What is an AI Score, and how is it calculated?

The AI Score is a single number, from 0 to 100, that measures how findable and recommendable a local business is to AI assistants like ChatGPT, Perplexity, and Google’s AI Overviews. It is the metric we built Local AI Registry around, and we publish exactly how it is calculated. A high score means AI can find you, trust you, and name you when a customer asks; a low score means it cannot.

The number matters because the results are stark. Across the local businesses we have scored, nearly two in three land below 45 out of 100, and barely one in seventy-five clears 65. Most local businesses are not close to being recommendable by AI, and most owners have no idea where they stand.

The AI Score exists to turn that invisible problem into one number you can see, act on, and track. It is the answer to the question every owner now has to ask: does AI recommend me, and why not?

The AI Score is not a black box or a single-metric gimmick. It is built from eight signals, each measuring a different part of whether an assistant can find and recommend you. Every weight is documented in our scoring methodology.

• Retrieval grounding: can AI find and cite you for the queries that drive bookings, not just your brand name.
• Off-property attestation: what independent sources across the web say about you.
• AI visibility: whether specific assistants already name you today.
• Reviews: volume, recency, rating, and whether you respond to them.
• Google Business Profile: how complete and accurate your listing is.
• Authority: the quality of the sites that link to and reference you.
• Speed and mobile: whether your site is technically eligible to be cited.
• Content and schema: whether your pages are clear and machine-readable.

Each signal is scored on its own, then combined into the final 0-to-100 number using a weighted geometric mean rather than a simple average. The distinction matters. A geometric mean leans on your weakest signal, so a business cannot paper over a missing off-property presence with a fast website. If any load-bearing signal is near zero, the overall score stays low, which mirrors how an assistant actually behaves: one strong attribute does not earn a recommendation when a critical one is absent.

The weights come from how much each signal moves a real AI recommendation, not from what is easiest to measure. That is why two businesses with identical websites can score very differently: the one the wider web corroborates scores higher, because corroboration is what the assistant grounds its answer in.

Because most local businesses score low, the bar to stand out is lower than it looks. Roughly:

• A score of 65 or above is strong: AI can reliably find and recommend you. Very few businesses are here yet.
• A score in the 45 to 64 range is fair: you are eligible and sometimes named, with clear gaps to close.
• Below 45 is weak: AI struggles to find or trust you, and a competitor is getting the recommendation instead.

The point of the number is not the grade, it is the diagnosis. Your AI Score comes with the specific signals dragging it down, so you know what to fix first.

No business can buy a higher AI Score, and no one can pay to be removed from a competitor’s comparison. We never sell rank, and we publish the method so anyone can check our work. That independence is the whole point: a score you could buy would tell an assistant nothing, and it would tell you nothing either.