Reading time: 12 minutes
An AI search visibility score measures a brand’s presence in AI answers, usually on a scale from 0 to 100. Each tool calculates its own. At Semrush, the AI Visibility Score compares the brand with competitors, while at Profound it measures the share of AI answers with the brand. So scores only compare within the same tool.
In short
- An AI visibility score sums up a brand’s presence in AI answers, but each tool does its own math. Semrush, for example, doesn’t publish the formula for its AI Visibility Score.
- Profound, Peec AI and Otterly publish their formulas, but they describe different bases. So the same 10 answers can score 40% in Peec AI and 50% in the Profound Visibility Score.
- None of the seven tools we reviewed publishes a margin of error in its documentation. By our math, 25 daily prompts on one engine give a weekly margin of ± 7 to ± 20 points.
- A high score doesn’t guarantee a recommendation, because the math can include critical mentions and prompts with the brand’s name. So read the score alongside citations, sentiment and traffic.
What is an AI visibility score?
An AI search visibility score counts how many AI answers mention a brand, for a given set of prompts, engines and dates. The tool runs the questions, stores the answers and counts the mentions. Then it turns that count into a percentage or an index from 0 to 100.
As a result, the same brand gets different scores depending on four choices the tool makes:
- Prompts: the tool’s own question database, as in Semrush’s AI Visibility Score, or the list you write, as in Peec AI.
- Engines and country: which AIs count and where the question comes from.
- Brand detection: in the Semrush database, its own AI system recognizes the brand and variations of its name. In Otterly, however, the name and the variations you enter are what count.
- Competitors: in a relative score, such as Semrush’s AI Visibility Score, the list of competitors changes the result.
If GEO is new to you, start with our guide to what GEO is.
Are AI visibility index and GEO score the same thing?
Not necessarily, because these names mean different things in different tools. At Ahrefs, the AI visibility index is the prompt database that Brand Radar checks. At Semrush, however, the same name refers to a public study of the most mentioned brands by industry. In Otterly, the Brand Visibility Index is a chart with the presence and position of the top ten brands.
GEO score also has two uses. The GEOscore, from the Brazilian company MonitorCNPJ, blends AI, SEO and social media into a score from 0 to 100, with weights by industry. In checkers such as the one from Rankai, however, GEO score measures how ready a page is for AI search. In other words, it doesn’t measure presence in answers.
How does each tool calculate its visibility score?
Three tools publish the formula for their AI search visibility score: Profound, Peec AI and Otterly. Semrush describes what goes into the math, but it doesn’t show the weights. Ahrefs, on the other hand, doesn’t sum up visibility in a single score. The table sums up the official documentation.
| Tool | Score name | How it’s calculated |
|---|---|---|
| Semrush | AI Visibility Score, 0 to 100 | Topics with the brand and how often it’s mentioned, without the math |
| Profound | Visibility Score, in % | Answers with the brand ÷ answers with any brand |
| Peec AI | Visibility Score, in % | Answers with the brand ÷ all answers |
| Otterly | Brand Coverage, in % | Prompts with the brand ÷ prompts run |
| Ahrefs | No single score | Mentions, citations, impressions and share of voice, separately |
At Semrush, the help center says the AI Visibility Score combines two factors. The first is how many topics include the brand, compared with other domains. The second is how often the brand shows up in the answers for those topics. On another page, however, Semrush compares the brand’s mentions with the median of its top competitors, which Semrush picks itself. The toolkit’s modules are in our Semrush AI Visibility Toolkit review.
At Profound, in turn, the Visibility Score divides the AI answers that mention the brand by those that mention at least one brand. Peec AI divides by all answers, with or without a brand. Finally, Otterly counts each prompt once per engine and per day. However, it leaves out the runs where an AI Overview doesn’t appear.
At Ahrefs, Brand Radar shows the metrics separately, as our Ahrefs Brand Radar review explains. NotHuman, which publishes this blog, works in the same market, so we left it out of the table.
Why can’t you compare scores from two tools?
Because each tool picks its own prompts, engines, brand detection and denominator. Even with the same answers, the math changes the number. Take 10 answers: the brand appears in 4, and 2 others mention no brand at all. Peec AI gives 40%, but the Profound Visibility Score gives 50%, because it drops the 2 answers with no brand.
The Profound Visibility Score has also changed its base in the documentation: the earlier version divided by all answers, like Peec AI. That’s why you should save the current definition with each report.
Even the definition of a mention varies. Otterly counts at most one mention per prompt, engine and day, even if the name repeats. Semrush’s Prompt Tracking, by contrast, counts how many times the name appears. And Semrush’s AI Visibility Score doesn’t even use your prompts, because it comes from the company’s database.
Therefore, compare a score only with itself, in the same tool and with the same setup. To choose a tool by collection, country and export criteria, see our GEO tools comparison.
Why does the visibility score change from week to week?
The AI search visibility score changes because AI answers change with every run, and the tool only measures a sample. In addition, the tool itself changes its prompt database, its competitors and its brand detection. So part of the weekly swing is noise, not the effect of GEO work.
For the same prompt, the brands, their order and the sources change with every run, as our measurement protocol with repeated prompts shows. Peec AI gives the same warning about sources. Its documentation says AI models pick sources with some built-in randomness, so a source can appear one day and be gone the next. Its advice:
“Focus on trends over weeks rather than daily changes.”
Peec AI, documentation (2026)
The tool also changes on the inside. At Ahrefs, for example, a change to the database can raise every brand’s mentions at the same time. For that reason, check whether the tool or the setup changed before you celebrate a rise.
What is the margin of error of the score?
None of the tools we reviewed publishes a margin of error for its AI search visibility score. However, with the method from our measurement guide, you can estimate the margin for a presence rate, at 95% confidence. The table shows one week of daily collection on the entry plans of three tools.
| Daily collection for one week | Answers | Margin of error |
|---|---|---|
| 15 prompts on 4 engines, as in Otterly Lite | 420 | ± 5 to ± 13 points |
| 25 prompts on 1 engine, as in the Semrush toolkit | 175 | ± 7 to ± 20 points |
| 50 prompts on 3 engines, as in Peec AI Starter | 1,050 | ± 3 to ± 8 points |
The real margin falls within the range, since repeating the same prompt produces similar answers. So more days of collection only lower the bottom of the range, while more prompts lower the whole range. Because both weeks have error, the gap between them has to exceed about 1.4 times the margin. With 25 prompts, that means 10 to 28 points.
Does a high visibility score guarantee a recommendation?
No. The Profound Visibility Score and the Peec AI score count answers that mention the brand, without separating a recommendation, a comparison or criticism. For example, a brand mentioned as an option to avoid raises the score just the same. Likewise, prompts with the brand’s name tend to produce answers with it, which inflates the result.
That’s why you should read the score for unbranded prompts on its own. One example is “qual o melhor software de gestão para clínicas” (“what is the best management software for clinics”). In those prompts, the brand only appears if the AI remembers it.
Position in the answer is also left out of these formulas. Otterly, for example, crosses presence with average position in its Brand Visibility Index. In that chart, a brand with many mentions that usually ranks lower in the answer falls into the low conversion quadrant. That is, the AI mentions it, but not as a top choice. Sentiment is left out too: in Otterly, it’s a separate index, from -100 to +100.
What numbers should you track besides the score?
Track the numbers that explain the score and the ones it doesn’t cover. After all, the score shows presence, but it doesn’t say whether the site becomes a source, how the AI describes the brand or whether it sends visits. So the report needs at least four numbers next to it:
- Presence by engine and by prompt: the average hides the engines and topics where the brand disappears.
- Site citations: the share of prompts where the answer cites the brand’s site, such as Otterly’s Domain Coverage.
- Sentiment: whether the AI speaks well or badly of the brand, always with the same criteria.
- AI traffic: visits and conversions that come from assistants, in the AI channel in GA4.
The brand’s share against competitors, its share of voice, is another metric, with its own math in each tool.
How do you read an AI search visibility score in a report?
Read the score as a trend, not as a scoreboard. In other words, compare the score with its own history and with the competitors in the same report. Then subtract the margin of error and open the answers. A score of 32, for example, is neither good nor bad on its own, because it depends on the tool and the category.
- Write down the setup: tool, prompt database, engines, country and period. Without it, no one can redo the math the next month.
- Compare within the tool: in Semrush, the AI Visibility Score is already relative to the industry. So 32 says little without the competitors’ scores next to it.
- Look at weeks, not days: Peec AI recommends following the trend over weeks.
- Account for the margin: a change within the margin is noise until it repeats.
- Open the answers: check the prompts and the text behind the score. In the Semrush report, for example, each prompt opens the full answer.
Finally, when you show the score to a client, name the tool and the formula on the same page.
What we found in the documentation of seven tools
We read the official documentation of Semrush, Ahrefs, Profound, Peec AI, Otterly, Radar GEO and GEOscore on the visibility score. Then we counted what each one publishes about the math, the base of answers and the margin of error:
- Formula for the main score: 3 of 7, that is, Profound, Peec AI and Otterly. Even so, each describes a different base.
- Factors without the math: 2 of 7, Semrush and GEOscore, which describe what goes in, but not the weights.
- Neither formula nor factors: Radar GEO, from the Brazilian company RD Station, mentions presence and sentiment, but doesn’t explain the math. Ahrefs, on the other hand, has no single score.
- Margin of error: 0 of 7.
- Versions of the same score: Semrush describes its AI Visibility Score in three ways, on three help center pages.
- Changed definition: Profound changed the base of its Visibility Score in the documentation, from all answers to answers with any brand.
In other words, anyone who gets a score without a formula can’t redo the math. So ask the vendor for the definition in writing before you use the score as a target.
Frequently asked questions
Yes, with limits. Semrush’s free AI Search Visibility Checker shows its AI Visibility Score and the brand’s mentions in ChatGPT, Gemini and Google’s AI. RD Station’s Radar GEO is also free and shows presence and sentiment in ChatGPT and Gemini. In both, however, you enter the brand or the site, not the prompts.
No. An AI search visibility score counts mentions in answers, not visits. Next to the score, Semrush shows an estimated audience for the topics where the brand appears, which isn’t traffic either. To measure visits and conversions from assistants, use the AI channel in GA4.
There’s no universal good score, because each tool does its own math and each category has different competition. In the Profound Visibility Score, for example, 50% means presence in half of the AI answers that mention any brand. The benchmark is your own history and the competitors’ scores in the same report. Peec AI gives an example: 15% visibility means more when a competitor reaches 45% on the same prompts.
Want to know how your brand shows up in AI answers today?
NotHuman is opening spots for agencies and brands to track how they appear in AI answers. Join the waitlist and we’ll let you know when a spot opens up for you.
Main sources
- Semrush, how Semrush calculates the AI Visibility Score (2026).
- Semrush, the AI Visibility Toolkit and the score against the competitors’ median (2026).
- Profound, Visibility Score and the other Answer Engine Insights metrics (2026).
- Peec AI, visibility formula (2026).
- Peec AI, how to read the answers and daily variation (2026).
- Otterly, definitions of the brand report metrics (2026).
- Ahrefs, the AI visibility index behind Brand Radar (2026).
- MonitorCNPJ, GEOscore and how its Digital Score is calculated, in Portuguese (2026).
