How we score.

Every audit puts the questions your buyers ask to ChatGPT, Gemini, Perplexity and Google AI Overviews. The score, out of 100, is 70% visibility (how often you are named, and how high) and 30% recognition (whether each AI knows who you are and describes you correctly). Recognition only counts in full once AI names you in at least 50% of answers: being well known means little if you are never recommended.

We research the questions buyers in your category really ask before they choose, such as “best X for Y” or “X alternatives”, and ask every one on every engine. A client's questions are fixed from month to month, so two audits measure the same thing.

Each answer earns points by where it names you. Being recommended first is worth far more than being one name in a long list:

We average the points over all your questions on each engine, then average the engines. An engine that could not answer in an audit is left out rather than counted as zero.

We ask each engine what your business is and check its answer against your own website: does it know you, and is what it says correct? This part is scaled by how often you are named, reaching full weight at 50% of answers.

AI assistants do not give the same answer every time. Two audits of the same questions differ by a few points even when nothing has changed. We work out that normal range for every audit, and a change inside it is reported as “about the same”, never as progress or a drop.

A recommendation only counts as having worked when at least two engines moved the same way.

• Citations: answers that link to your website as a source.
• Sources: the other sites AI leans on for your questions, and the share of answers citing each.
• Share of voice: of every business named in your category, the share that is you.
• Tone: whether answers that name you describe you favourably, neutrally or critically.
• Competitors: who is named instead of you, local and international.

Questions about the method? Talk to Recvis.