Measuring AI visibility: share of model and citation rate
Google rankings say almost nothing about whether ChatGPT or Perplexity recommend you. Which metrics really capture AI visibility, which tools are worth it, and how to measure it yourself in 5 steps.
Measuring AI visibility means checking how often, in which context, and with which tone ChatGPT, Perplexity, Gemini, and Google AI Mode name your company when someone asks about your service. It is a discipline of its own, because the classic SEO metrics are blind to it: according to an Ahrefs analysis of 15,000 queries, only 12 percent of URLs cited by AI sit in Google’s top 10.
So whoever ranks at position 3 can still be missing from the AI answer. And whoever is missing there does not exist for a growing share of B2B research. This article shows you the four metrics that count, the tools we use, a 5-step measurement method to do it yourself, and how we track it monthly for clients.
Key takeaways
- Google rankings and AI mentions run largely separately: only 12 percent of AI-cited URLs sit in Google’s top 10 (Ahrefs, 15,000 queries).
- The AI systems also differ among themselves: only around 11 percent of domains are cited by both ChatGPT and Perplexity (5W Research 2026).
- The four metrics: share of model (share of mentions), citation rate (share with link), mention context (recommendation vs. mention), and sentiment.
- Measurement works with Ahrefs Brand Radar, manual prompt spot checks, and Bing Webmaster Tools. Without a fixed prompt set and competitor benchmark, every number is worthless.
- AI answers are volatile: in September 2025, Reddit’s share of ChatGPT citations fell from 60 to 10 percent within weeks. Measuring once is not enough.
Why do Google rankings say nothing about AI visibility?
Because AI systems select their sources differently from how Google ranks. In 2025, Ahrefs compared for 15,000 queries which URLs AI tools cite and which sit in Google’s top 10: only 12 percent overlap. The remaining 88 percent of AI citations come from pages that do not appear on page 1 at all.
On top of that, the AI systems barely agree among themselves. According to the 5W report "State of AI Citations 2026", only about 11 percent of domains are cited by both ChatGPT and Perplexity. ChatGPT accesses the web via Bing and cites Wikipedia heavily, Perplexity prefers Reddit, G2, Gartner, and comparison portals, Google AI Mode leans on LinkedIn, YouTube, and its own services.
And one more thing ranking tools do not see: according to 5W, ChatGPT names brands around 3.2 times more often than it links them. So most recommendations come without a click, without a referrer, without a trace in your analytics. Whoever only watches Search Console and rank trackers is measuring a channel that says little about their customers’ AI research. How SEO and GEO differ fundamentally is in the article on generative engine optimization.
Which metrics measure AI visibility?
AI visibility can be captured with four metrics: share of model, citation rate, mention context, and sentiment. Each answers a different question, and only together do they form a picture.
Share of model
Share of model is the share of AI answers on your core topics in which your company is named, relative to your competitors. The counterpart to share of voice from classic brand measurement. Example: you ask 50 prompts about "web agency for B2B software" in ChatGPT. If you are named in 8 and your strongest competitor in 22, your share of model is 16 percent versus 44. This number is the most important one because it is relative. Absolute mentions fluctuate with every model update; the ratio to competitors stays comparable.
Citation rate
The citation rate measures how often the AI not only names you but links a page of yours as a source. That is the part that brings traffic and that you find in analytics under referrers like chatgpt.com or perplexity.ai. A high mention rate with a low citation rate means: the AI knows you from third-party sources (directories, press, reviews), but your own content is not citable enough. That is a content problem, not an awareness problem.
Mention context
Not every mention is worth the same. "INSYNC is a web agency from Bavaria" is a mention. "For B2B companies with high design standards, INSYNC is recommended" is a recommendation. And whether you are named as the first, third, or eighth option decides whether the reader even gets to you. We categorize every mention into three levels: recommendation, neutral mention, exclusion ("less suitable for"). The position in the list is added as a fourth dimension.
Sentiment and accuracy
AI systems do not form their own opinion about you; they reflect the tone of their sources. If the cited sources contain old criticism, an outdated service portfolio, or a long-resolved incident, that ends up in the answer. 5W describes this as "durable crisis memory": AI models cite documented crises in the present tense years later. That is why measurement also includes the question: is what the AI says about you correct? Wrong prices, wrong locations, wrong services are visibility that hurts you.
The four metrics at a glance
| Metric | Question | Measurement |
|---|---|---|
| Share of model | How often am I named compared with competitors? | Prompt set, count mentions, calculate share |
| Citation rate | How often is my page linked as a source? | Brand Radar, analytics referrers, Bing Webmaster |
| Mention context | Recommendation, mention, or exclusion? At which position? | Manual categorization per answer |
| Sentiment | How does the AI talk about me, and is it correct? | Manual review, trace the answer’s sources |
Which tools measure AI visibility?
Three tools are enough for measurement: Ahrefs Brand Radar for ongoing tracking, manual prompt spot checks for context and sentiment, and Bing Webmaster Tools for ChatGPT’s access to your site.
Ahrefs Brand Radar tracks mentions and citations of your brand in ChatGPT, Perplexity, Gemini, Google AI Overviews, and AI Mode across millions of prompts. You see share of voice against competitors, cited pages, and the trend over months. It is the tool we work with because it sits in the same system as keywords and backlinks. Alternatives with a similar principle: Semrush AI Toolkit, Peec AI, Profound.
Manual prompt spot checks do not replace a tool, but no tool replaces them. Only whoever reads the answer themselves sees whether the mention is a recommendation or a side note and whether the facts are correct. 20 to 50 prompts a month, in every relevant system, logged in a spreadsheet.
Bing Webmaster Tools are the underrated building block. ChatGPT searches via Bing. If your site is poorly indexed in Bing, ChatGPT cannot cite it, no matter how well it ranks on Google. Bing index coverage is thus a direct precursor of the citation rate.
Analytics deliver the flip side: referrers from chatgpt.com, perplexity.ai, copilot.microsoft.com, and gemini.google.com show how much traffic actually arrives. For most B2B sites this is still under 3 percent today, but with a noticeably higher conversion rate than Google traffic, because pre-qualification already happened in the AI answer.
Define the prompt set
Collect 20 to 50 questions your customers would ask before buying: "Which agency for X?", "Best providers for Y in Germany", "X vs. Y". Mix recommendation, comparison, and knowledge questions. This set stays fixed over months, otherwise the numbers are not comparable.
Set the competitors
Choose 3 to 5 competitors against which you calculate share of model. Take the ones your customers actually compare, not the ones you would like as competition.
Run spot checks in all systems
Ask every prompt in ChatGPT, Perplexity, Gemini, and Google AI Mode, each in incognito mode without login history. Log: mentioned yes/no, position, context (recommendation/mention/exclusion), linked source, tone, factual errors.
Set up tool tracking
Set up your brand and competitors in Ahrefs Brand Radar and add Bing Webmaster Tools for index coverage. The tool delivers breadth, your spot checks deliver depth.
Evaluate monthly and trace sources
Calculate share of model and citation rate per system, compare with the previous month. Most important step: for every answer where a competitor sits ahead of you, check which sources the AI cited for it. Those sources are your to-do list.
How does INSYNC measure AI visibility?
We measure AI visibility for our clients monthly with a fixed prompt set, Ahrefs Brand Radar, and manual spot checks in four systems. The result lands in the same report as the Google rankings, because only both together show where a company is really visible.
Concretely, it works like this: in the setup, we define 30 to 50 prompts per client, derived from the keyword strategy and the real customer questions from sales calls. Plus 3 to 5 competitors as benchmark. From then on, we track share of model, citation rate, and mention context per system monthly, plus Bing index coverage and the AI referrers from analytics.
The part that makes the difference is the source analysis. For every answer where a competitor sits ahead of our client, we pull the cited sources: directories, trade media, comparison sites, Reddit threads, LinkedIn posts. From that comes the list of places where the client needs to be present. That is the core of our GEO work: not guessing which content AI systems like, but replicating what is demonstrably cited.
An honest note on interpretation: AI answers are volatile. In September 2025, Semrush observed how Reddit’s share of ChatGPT citations fell from around 60 to 10 percent within six weeks. A single monthly value says little. The trend over three to six months and the gap to competitors are the numbers we look at.
What to do when the AI does not name you?
If the measurement shows you are missing, it is almost always due to missing third-party sources, not your website. According to 5W, brand search volume is the strongest known predictor of AI mentions, with a correlation of 0.334, noticeably stronger than backlinks. AI systems recommend what they find confirmed in several places.
The order that works for us: first, make your own content citable, with direct answers in the first paragraph, clear question headings, and structured data. What that looks like is in the article on LLM SEO. Second, cover the sources the AI cites for your competitors: industry directories, comparison portals, trade media, review platforms. Third, make the brand itself better known, because search volume for your name directly increases the probability of being mentioned.
And then: keep measuring. Without a baseline you do not know whether it works. Whoever wants to know where they stand today starts with a GEO audit from 1,500 euros, which runs exactly this measurement once completely and delivers the source list.
Frequently asked questions
AI visibility describes how often and in which context AI systems like ChatGPT, Perplexity, Gemini, or Google AI Mode name your company or cite it as a source when users ask about your service. It is independent of Google rankings: according to Ahrefs, only 12 percent of AI-cited URLs sit in Google’s top 10.
Share of model is the share of AI answers in a topic area in which your brand is named, relative to competitors. If you are named in 8 of 50 answers and your competitor in 22, your share of model is 16 versus 44 percent. The metric is relative and therefore more stable than absolute mentions.
With a fixed set of 20 to 50 customer questions that you ask monthly in ChatGPT in incognito mode and log: mentioned or not, at which position, as recommendation or mention, with which source. For breadth, add a tool like Ahrefs Brand Radar, which tracks mentions across millions of prompts.
Ahrefs Brand Radar, Semrush AI Toolkit, Peec AI, and Profound track mentions and citations in ChatGPT, Perplexity, Gemini, and Google AI. Add Bing Webmaster Tools, because ChatGPT searches via Bing, and the referrer reports in your analytics. No tool replaces manual spot checks for context and sentiment.
Because AI systems select their sources differently. They weigh third-party sources like directories, comparison portals, trade media, and Reddit more heavily than your own website, and ChatGPT searches via Bing instead of Google. According to 5W Research, brand search volume is the strongest predictor of AI mentions, stronger than backlinks.
Monthly, with the same prompt set. AI answers are volatile: in September 2025, Semrush observed Reddit’s share of ChatGPT citations change from 60 to 10 percent within six weeks. A single value says little; the trend over 3 to 6 months and the gap to competitors count.
More Articles
Let's talk about your project.
20-minute call, no sales pressure. You describe what you have in mind, we tell you if and how we can help.