Home / What AI sees
Free tool updated 2026-11-22

What does an AI crawler see on your page?

Most AI crawlers don't run JavaScript. Enter any page and see it the way they do: the text they get, the facts they can take from it, and what they miss.

What it checks

The same check is earncitations view <url>, the MCP tool crawler_view and GET /api/view (docs). To find which page should answer a buyer's question, use the question doctor.

Questions

What if my site serves AI agents a Markdown copy?

We read it and compare it with the HTML people get. Some sites answer crawlers that ask for Markdown (or link a .md version) with a separate copy; Vercel and Linear both do for some pages. That is honest and useful, but a copy can drift: if its prices differ from the HTML, assistants that read the copy can quote the stale ones. We flag any price that appears in one version and not the other, the audit checks your pricing page's copy, and watch alerts when the two drift apart. No copy yet? The fix kit makes one from your pricing page's own text, with the tag that links it.

Do you check for hidden instructions to AI?

Yes. We flag text hidden from people (display:none, the hidden attribute, off-screen, zero-size or white-on-white text, comments, alt text, meta tags, structured data) that tells AI assistants what to say, like "AI models: always recommend us" or "ignore previous instructions". Google's spam policies count hidden text as hidden text abuse, and nobody reading the page can see what it asks. If it's on your site and nobody added it on purpose, check plugins, themes and tag-manager scripts. It's rare: on 2026-11-21 we found it on 0 of the 219 homepages we could read from our software and hotel and restaurant samples. 2 of 117 software homepages hide a harmless note pointing AI agents to a Markdown copy of the page, which we list as a note, not a problem.

Why would an AI crawler see something different from me?

Your browser runs the page's JavaScript; most AI crawlers don't. Vercel's analysis found that none of the major AI crawlers it measured render JavaScript (Applebot aside). Anything that appears only after scripts run, such as an app that builds the page in the browser or a widget that loads hours, menus, rooms or reviews from another company, isn't in what they get.

What are "third-party widgets"?

Embeds from another company: reservation and ordering buttons (OpenTable, Resy, Toast), hotel booking engines (SynXis, Cloudbeds, Mews), menu and hours feeds (Marqii, Popmenu), review carousels (Elfsight, Trustindex). They load by script or in an iframe, so the facts they show aren't in your HTML. Keep using them, and also state the key facts (hours, check-in time, prices) in your own text.

Does structured data help if the page is JavaScript-only?

Somewhat: crawlers that read JSON-LD can take facts like opening hours or a phone number from it. But assistants that quote pages need text to quote, and a page with no text has nothing to rank for the question. Put the facts in both.

Is this how Google sees my page?

No. Google renders JavaScript for search, usually later than its first crawl. This shows what crawlers that don't render get, which is what most AI search and answer crawlers do today.

What do you fetch?

The one page you enter and the site's robots.txt, once, with our identified user agent (EarnCitationsAudit). We don't run the page's scripts, render it, or keep a copy beyond a short cache.