AI visibility score

What does a single score hide?

The table at the top of this page is our own workspace, three of the 61 questions it tracks, read over the Attensira MCP for the 30 days to 4 October 2026. Blend those three rows into one score and Attensira is named in 65 of 237 answers, 27.4%. That number is true and useless. No question behaves like it:

• When a buyer asks about us by name, we are there. "Attensira vs Profound" named us in 98.5% of answers and linked our site in 61 of 66. Nothing to fix.
• When a buyer asks for the category, we are absent. "free AI visibility checker" and "Peec AI alternatives" named us in none of 171 answers, while Ahrefs and Peec AI owned them. The fix is pages and placements for those questions.

A score says neither. Across all 61 questions the blend falls to 4.35% of 4,524 answers, which sounds like a mediocre brand rather than a strong branded presence and an empty category one. And a score can rise while the thing you care about falls: more citations and fewer mentions can average out to "up".

What is an AI visibility score?

It is a vendor's composite. Most blend some of: the share of answers naming the brand, the share linking its domain, where the brand appears in the answer, and sometimes sentiment, then rescale to a friendly range. The formula, the questions it is computed over, the engines and the countries are all the vendor's choice.

That makes the number internally consistent at best. It is not comparable across tools, and often not across time within one tool when the question set or engine list changes.

Why is there no standard visibility score?

Because every input is a choice. Which questions, phrased how? Which engines, read through the consumer app or the API? Which countries? How many readings per question per day? How are mentions and citations weighted? Does a competitor list affect the denominator? No standards body has settled any of these, and the engines themselves publish nothing to calibrate against. Google reports AI feature traffic inside Search Console's ordinary Web totals and, since 31 August 2026, impressions for AI Overviews and AI Mode in its generative AI report. ChatGPT, Perplexity and Claude publish nothing comparable.

What should you measure instead?

Separate rates, each with its sample size:

• Share of voice

  Question it answers

  In what share of answers were you named?

  Glossary

  Share of voice

• Citation rate

  Question it answers

  In what share of answers was your domain linked?

  Glossary

  Source citation

• Position

  Question it answers

  Where did you come among the names you track?

  Glossary

  Citation position

• Sources

  Question it answers

  Which domains did the answers lean on?

  Glossary

  AI content ranking

• Sentiment

  Question it answers

  When an answer judged you, was it favourable?

  Glossary

  Sentiment (docs)

Attensira reports these and no composite. Its documentation is explicit: there is no visibility score in the product, and "visibility" is only a label on a section of the app.

What are the objections to dropping the visibility score?

"Executives want one number." Give them two, with n: named and linked. They are just as short, and each one tells someone what to do.

"Our score went up 8 points." Ask what moved, on which engine, and over how many answers. If nobody can say, nothing was learned.

"Scores let us compare against competitors." Only inside one tool, and only if the competitor list does not change the denominator. Attensira's share of voice divides by all answers, so adding a competitor never moves your number.

"A score is easier to put in a report." A report that cannot be acted on is decoration. Separate rates are as easy to chart.

How does Attensira report AI visibility?

Attensira reads your questions on each engine every day and reports share of voice and citation rate separately, per engine and per country, each as a value with the number of answers behind it, and it tests every change for significance before calling it real. Its crew then works on whichever gap the numbers show, ships each change after your approval, and measures before and after. Our methodology sets out how each rate is computed and when a change counts as real.