A single number from 0 to 100 describing how well a business is represented across AI answers and search. Here is exactly what goes into it, how it is calculated, and what it deliberately does not measure.
We publish the method because a score nobody can interrogate is worth nothing.
Every marketing supplier has a proprietary score, and almost none of them will tell you how it works. That is usually because the answer is embarrassing.
We take the opposite position. The whole method is on this page: the four components, their weights, how the question set is built, how often it is measured, and what the score cannot tell you. You are free to disagree with the weighting — several clients have, and one of them was right.
If a supplier won’t show you the formula behind their score, the score is a retention device rather than a measurement. Ask them. The reaction is informative.
Each component is scored 0–100 in its own right, then combined.
| Component | Weight | What it measures | How it is calculated |
|---|---|---|---|
| Answer Coverage | 35% | How often the business appears at all | Percentage of tracked questions where the business is named in the answer. |
| Answer Accuracy | 25% | Whether what is said is correct | Percentage of factual claims that match reality. Severe errors — wrong address, phantom closure — carry double weight. |
| Answer Trust | 20% | Strength of independent corroboration | Breadth and quality of independent sources confirming the business’s core facts and claims, scored against a fixed source list. |
| Share of Answer | 20% | Position against named rivals | Business mentions divided by all business mentions across the tracked question set. |
Coverage 58, Accuracy 74, Trust 51, Share 44. Weighted: (58×0.35) + (74×0.25) + (51×0.20) + (44×0.20) = 57.8, rounded to 58.
Two businesses can both score 58 and need completely different work. Always read the components.
A 58 built on high accuracy and low coverage is a well-run business nobody can find — the work is visibility. A 58 built on high coverage and low accuracy is a business being found and misdescribed — the work is urgent and much faster to complete.
This is why we refuse to report the headline number on its own, and why the four components appear on every page of every report.
Bands are descriptive, not graded. A 55 is not a failure — it is the normal starting point for a business that has never worked on this.
| Band | Score | What it looks like in practice |
|---|---|---|
| Absent | 0–39 | Rarely or never named. Often accompanied by factual errors nobody has noticed. Most of the available gain is here. |
| Marginal | 40–59 | Appears occasionally, usually for direct name searches rather than category questions. The most common starting band. |
| Present | 60–74 | Reliably named for some category questions. Competitors still take the majority of the shortlist positions. |
| Strong | 75–89 | Named in most relevant answers, described accurately, well corroborated. Work shifts from fixing to defending. |
| Dominant | 90–100 | Consistently the first or only business named. Rare, and generally only sustainable in a narrow specialty or a small market. |
The score is only as good as what it asks. This is the part most scoring systems gloss over.
Questions are written the way a customer would ask them, not the way a marketer would. “My boiler is leaking, who can come out tonight?” rather than “emergency plumber services.”
Enough to be stable across runs, few enough to re-run monthly at reasonable cost. The set is agreed with the client at the start and stays fixed — changing the questions changes the score, which would make the trend meaningless.
Urgent need, planned purchase, comparison, and direct-name lookup. A business can score well on one and badly on another, and the difference tells you where the work is.
AI answers vary between runs. A single query is noise. Each question is asked three times and the majority result is recorded, which removes most of the run-to-run variation.
Stated plainly, because a metric that claims to measure everything measures nothing.
Yes. The free scan returns a Search Presence Score and the four components. It uses a shorter question set than a paid engagement, so treat it as indicative rather than the full measurement, but the method behind it is the one on this page.
Because it would be trivially gameable and it would mislead you. A business appearing constantly with wrong information, or appearing only for searches of its own name, would score highly on a raw count while being in poor shape. The four components exist to stop a single easy number from flattering anyone.
Whatever beats the businesses named alongside you. That is what Share of Answer is for. An absolute score of 70 is excellent in a competitive city category and unremarkable in a rural one, so we report your score next to the competitors we track rather than against a national average that would mean nothing.
Partly, and we thought about that before publishing. But the components are things you can only improve by genuinely being more findable, more accurate and better corroborated. “Gaming” this score mostly means doing the work. We are comfortable with that trade.
Monthly, on the same question set, at roughly the same point in the month. Consistency of measurement matters more than frequency — measuring weekly would mostly produce noise and a larger bill.
Free, about a minute, and it comes back with all four components rather than a bare number.