Methodology

How the engine decides whether AI can cite a page: 172 checks in 8 groups, what each measurement rests on, and what it deliberately does not measure.

• What GeoScan measures, and what it admits it cannot

  The engine runs 172 checks in 8 groups on every page. Ten of its thresholds come from published measurements. The other sixty-five variables are judgements informed by that research, and this page says which is which, because a tool that hides the difference is asking you to trust it rather than check it.

• Version history: v1.0 to v1.2, including the score that fell from 96 to 63

  Three versions in five weeks. The first scored pages and called it AI readiness. The second asked models directly and found the gap. The third was rebuilt around it, and the same site that scored 96 now scores 63. The lower number is the honest one, and this page explains why we published the drop instead of quietly rescaling.

• The first 30% rule: where AI actually reads your page

  44.2% of all AI citations are pulled from the first third of a page's text. Not from the best paragraph, not from the conclusion you polished for an hour. From the top. If there is no self-contained answer of 40 to 80 words inside that first third, the model has nothing to lift, and it moves on to a page that does.

• Numbers beat adjectives: the 33% that AI citation research is built on

  Adding concrete numerical data to a page raises its visibility in AI answers by 33% to 41%, according to GEO-bench measurements from Princeton. Not "significantly improved results" but "conversion rose 23%". This is one of the few findings in AI search optimisation that comes from controlled measurement rather than from someone's opinion, and it is also the cheapest change on the list.

• Expert quotes are the strongest content lever, and almost nobody uses them

  Adding attributed expert quotes to a page raises AI citation rates by 28% to 41%, which makes it the single strongest content change measured in the GEO-bench study. It also happens to be the change most businesses skip, because it requires finding a real person willing to say something specific on the record.

• The trust overload paradox: how doing everything right makes AI trust you less

  Piling trust signals onto a page drops a challenger brand's chance of breaking into an AI recommendation from 55% to 21.2%. That figure comes from research on brand bias in language model recommendations, and it inverts the standard optimisation instinct. More badges, more quotes, more schema and more certifications do not add up. Past a point they subtract.

• One score hides the problem: why AI readiness needs two numbers

  A single AI-readiness score cannot describe your situation, because two independent things decide whether a model cites you: whether your page can be extracted, and whether the model has ever heard of your brand. A site can be perfect on the first and invisible on the second. Blending them into one number hides exactly the case that matters.

• FAQ markup is the only place AI quotes you word for word

  An answer inside FAQPage markup is already a complete, labelled, self-contained unit, which is exactly the shape a retrieval system is trying to build out of your page anyway. Everywhere else the machine has to guess where an answer starts and stops. In FAQ markup you hand it the boundaries, and the wording that comes back in the answer is often yours, unchanged.

• We gave llms.txt a weight of zero, and we still recommend it

  Llms.txt is a proposed convention for telling language models what your site contains and how to read it. No major AI provider has committed to honouring it. In our own scoring engine it carries a weight of exactly zero, and we still tell people to publish one. Those two positions are compatible, and the reason they are compatible is worth more than the file itself.

• Your best content is a video, and AI cannot watch it

  A video embedded on your page contributes nothing to what a language model can cite, because retrieval operates on text and an embed is a URL wrapped in markup. The forty minutes of expertise inside it are invisible. A transcript on the same page converts that forty minutes into the most quotable content you own, and almost nobody publishes one.

• Our site scored 96 out of 100 and got zero AI citations

  We built an AI-readiness auditor, ran it on our own site, and got 96 out of 100. Then we asked four language models whether they would name us when someone asks for an agency in our field. Twenty-four answers. Zero mentions. The score was not wrong about what it measured. It was wrong about what it claimed to predict, and rebuilding it around that gap took most of a month.

• The denominator problem: nobody can measure AI visibility and most tools pretend otherwise

  To state a share of AI visibility you need to know how often you were cited and how often you could have been. The second number does not exist. Nobody knows how many relevant conversations happen inside AI assistants, what was asked, or what was answered. Every percentage of "AI share of voice" you have seen has an invented denominator underneath it.

Terms used across this section

Citability

Whether a model can find, read and use the page as a source. Content multiplied by access, not added to it.

Memorability

Whether a model knows the brand at all, before any page is fetched. Returned empty below three independent external signals.

Check

One separate verdict inside the report: a single yes or no, not a group and not a signal source.

Blocker

A condition that zeroes the whole score regardless of everything else, such as noindex or a robots.txt ban.