Paste a URL. See whether ChatGPT, Claude, Perplexity, and Google AI can actually reach and read your site, and exactly what to fix.
Free, no login, no email. Nothing stored beyond a 24 hour cache. API and MCP
Paste any URL
No login and no email. We check that page plus your robots.txt, llms.txt, and sitemap.
We check it as an AI crawler would
We fetch your page twice, once as a browser and once as GPTBot, then render it to measure how much content needs JavaScript.
Get a score and the fixes
A 0 to 100 score, four category scores, and a fix for every failure with steps for your platform.
For twenty years the job was to rank a page. An answer engine does something different. It reads your page, decides whether it answers the question in front of it, and quotes a passage back with a link. Ranking well is no longer the same as being quoted.
Two differences matter most. The first is JavaScript. Googlebot renders it. Most AI crawlers do not. A site that scores well in every SEO tool can be almost invisible to ChatGPT, because the copy arrives in the browser rather than in the HTML. That is what the readability check measures, and it is the single most common reason a good-looking site scores badly here.
The second is access. Everyone knows to check robots.txt. Almost nobody checks whether the CDN in front of the site is quietly returning 403 to AI crawlers. Cloudflare shipped a one-click toggle to block AI scrapers and a lot of people turned it on without realising it also removes them from AI answers. Your robots.txt says yes. Your edge says no. Nothing in your CMS tells you.
Access, 35 points. Whether AI systems can reach you at all. We evaluate every crawler in our registry against your robots.txt, then fetch your page as GPTBot and compare the response to what a browser gets. We also check for llms.txt and a valid sitemap.
Readability, 30 points. Whether they can read what they reach. We render your page in a real browser and compare it against the raw HTML to work out how much of your content only exists after JavaScript runs. We also check that your title and meta description are in the HTML rather than injected later.
Structure, 25 points. Whether they can understand it. One H1, an outline that does not skip levels, structured data that actually parses, and content shaped like an answer rather than a brochure.
Identity, 10 points. Whether they will cite you correctly. Organization schema, links to your official profiles, and a brand name that is consistent across your title, your schema, and your Open Graph tags.
Plenty of publishers block AI crawlers on purpose, and that is legitimate. It is also why we split the list in two. Blocking CCBot, Google-Extended, or Applebot-Extended keeps your writing out of training data and costs you nothing in AI answers. We report that as a warning with an explanation, never as a failure. Blocking OAI-SearchBot or PerplexityBot is different: those are the crawlers that decide whether you can be cited at all.
It is a 0 to 100 measure of how easily AI answer engines can reach, read, understand, and attribute your page. We run 12 checks across four groups: Access (can they reach you), Readability (can they read it), Structure (can they understand it), and Identity (will they cite you correctly). A pass earns full points, a warning earns half.
The same run this page does, as an HTTP endpoint and as the MCP tool check_ai_visibility. No account, no API key. Full reference.
curl -sS https://usesuperflow.ai/api/tools/ai-visibility \
-H 'Content-Type: application/json' \
-d '{"url":"example.com"}'Returns { ok, report: { score, grade, scoredOutOf, categories[], findings[] with why and fix, detection }, cached, ageSeconds }. 10 runs per hour per IP. Allow up to 75 seconds.
claude mcp add --transport http superflow https://usesuperflow.ai/api/mcpThen ask your agent for check_ai_visibility. Setup for Claude Desktop, Cursor and VS Code is on the reference page.
No login, no email, no ads. Same as this one.
Superflow agents check every page of every site you ship, on every change, against your own QA rules. Your team approves, then your client signs off. No client login required.