The category
AI Search Intelligence
The discipline of measuring how AI systems find, understand and recommend a business — and why that is a different measurement problem from the one SEO reporting was built to solve.
A different question, not a better answer
For twenty-five years, being findable online meant occupying a position in a list. The list was ranked, the ranking was measurable, and an entire industry grew up around moving documents up it. The measurement model followed the interface: keywords, positions, impressions, click-through rates. It worked because the interface was stable and the unit was unambiguous — a page either ranked third or it did not.
Generative answers break that model in a specific way. When somebody asks an assistant “who should I use for commercial solar in Brisbane?”, they do not receive ten links. They receive a paragraph that names two or three businesses, characterises each one, and possibly cites a few sources. There is no third place. There is inclusion or absence, and if you are included there is the separate question of what was said about you.
That is why AI Search Intelligence is not a rebadged SEO report with a new chart on the front. It is a different measurement object. SEO asks where do my pages rank. AI Search Intelligence asks what does the system believe about my business, and does that belief lead it to recommend me. The first is a position. The second is a claim — and claims can be incomplete, out of date or simply wrong in ways a ranking never could be.
You are the fourth result for this keyword.
A ranking says
Two firms come up most often for this. One specialises in commercial work.
An answer says
Were you one of them, and was the description of you true?
So we must ask
Where it differs
SEO reporting and AI Search Intelligence, side by side
Not a replacement. The two disciplines share inputs and diverge almost everywhere else.
| Dimension | SEO reporting | AI Search Intelligence |
|---|---|---|
| Unit of measurement | Position in a ranked list | Presence in a synthesised answer |
| Result shape | Ten links, stable between users | A paragraph naming two or three businesses |
| Query input | Keywords and search phrases | Conversational buyer questions |
| Reproducibility | High — the same query returns near-identical results | Variable by model, location, personalisation and time |
| What can go wrong | You rank too low | You are absent, or described incorrectly |
| Source of truth | An index of documents | A model plus retrieved sources plus citations |
| Correct output | A rank trend over time | A distribution across repeated observations |
| Primary lever | Relevance and link authority | Legibility, corroboration and answerability |
The practical consequence of row four is the one most businesses underestimate: because generative answers are not reproducible, any tool that reports a single AI “rank” without stating how many observations it is based on is over-claiming. See how we handle that in the score.
The measurement set
Four things worth measuring, in order
Everything else is a refinement of one of these.
- 1
Presence
Across a defined library of buyer questions for your category and market, how often are you named at all? This has to be measured against questions customers plausibly ask, not against your brand name — almost every business appears when you ask an assistant about it directly.
- 2
Accuracy
When you are described, is the description true, current and complete? Wrong service areas, outdated specialisations and invented capabilities are common, and they cost more than a missing mention because they actively misdirect a buyer.
- 3
Competitive share
Of the businesses named in answers to your questions, what proportion are you? Absence only becomes a commercial argument when you can see who occupied the slot instead, and on which kind of question they did it.
- 4
Retrievability
Is your own site legible enough to be the source? Structured data, crawl access, explicit service and location statements, question-shaped headings and answer-first writing all determine whether a system can quote you rather than paraphrase somebody else about you.
How AI Says approaches it
Three commitments that shape the whole product
Observed and estimated are never mixed
Every component score carries its provenance. A number derived from live AI answers is labelled observed. A number derived from measured website and entity signals is labelled estimated. They are never averaged into an undifferentiated figure, because they are not the same kind of evidence.
Confidence is reported, not folded in
How much data an assessment rests on is shown separately from the score itself. A 62 from many observations and a 62 from two are different claims, and flattening them into one number would hide exactly the thing a decision-maker needs.
No number without evidence
Every audit result shows the raw signal behind it — the title we read, the schema types found, the URL count in your sitemap. If a check could not run, it is reported as not tested rather than counted as a failure.
AI Says Score™ is a proprietary visibility assessment developed by AI Says. AI-generated answers may vary by platform, model, location, personalisation, prompt wording and time.
Related disciplines
The vocabulary, untangled
Three acronyms circulate in this space and they are not interchangeable. Each has a page explaining it properly.
AEO
Answer Engine Optimisation
Making an individual page structurally capable of supplying an answer: schema, headings, question coverage, answer-first prose. Measurable today, and the basis of our readiness audit.
Read the pageGEO
Generative Engine Optimisation
Being retrieved and cited by generative systems. Broader than AEO, dependent on sources beyond your own site, and the area where the least is publicly known.
Read the pageMonitoring
AI brand monitoring
Watching all of the above change over time, because a single measurement of a non-deterministic system is a snapshot rather than a state.
Read the page
All of it rolls up into 7 scored components. Browse the resources or run a free check to see where your business currently sits.
Questions
Questions about the discipline
No, though they overlap. SEO measures position in a ranked list of documents. AI Search Intelligence measures inclusion in a synthesised answer — whether you are named at all, how you are described, whether you are recommended and which sources the answer drew on. The unit of analysis changes from a ranking to a statement, and statements have to be assessed for accuracy as well as presence.
Because there is no stable list to rank within. Ask the same question twice and the wording, the businesses named and the citations can all differ. The honest measurement is distributional: across a defined set of buyer questions, run repeatedly, how often are you present, recommended, cited and correctly described? A single answer is an observation, not a position.
Four things. Presence — are you named in answers to the questions your customers actually ask. Accuracy — is what is said about you true and complete. Competitive share — who is named instead of you. Retrievability — is your own site legible and quotable enough to be the source. Anything beyond those four is refinement.
Often enough to separate signal from noise. A single check establishes a baseline; the second check tells you whether the first was typical. Because model updates, new content and competitor activity all move the result, a cadence of weeks rather than days or quarters tends to be the useful compromise.
The opposite, in many categories. A large brand is usually described adequately because a great deal has been written about it independently. A smaller business is far more dependent on its own site being legible, its entity being corroborated and its services being stated explicitly — which is precisely what this discipline measures and improves.
Find out what AI says about your business
Enter your website and get your AI Says Score. No credit card required.