We Scored 11.7 Million Websites on How Well AI Can See Them. 97% Are Failing.
Search is quietly moving from Google's ten blue links to a single, confident answer written by ChatGPT, Claude, and Gemini. So we asked a simple question at scale: when an AI writes that answer, whose website can it actually read? We built the largest dataset of its kind to find out. The results are worse than we expected.
AI visibility score distribution across 11.7M scored US business sites. Source: Engagemii research, July 2026.
The headline numbers, across 11.7 million US business sites: the average AI visibility score is 4.53 out of 10, only 3.0 percent score a 7 or higher, 16 percent are near-invisible to AI, and just 27 sites out of 11.7 million scored a perfect 10.
The web was built for a reader that no longer shows up
For twenty years, a website's job was to rank. You wrote for Google's crawler, you chased the ten blue links, and a human clicked through to you. That entire motion is being compressed into one step. A person asks an assistant a question and gets back a single, confident answer, and increasingly they never click anything at all.
That answer is assembled from whatever the model can find and parse on the open web. Not what looks good to a human. What is legible to a machine: structured data, clean entity signals, files that explicitly tell an AI crawler what a business is and does. Most of the web has none of it, because it was never built for this reader.
The average US business scores 4.53 out of 10. Four out of five sites sit in a mushy 4 to 6 middle: crawlable, but ambiguous. To an AI deciding whose name to put in the answer, ambiguous is the same as absent.
What the 3% do that everyone else doesn't
We score every site across six dimensions of how an AI actually reads a page. The sites that break 7 aren't better companies. They're just legible ones. Structured data: Schema.org markup that states facts an AI can lift verbatim, instead of guessing them out of prose. Content structure: headings and answer-shaped passages a model can quote cleanly. Entity clarity: who this business is, where, and what it sells, with zero ambiguity. E-E-A-T signals: the experience, authority, and trust markers models weight. Technical AEO: robots, sitemap, and an llms.txt that welcomes AI crawlers instead of blocking them. AI discoverability: presence in the sources these engines actually pull from.
None of it is exotic. Almost all of it is mechanical. Which is exactly why the gap is so striking: the winning 3 percent aren't doing anything a competitor couldn't copy in an afternoon. They just did it, and everyone else assumed their old SEO would carry over. It doesn't.
The good news is hiding in the fat middle
Look at the chart again. The bulk of the web isn't dead, it's stuck at 4 and 5, one or two concrete fixes away from a real jump. A site with strong content and clean entities but no structured data is leaving points on the table that take an afternoon to claim. The distance from invisible to cited is smaller than it looks. It's just been invisible itself, because nobody was measuring it.
How we measured it
The dataset spans 11.7 million US business domains, sourced from public web-scale lists (Common Crawl, the Chrome UX Report, Tranco, Majestic) and scored by a fleet that fetches each site, runs a US-relevance check, and grades the six dimensions above into a single 0 to 10 number. No paywalled data, no scraping behind logins. New sites are added and re-scored continuously, so the distribution above is a living snapshot, not a one-time crawl.
Want to know where your site lands? We'll score any business against the same 11.7 million for free, and hand you the exact fixes to climb. No pitch. See your free AI visibility score at engagemii.com/aeo.
If you want to cite this article, the URL is engagemii.com/blog/we-scored-11-million-websites-on-ai-visibility-97-percent-failing.
Ready to find out if AI can cite your brand?
Get Your Free AEO Score