How the score works

1. Which AI we actually ask

This is the honest part that most tools leave out. The ChatGPT API with web search is not the ChatGPT app you use. They are different systems and they sometimes answer differently. We read the surfaces below, we label every result with the one it came from, and we never claim to be reading the consumer app.

Google AI Overviews is the exception, and it is the interesting one: that is captured from a live search result page, so it is literally what a person in your town sees today.

There is no official way to read the consumer apps, and scraping them breaks constantly and against their terms. Some tools do it anyway. We would rather label the surface honestly and publish how closely it agrees.

2. AI answers are not repeatable

Ask the same question twice and you can get two different answers. That is how these models work, and any tool reporting a single run as a precise fact is overstating what it knows. So you can ask each question several times and we report the spread rather than pretending to a precision we do not have. Repeats cost a credit each, which is why they are off by default.

3. What each result is worth

Every question × engine gives one of four verdicts:

Being recommended third is worth less than being recommended first, so points decay by 0.85 for each place below first, so third place is worth 0.85² of first.

4. Readiness, and what each check is worth

Readiness is a weighted average across the checks that apply to you. A pass scores full, a warning scores half, a failure scores nothing, and a check that does not apply is left out of the sum entirely rather than counted against you.

A Bing indexation check is planned but does not run yet. Until it does, it carries no weight in any score and is not listed above.

When we add a check, the check registry version goes up. Scans already run keep the score they were given. Your next scan is scored over the longer list, so readiness can move with nothing changed on your site, and History shows the added checks as new rather than subtracting the two scores. Version 6 added share previews, page titles and summaries, page structure, and complete business details. Version 7 tightened how FAQ pages, sitemaps, training crawlers and llms.txt are judged, so a site is no longer told to add what it already has.

4b. Local presence, for businesses customers choose by where they are

Local presence is a separate score covering your Google Business Profile. It is kept apart from readiness on purpose: one is your website and one is an asset that lives on Google’s platform, and adding a photo to a listing should not move the score that describes your site. It is scored the same way: a weighted average, a pass full, a warning half, and anything that does not apply left out of the sum rather than counted against you. Local registry version 2.

Three of these need your permission before we can check them, because Google will only tell the account that owns a listing whether its reviews have been replied to. Until you connect it, those three are excluded from the sum, not counted as failures. And if you have told us customers do not choose you by location, none of this applies and no score is shown at all; an absent score and a zero mean different things here, and a zero means we searched and found no listing.

5. The overall score

overall = 0.6 × visibility + 0.4 × readiness

Visibility carries the larger share deliberately: being recommended matters more than being technically tidy. Readiness is what you can change this week, which is why it is on the same scale rather than hidden.

6. A worked example

Eight results across five engines, and four checks. Every figure below is produced by the same code that scores your scan, so you can add them up.

An eighth result, Google AI Overviews, showed no AI answer, so it is left out entirely. The seven that answered give 7.99 points out of a possible 21, which is a visibility score of 38.1.

This example business is online-only, so the name, address and phone check does not apply and its weight leaves the sum. The remaining three give a readiness score of 56.5.

7. Freshness, and what we reuse

Questions that name no business (“best plumber in Chicago”) have the same answer for everybody who asks them, so we ask once and share the answer. That is what makes the product affordable, and it is why most of your questions cost you nothing.

Questions naming a business are never shared. Not yours, not a competitor’s. They are asked fresh, for you, and never contributed to anyone else’s results.

Shared answers are re-asked on this schedule. Past half the window a result is shown with a reduced-confidence marker rather than presented as current, and you can always force a fresh answer.

We also fingerprint your site. If you change it, any result that predates the change is flagged as such rather than shown as though it reflected your new pages, which matters most at exactly the moment you have just fixed something.

8. What a score does not mean

A 62 is not a grade and it is not a ranking. It says: across the questions you chose, on the surfaces listed above, on the day we asked, this is how often you were named, recommended or cited, weighted against how readable your site is to the engines. Ask different questions and you get a different number, which is why the questions are yours to edit and are listed on every report.