What an audit does
We make about 11 requests to your site: the homepage twice (once asking for Markdown), robots.txt, sitemap.xml, llms.txt and six standard /.well-known paths. If we can't read the homepage, we stop after robots.txt. We don't run JavaScript, crawl other pages or send requests as anyone but ourselves. Our user agent is AI-Visibility-Checker/1.0.
A lookup audits a site at most once per calendar day, so the first lookup each day runs a fresh audit and later ones reuse it. The re-check button runs a fresh audit once every 24 hours.
The scores
Crawler access
Can crawlers get to your content? We check HTTPS, robots.txt, your sitemap, whether robots.txt blocks AI search crawlers, and whether the homepage has real text before JavaScript runs. Blocking training crawlers while allowing search crawlers is a legitimate policy, so it doesn't lower this score.
Search fundamentals
Titles, descriptions, structured data, canonical URLs, viewport, language, image alt text and Open Graph tags.
Emerging standards
llms.txt, Content Signals, Markdown negotiation, Link headers, API catalogs, MCP server cards, OAuth and OpenID discovery, and Web Bot Auth. Few sites need all of these, and none has proven ranking value, so we report how many you've adopted and leave them out of the headline score.
Headline score
The average of crawler access and search fundamentals. Within each, a pass counts fully, a warning counts half and a failure counts zero. If robots.txt blocks a search crawler (not counting ChatGPT-User, which OpenAI says isn't used to determine what appears in Search), or the homepage has almost no text without JavaScript, the headline is capped at 49, because the rest matters little if crawlers can't read you.
Levels: Excellent from 90, Good from 75, Fair from 50, Needs work from 25, otherwise Critical.
AI Visibility
The free check doesn't measure this. The AI visibility page shows it with sample data, and reports show a blurred sample rather than results for your site. If you want to use it, reach out to us.
How we avoid false results
- A 200 isn't enough. Many sites return their normal page for any URL. A file only passes if it has the right content, such as valid JSON with the fields the standard requires.
- We read robots.txt rules. Mentioning a crawler doesn't count as allowing it. We apply the matching group's rules, following RFC 9309.
- No score without a measurement. If a site blocks our bot or doesn't respond, the report says "Couldn't audit" instead of giving it a low score, and it's left out of benchmarks.
What we can't see
- We only audit the homepage and a few fixed paths, not your whole site.
- Crawler policy comes from robots.txt. A firewall or CDN can block crawlers that robots.txt allows.
- We can't tell how much crawlers actually visit, or whether they cite you.
- A high score means no technical barriers were found. It doesn't mean AI systems will recommend you.
What the evidence says
Google's AI optimization guide says its AI features rely on core Search, and that you don't need new machine-readable files or special markup to appear in them. That's why fundamentals count toward your score and llms.txt doesn't.
The GEO paper (KDD 2024) found that citing sources and adding statistics and quotations improved visibility in a simulated engine, while keyword stuffing didn't help. Content is outside what this tool measures, and it probably matters more than anything here.
Sources for the numbers on our site
- Synchrony, March 2026: 56% of US consumers used generative AI during the 2025 holiday season. Of those, 34% compared products and 29% hunted for the best price. 1,000 US consumers, Oct–Dec 2025.
- Digital Commerce 360, August 2026, reporting Adobe Analytics: in July 2026, AI referrals to US retail sites rose 62% year over year and converted 60% better than other traffic, based on more than 1 trillion visits. The monthly figure moves around, and this covers retail only. We read this report rather than Adobe's own.
- Pew Research Center, July 2025: Google users clicked a result in 8% of visits with an AI summary and 15% without. 900 US adults, March 2025. Google only, and Google disputes the method.
- OpenAI: sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers," though they "can still appear as navigational links."
- Vercel and MERJ, December 2024: none of the major AI crawlers tracked rendered JavaScript, while Gemini uses Googlebot's rendering. This is a 2024 measurement and may have changed.
- Gartner, May 2026: in a survey of 846 US consumers, among those who used AI for a recent purchase, 54% had to double-check everything it told them and 62% said its information ended up being a waste of their time. Gartner also reports that willingness to let AI make the purchase topped out at 11%.
Example reports and the AI visibility page are illustrative and labelled as such. Numbers from our own audits are real and show their sample size.
Extra pages
On the free plan you can check your homepage plus up to 4 more pages on the same site, once a day. Each extra page takes one request. We look at the page's title, description, canonical, headings, word count and structured data.
Benchmarks
Benchmarks use each site's latest audit. Sites people chose to check are reported separately from sites we added from the Tranco top-sites list, because the first group is self-selected. We hold back per-check comparisons below 30 sites and percentiles below 100.
The traditional sectors group covers real estate, travel and airlines, hotels, insurance, banking, healthcare and automotive. From the same Tranco ranking, we took the highest-ranked real companies in each sector, one per brand, leaving out directories, media sites and Russian banks. We fixed the list before scanning any of them. That selection is ours, so it isn't a random sample, and we report these sites by sector rather than by name.