Web Check · feature

AI visibility (can AI crawlers read your page?)

AI visibility comes down to one practical question: can AI crawlers actually read your page? Web Check answers it in seconds. Enter a page address on /ai-lite and you get a score from 0 to 100, a colour group (red, yellow or green) and five signals shown as a traffic light. The check is built for site owners, SEO specialists and developers who want to see a page the way an AI crawler sees it.

AI visibility report: score 0–100 and five traffic-light signals for AI crawlers

What AI visibility means for a site owner

AI crawlers other than Google do not execute JavaScript. They read only the raw HTML that your server returns. If the text of a page is drawn by scripts after the load, such a crawler gets an empty frame and nothing to quote. This is the principle behind the check: the js-render check counts words in exactly that raw HTML.

Poor AI visibility usually has a short list of causes: robots.txt rules that close the page to named AI bots, too little text in the raw HTML, no llms.txt file, no structured data. Web Check turns every cause into a separate signal, so you see what to fix first instead of one vague number.

The five signals behind the score

AI crawlers. The rules in robots.txt for 17 named bots: GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, Claude-Web, anthropic-ai, PerplexityBot, Google-Extended, CCBot, Bytespider, Applebot-Extended, Amazonbot, Diffbot, FacebookBot, ImagesiftBot, Omgilibot, YouBot. The result reads as how many of the 17 are allowed.
Content for AI. How many words are visible in the raw HTML without running scripts. Fewer than 300 is a fail, 300 to 600 is a warning, 600 and above is normal.
llms.txt. The file is missing, basic or full. Web Check also looks for llms-full.txt.
Structured data. None, shallow or rich Schema markup on the page.
GEO-readiness. A scale of nine signals: FAQPage or QAPage markup, BreadcrumbList, datePublished, dateModified, author, sameAs and others. The report shows "N of 9" and lists what is missing.

How a check runs, step by step

You enter the page address on /ai-lite or call GET /api/ai-lite with the url parameter.
Web Check reads the raw HTML of the page together with robots.txt, llms.txt and llms-full.txt.
Each of the five signals gets a traffic-light status with its own result: bots allowed, word count, llms.txt level, markup level, GEO points.
The report shows the total score from 0 to 100 and the colour group.
For every signal that did not pass, the page offers a quick win: a ready llms.txt template or an FAQ block with a copy button.

What you get in the report

A score from 0 to 100 and a red, yellow or green group.
Five signals with concrete numbers instead of a verdict without reasons.
A quick win for each failed signal, ready to paste into the site.
The same data as JSON, if you call the API instead of opening the page.

The full audit goes deeper with related checks: ai-access (bot by bot, plus the x-robots-tag header), llms-txt and llms-validate (the links listed in llms.txt), schema-org (the JSON-LD types found on the page) and geo-readiness. Improving AI visibility step by step then means closing the items from this list one by one.

Where it helps in practice

A site built on a JavaScript framework. You see how many words survive without scripts and whether the page lands in the fail, warning or normal band.
A site with a hand-edited robots.txt. You see at once which of the 17 named AI bots are allowed and which are not.
A freshly added llms.txt. You confirm that it is found and that the links inside it are checked by llms-validate.

The check takes seconds, so you can run it again after each change and watch your AI visibility score move.

Start a check and know the limits

Open https://webcheck.asibots.pro/ai-lite or call the API:

GET https://webcheck.asibots.pro/api/ai-lite?url=PAGE_URL

The answer is JSON with the five signals and the score. Keep in mind:

If a page returns a protection stub (for example from DDoS-Guard) or an almost empty JavaScript shell, Web Check does not invent a score. The blocks are marked "unreliable" with the reason.
The API accepts 2 requests per second. Above that you get a 429 answer with a plain explanation.
Addresses inside an internal network are rejected with a 400 answer.
The AI visibility result describes the page itself: its rules, raw HTML, files and markup.

Frequently asked questions

Why does the check count words in raw HTML and not on the rendered page?

Because AI crawlers other than Google do not execute JavaScript. Text that appears only after scripts run is invisible to them, so js-render counts words in the raw HTML. Under 300 words is a fail, 600 and above is normal.

Which AI bots does the check cover?

Seventeen named bots in robots.txt , among them GPTBot , ClaudeBot , PerplexityBot and Google-Extended . The result is the number allowed out of 17. The ai-access check in the full audit goes bot by bot.

Can I read the AI visibility result through the API?

Yes. GET /api/ai-lite?url=… returns the five signals and the score as JSON within seconds. All 53 checks together are available through audit-full .

Check my page