexity, Gemini and Google AI Overviews: crawler access, freshness, schema, author and extractable answers. Free, no signup. Every check is a real read of your page, not a guess.

How to use it

Paste one page URL, not your homepage. Use the specific article or product page you want an engine to quote. Homepages score badly here and that is usually correct: they are rarely what gets lifted into an answer.
Read the failed checks, not the number. The score is only the sum of the checks. The value is in which ones failed and why.
Fix, wait, re-run. Each failed check comes with the specific change and a link to the longer explanation.

It works on any public URL, so you can also point it at a page you have watched get quoted in ChatGPT and compare it to yours line by line. That comparison is usually more useful than your own score in isolation.

The seven checks, and what each is worth

Listed in the order they appear in your result, with the points each contributes. Nothing is hidden: the score is these seven numbers added together, and every check reports what it looked for.

Where the weights come from. The seven checks are drawn from what AI crawlers and answer engines document publicly: robots.txt access, server-rendered text, schema, a self-contained opening answer. The point values are our own editorial judgement about which failures cost you most, not a measured coefficient from any study. We weight crawler access highest because a page an engine cannot fetch cannot be cited at all, and the rest follows that logic. Treat the score as a structured checklist, not a prediction.

Check	Points	What has to be true to pass
AI crawlers can reach the page	20	Your robots.txt does not shut out any of the crawlers behind ChatGPT, Perplexity, Claude or Gemini. A blocked crawler can never quote you, which is why this is the heaviest check.
Freshness, meaning a real visible date	20	The page carries a date signal: a published or modified date in its schema, a `<time>` element, or a visible line such as “Last updated March 2026”. Undated pages read as unmaintained.
Schema markup	15	The page carries at least one machine-readable schema block, and the tool reports how many it found. This is what lets an engine parse what the page is and who stands behind it.
Named author or byline	15	A real person is credited, either in the schema, in an author link, or as a visible byline near the top. Anonymous pages are harder for an engine to justify quoting.
Self-contained opening answer	15	The first paragraph is between 25 and 120 words. Shorter is usually a fragment; longer is usually a wind-up. In that band it is a self-contained passage an engine can lift whole, which is exactly what gets quoted.
Question-shaped headings	10	At least one heading on the page is phrased as a question, either ending in a question mark or opening with a question word. It is a low bar on purpose: one is enough to show the page is organised around what people ask.
Title and description	5	A real page title and a description of at least 20 characters. Worth few points because almost everyone passes, but failing it means something is genuinely broken.

How to read the score

The number lands in one of three bands, and each one implies a different next move:

Score	Band	What to do
80 to 100	AI-ready	The mechanical work is done. Stop optimising this page and go earn mentions elsewhere, because that is now your only remaining lever.
50 to 79	Needs work	One or two structural things are missing, usually the date or the opening answer. These are hours of work, not weeks. Fix them before you do anything else.
Below 50	Not ready yet	Something fundamental is off, and it is often crawler access or a page that returns almost nothing to a plain fetch. Read the failed checks in order of points.

An honest note on the number itself. No engine publishes a scoring formula for citations, and anyone who tells you they have reverse-engineered one is selling something. These weights are our judgement about what matters, applied consistently, and the score is simply their sum. That makes it useful for comparing two pages against each other and for tracking one page over time. It is not a prediction, and a 90 does not mean you will be quoted.

What this tool does not do

Every checker has a boundary. Here is ours, in full, because a limit you do not know about is how you end up drawing the wrong conclusion:

It does not tell you whether any engine mentions your brand. This scores a page, not a brand. Not one line of it talks to ChatGPT, Perplexity or Gemini. To find out whether they name you, ask them, which is what the Prompt Generator is for.
It does not run JavaScript. It reads the HTML your server returns. If your page builds its content in the browser, this tool sees an emptier page than you do and will score it lower. Worth investigating rather than dismissing: how much JavaScript each AI crawler executes varies, so a page that only exists after rendering is a real risk, not just a quirk of our checker.
It checks that things exist, not that they are correct. The schema check counts blocks; it does not validate them. A page with broken markup passes. Use a schema validator for correctness.
The opening-answer check reads the first paragraph in the source, which is not always the first paragraph you see. A cookie line, a banner or a promo blurb sitting earlier in the HTML will be measured instead of your actual opening.
The author check is pattern matching. A page saying “by Jane Smith” passes whether or not Jane is credited properly, and an unusual byline layout can fail even when a real author is named.
The crawler check only catches a blanket block in robots.txt. Firewall and bot-management rules that turn crawlers away at the network edge are invisible to it, and so are page-level directives such as noindex.
One URL at a time. It does not crawl your site. To assess a template, run one representative page from each type.
The page has to be public and reasonably quick. Anything behind a login, or slower than about nine seconds to respond, returns an error rather than a score. That is deliberate: an error is honest, a fabricated score is not.
Results are cached for a few minutes. Fix something and re-run immediately and you may see the old result. Give it five minutes.

Fixing the checks that fail most

In the pages we run through this, four failures come up again and again, and all four are cheap to fix:

No date. Add a real, accurate “Last updated” line that a reader can see, and make it match reality. Faking a date to look fresh is the one fix here that can actively cost you trust.
No opening answer. Your first paragraph is scene-setting. Replace it with a self-contained answer to the page's core question, in roughly 40 to 60 words, that would still make sense quoted alone with no context around it.
No named author. Put a real person on the page with a link to who they are. This is a five-minute change with a disproportionate effect on how quotable a page reads.
Blocked crawlers. The heaviest check and usually an accident. The full fix is in allowing AI crawlers in robots.txt, and the AI Bot Checker shows you which specific ones are shut out.

The wider version of this work, beyond the seven mechanical signals, is in answer engine optimization and how to get cited by ChatGPT.

When to run this

Before you publish anything you want quoted. Catching a missing date or a buried opening answer before launch costs nothing; catching it six months later means the page has been underperforming the whole time.
On a page that gets Google traffic but never shows up in AI answers. That gap has mechanical causes worth ruling out before you assume it is an authority problem.
On a page you have watched get cited. Run a competitor page that keeps appearing in ChatGPT answers, then run yours, and compare which checks each one passes. The difference is your work list.
Once per template. If your blog posts all share a layout, one failure is usually every post's failure, and one template fix moves the whole library.

What a high score does and does not mean

A high score means your page clears the on-page bar: the engines can reach it, parse it, and lift a clean answer out of it. That makes you eligible. It does not make you cited. The honest truth is that the biggest remaining lever is off-page, and it is genuine authority and mentions of your brand across sites the engines already trust. Think of this as the checklist you complete first so that the authority work has something to land on, then go and do the harder half in improving your brand visibility in AI search.

Questions people ask

Does it check whether ChatGPT mentions my brand?
No. It reads your page and scores how quotable it is. Whether an engine currently names your brand is a different question, answered by asking the engines directly and repeatedly. The Prompt Generator builds the questions, and how to measure AI search visibility covers turning that into something you can track.

Does it really read my page?
Yes. A small function on our side fetches the URL you enter and its robots.txt, then checks the actual HTML. There is no model in the loop and no estimate anywhere. If we cannot reach the page you get an error, never an invented score.

My page looks great but scored badly. Why?
The most common cause is a page whose content is assembled in the browser, because this tool reads what your server sends rather than what renders. The second most common is a first paragraph that is not the one you think it is. Both are listed in the limits above, and both are worth checking rather than dismissing.

What counts as a good score?
80 or above clears the bar. Below that, read the failed checks in order of points: the two twenty-point checks carry the most weight individually, so start there.

Can I check a page I do not own?
Yes, any public URL. Running the pages that already get cited in your category is one of the more useful things you can do with this.

Is anything stored?
No. The URL is used for the fetch and nothing else. No account, no log, no database. The result is cached briefly at the edge, which is why an immediate re-run can repeat itself.

Is it free?
Yes. No signup, no account, no trial that expires.

The other three tools

AI Bot Checker reads your robots.txt and tells you which of eight AI crawler families are allowed in. Prompt Generator builds the buyer questions to paste into ChatGPT and Perplexity so you can see where you are missing. llms.txt Generator builds a curated index of your best pages. All free, all on our tools page.

Is your page ready to be cited by AI?