Research
The AEO Gap Report
We scored 615 top-ranking pages across 13industries on how ready they are to be cited by AI answer engines. Even the winners leave citability on the table - and it's the same three gaps every time.
of top-ranking pages have no proper Article or FAQ schema - the markup that tells an engine what the page is.
don't phrase section headings as the questions people actually ask.
pages returned no readable HTML to a standard crawler at all - before AEO even begins.
These aren't obscure pages. Each ranks on page one for a real, high-intent question - “how long does probate take”, “what is APR”, “how much does a new roof cost” - the kind an AI answer engine now fields directly. We ran every page through the same deterministic checker behind our free AI Visibility Check: eight structural signals of citability, no AI in the loop, fully reproducible. The finding isn't that these pages are bad - it's that ranking and being citable are no longer the same thing.
Finding 01
Ranking well isn't the same as being citable
Most pages score respectably - they're page-one results. But nearly a third fall below 80, and the points they lose cluster in the signals answer engines lean on hardest.
AEO score distribution · n=615
Finding 02
Three gaps account for almost all the lost ground
When a top page loses points, it's the same short list every time - and none of them are about substance. They're about structure an engine can extract.
Share of pages missing each signal · n=615
The headline
You can't be cited if you can't be read
Before structure even matters, access does. More than one in five pages we sampled - 161 of 776 - returned no readable HTML to a standard automated request, whether from bot-blocking or JavaScript-only rendering that leaves nothing in the page source. Of the pages that did respond, another 5.8% explicitly block AI crawlerslike GPTBot and ClaudeBot in robots.txt. A large share of the web's best content is simply invisible to the engines now answering the questions it was written for - and it's the most fixable gap of all.
Finding 03
The sectors that most need trust are least ready
Split by industry, a clean divide appears: commercial and how-to niches score in the high 80s, while the high-stakes “your money or your life” sectors - legal, finance, healthcare - trail by up to twelve points. Exactly where a cited, authoritative source matters most, the content is least built to be one.
Mean AEO score by industry · /100
Read the full breakdown for each sector: Real estate, SaaS & software, Ecommerce, Home services & trades, Travel & hospitality, Healthcare, Finance & fintech, Legal & law firms, Education & e-learning, Beauty & wellness, Professional services, Manufacturing & industrial, Automotive.
What it means
The takeaways
The citability gap is structural, not editorial
Only 3.2% of pages were too thin. The winning pages already have the substance - they just don't present it in a form an engine can lift.
Half of top pages have no proper schema
Without Article or FAQ markup, an engine has to guess what the page is. It's a template-level fix that lifts every page on a site at once - the highest-leverage change on this list.
49% don't ask a single question
Their headings are statements. Rewriting “Our roofing process” as “How long does a roof replacement take?” maps a section straight onto a query - a five-second change.
Access is the silent killer
One in five pages can't be read at all, and 6.6% block AI bots outright. Check your robots.txt and your rendering before anything else - citability starts with being readable.
Methodology
How we measured it
We took high-intent informational queries across 13industries - the “how / what / how much” questions AI answer engines now field directly - and collected the top-ranking content pages for each (776URLs). Each page was scored 0-100 by a deterministic checker (the same one behind our free AI Visibility Check): title and meta length, a single H1, question-shaped H2s, an FAQ, Article/FAQ schema, word count, and whether robots.txt allows AI crawlers. No AI is used in the score, so it's fully reproducible. 161 pages returned no usable HTML and are reported as an access finding rather than scored.
For comparison: we ran the same checker across 450 pages built specifically to be answer-first. They averaged 87, with their gaps clustering in cosmetic title length, not structure - purpose-built content closes exactly the three gaps this sample reveals.
How does your page score?
Run any URL through the same free check we used for this study - a score and the exact fixes that matter most.