How we measure
We score every business against the same criteria, ask the same kinds of questions of the same assistants and report the result the same way every time. This page sets out exactly how, because a measurement you can’t inspect isn’t a measurement.
Three angles, scored separately
Ordinary search results, the map results and what AI assistants do with your pages. Different systems produce them, so we never combine them into a single number.
- SearchCan a search engine find, read and rank these pages? The oldest and best-understood of the three.
- MapDoes the business appear when someone nearby is ready to book? Highest intent, and driven by the business profile as much as the site.
- AI answersDoes an assistant name you, or does it only cite you? Different inputs, and we’ve measured cases where it disagrees with search.
The words for this, if you want them
You may have seen the acronyms. They describe the three angles above, and we use them here rather than on the pages a first-time visitor reads.
- SEO — search engine optimisation. Being found in the list of blue links.
- AEO — answer engine optimisation. Being the source used when a system answers the question directly instead of returning a list.
- GEO — generative engine optimisation. Being named inside an answer written by a generative assistant such as ChatGPT or Gemini.
The boundary between the second and third isn’t settled, and different firms draw it differently. We use them as shorthand for the three angles we score, and nothing more.
If a firm sells you “GEO” without showing you the questions it will ask and how it will score the answers, the acronym is doing the work that a measurement should be doing.
A fixed way of scoring, so the number means something next quarter
Every business is scored against the same criteria, with the same weightings, by the same method. A score that moves because we changed the ruler isn’t a result.
A score on its own tells you nothing — 61 out of 100 is meaningless without context. So every score is reported alongside other businesses in your area, measured the same way on the same day.
The question sets
We don’t ask an assistant about you. We ask it the question a customer would ask, and see who it names.
- AI assistants — 20 questions. Treatment plus location, treatment plus concern, and “best / who should I see” phrasings. Run against ChatGPT and Gemini. Scored on two things we keep apart: whether you’re named in the answer text, and whether you’re cited as a source.
- Map results — 10 questions. The phrasings someone close by and ready to book would use.
- The sets are yours, built from the treatments you told us you want to grow. We revise them each quarter so they keep pace with what you’re trying to grow — and we tell you when we do, because changing the questions changes the numbers.
What we measure, and when
The type column matters: what we did and what moved are reported side by side and never added together.
| What | Type | Monthly | Quarterly |
|---|---|---|---|
| Work log — what we did | Execution | Yes | Yes |
| Google Search Console — impressions, clicks, queries | Outcome | Yes | Yes |
| Position in the map results (10 questions) | Outcome | Yes | Yes |
| AI assistant answers (20 questions) | Outcome | Yes | Yes |
| Full scoring across all three angles | Execution | — | Yes |
| Comparison with other businesses in your area | Reference | — | Yes |
We run the AI measurements every month against the fixed question set, but we act on them at the quarterly reset. Month to month the movement can be noisy: the same question can name a business one month and not the next, with nothing changed on the site. A single month isn’t a trend in either direction.
What the measurement can’t do
Saying this here rather than in the small print.
- It doesn’t predict. We don’t forecast rankings, citations, traffic or enquiries.
- It doesn’t establish cause. We report what we did and what moved, and we don’t claim one produced the other.
- It’s a sample, not a census. Twenty questions isn’t every question a customer might ask, but it’s enough to stay comparable over time.
- We run the AI measurements by hand. That’s why the question set is fixed and modest in size.
Common questions
- Can we see the criteria?
- Yes. They come with the review, along with your score against each criterion and the reasoning, so you can push back on it.
- Why only twenty questions?
- Because we run them by hand, and we’d rather ask twenty questions consistently every month than a hundred once. Consistency is what makes the numbers comparable.
- Do you use a tool for the AI measurements?
- We run them ourselves against a fixed set of questions. The observations we quote on this site come from that work, not from a vendor’s data.
- What if our competitors are measured differently?
- They aren’t. The comparison is measured the same way, on the same day, against the same criteria. That’s what makes the number mean anything.
Where to go from here
The review is where these measurements are first taken, and the report is where they arrive every month afterwards.