About CrawlRadar
CrawlRadar scores any page 0–100 on whether AI answer engines can find, parse and cite it. It exists because “is my site ready for AI search” had a lot of opinions attached to it and no instrument.
By the CrawlRadar team · Published · Last updated
What the score is
The score is a measure of eligibility, not a prediction. It says whether an answer engine can retrieve, attribute and quote a page, not whether it will, which depends on the query and the competing sources and is not observable from outside.
What it deliberately does not do
Three things, each of which would make the number bigger and less honest.
- It does not guess. When a measurement fails, whether a robots.txt that times out or a challenge page, the check reports not checked and its points leave the denominator. Awarding full marks there once scored a site that blocks GPTBot as though it welcomed everyone.
- It does not oversell a check. llms.txt is worth 4 points and the fix text says the evidence for it is nil. A tool that inflates a check to seem comprehensive is worth less than one that admits the check is minor.
- It does not hide a heuristic. Where a check infers rather than measures, the summary says so. False positives cost more trust than they save time.
How the extension handles your data
Scanning does not touch it. The extension installs with activeTab, scripting and storage, with no host permissions, no content scripts running in the background, no analytics. Settings stay in local browser storage, and a scan talks only to the site it is scoring.
Three optional features do reach the network, each behind a permission you grant when you enable it, and only one of them sends page content anywhere: AI checks, which post the text straight to your own provider key. The privacy policy covers all three.
The website is different by necessity: it has to fetch the URL you give it. It fetches that page once, anonymously, with no cookies, and does not store the URL unless you explicitly create a share link.
The scoring model
Fourteen checks across three levers, weighted by how much failing each one costs. The full breakdown is on the GEO score checker page.
Frequently asked questions
Who makes CrawlRadar?
CrawlRadar is an independent tool, not affiliated with OpenAI, Anthropic, Google, Perplexity or any other AI vendor. The engines are named because they are what the scoring model is about, not because anyone endorses this.
Does the extension send my browsing data anywhere?
Scanning does not. The extension installs with activeTab, scripting and storage, no host permissions and no analytics, and the scoring engine runs inside the page you are looking at, and the only requests a scan makes are same-origin fetches of that site’s own robots.txt, llms.txt and raw HTML. Three optional features can send data, each behind its own permission you grant at the moment you enable it: in-editor scoring on a domain you choose, a plan check against crawlradar.xyz that carries no page data, and AI checks that send page text directly to your own Anthropic, OpenAI or Perplexity key. See the privacy policy for what each one does.
Is the scoring model the same in the extension and on the website?
Yes, literally the same code. The scoring engine is a shared package with no IO of its own; the extension feeds it a rendered DOM from your tab, and the website feeds it a server-parsed document. Scores can still differ, because those are genuinely different views of a page: the website sees what a non-JavaScript crawler sees.
How are the weights decided?
By judgement about how much each failure actually costs a citation, not by a model fitted to data. There is no public dataset of what answer engines cite. The weights are documented, versioned and stated openly so you can disagree with them: 25 points for retrieval, 15 for source trust, 60 for extractability.