How brands measure how often AI answers mention them
Brands count the share of AI answers to a fixed set of questions that name or cite them, and tracking tools rerun those questions every day.
Brands measure their visibility in AI answers as the percentage of answers to a fixed set of questions that mention or cite the brand. Tracking tools put the same questions to ChatGPT, Google AI Overviews, Perplexity and other engines every day, because one answer changes from run to run. A team without a tool can run a smaller version of the check by hand once a month.
TL;DR
- Visibility in AI answers is the percentage of answers to a fixed set of unbranded questions that mention or cite a brand.
- Peec AI and Profound divide by different totals when they calculate a visibility score, so the two tools can report different numbers for one brand.
- Marketing leads should track one fixed question set per engine, read the trend over several weeks, and check GA4 for visits from AI assistants.
Definition and calculation
AI visibility is a brand metric that gives the percentage of AI answers to a fixed set of questions that mention or cite the brand. Peec AI’s documentation calculates its Visibility Score as the number of responses that mention the brand, divided by all responses, multiplied by 100. Peec AI is a tracking tool that sends a project’s prompts, the questions it follows, to several AI engines.
Profound, another tracking tool, divides by a smaller total. Profound’s glossary sets the brand’s mentions against only the responses that name at least one brand. Profound runs each prompt once a day on each engine it tracks. Answers that name no brand at all lower a Peec AI score and leave a Profound score unchanged, so one brand can score differently in the two tools.
The number is the baseline for any work on generative engine optimization, which aims to get a brand named and cited in those answers.
Share of voice, position and sentiment
AI share of voice measures a brand’s portion of all the brand mentions in the answers, so it compares the brand with the competitors a team chooses to track. Peec AI’s metrics overview counts it as the brand’s mentions divided by the mentions of every tracked brand, as a percentage. Adding a competitor that the answers often name therefore lowers the figure, even when the answers stay the same.
The same overview defines two more brand metrics:
- Position: Peec AI averages the brand’s rank across the answers that name it, so 1 means the brand always came first and lower is better.
- Sentiment: Peec AI scores how positively the engines describe the brand on a scale from 0 to 100.
Peec AI also separates brand visibility, where the answer names the brand, from source visibility, where the answer uses or cites the brand’s pages without naming it. A brand can be named without its site being used, and its site can be used without the brand being named, so a useful report shows both.
How tracking tools collect the answers
AI search visibility measurement for GEO agency reporting on Google AI Overviews, ChatGPT and Perplexity relies on visibility tracking tools, which agencies in 2026 can set to rerun questions daily. Peec AI’s guide to its chats says each chat is an answer produced by running a project’s prompts every day. It adds that AI models pick sources with some randomness, so a page can appear one day and be gone the next, and it advises reading trends over weeks.
AI answers can change each time someone asks, so a single run is a sample of one. Si Quan Ong, SEO and marketing educator at Ahrefs, wrote on 22 October 2026: “AI responses vary, so groups of relevant prompts are usually more useful than individual answers.”
Each chat in Peec AI records two lists of pages. Sources are every URL the model opened while it wrote the answer, and citations are the sources the answer text refers to. An answer might cite 5 pages and show 8 in its sidebar. ChatGPT lists no sources when it answers without a web search, which where ChatGPT search gets its information explains.
Measuring without a tracking tool
A brand can measure without a tracking tool by asking each engine the same unbranded questions once a month and recording which answers name it and which pages they cite. The questions should be the ones buyers ask before they know the brand, such as which supplier to hire for a given job. A question that names the brand gets a mention by default, so it measures nothing.
Keep the wording identical every month, because a reworded question is a different question. For each answer, note whether the brand appears, where it sits in any list, and every page the answer cites. Visibility is the number of answers that named the brand, divided by all answers, times 100, worked out for each engine.
The cited pages show whether the engines rely on the brand’s own site, and how often AI answers cite a brand’s own website gives a benchmark for that share.
A monthly check by hand is one answer per question per engine, so it shows direction and misses day-to-day swings. The steps for reading the sources behind each answer are in how to check whether an AI assistant has cited your site.
What Search Console and GA4 show
Search Console and GA4 count the clicks and visits that AI answers send to a site, and an answer naming a brand without a click leaves no record in either. Google’s documentation on AI features, last updated on 10 December 2025, says Search Console counts sites shown in AI Overviews and AI Mode with other Google Search traffic. Those clicks sit in the Performance report’s “Web” search type, mixed with clicks from ordinary results.
As of 29 October 2026, Google Analytics’ default channel definitions include an AI Assistant channel for visitors who arrive from sources such as ChatGPT, Gemini, DeepSeek, Copilot or Grok. That channel leaves out Google’s AI Overviews and AI Mode, so visits from those two never appear in it. Teams that track ChatGPT and AI assistant traffic in GA4 begin with that channel.
Which pages Google uses for its own AI answers is the subject of what Google AI Overviews are and how they choose pages.
What it means
Marketing leads who want a reliable number should set up one fixed question set and read it the same way every month, in this order.
- Write the questions your buyers ask before they know your brand, and freeze their wording.
- Run the questions on the engines your buyers use, daily in a tracking tool or monthly by hand.
- Record visibility, share of voice and the cited pages for each engine separately.
- Compare months, and act only on a change that holds for several weeks.
- Check GA4’s AI Assistant channel and Search Console’s Web totals for the visits those answers send.
FAQ
How many prompts should a brand track?
Tracking tools size their plans in the tens to hundreds of prompts. As of 29 October 2026, Peec AI’s pricing page lists three monthly plans with 50, 150 or 350 prompts, each tracked daily on three models the buyer chooses. Its own example shows the scale: 25 prompts on 3 models for 30 days produce 2,250 answers. A check by hand works with fewer questions, since each one is asked and recorded on every engine.
How often should a brand read the number?
Once a month is enough for a report, as long as a change counts only when it holds for several weeks, since daily figures move with the engines’ randomness. A new page also needs time before it shows up in answers at all, which how long GEO takes to show up in ChatGPT answers measures. Judge a new page after a month at the earliest.
What is a typical score?
A typical score is low, and it depends on the category, the market and the question set. Agenzy’s 2026 study of 187,810 AI answers found the median European SMB brand appears in just 5% of AI answers about its own product category (n=90 tracked market projects, 2026-04 to 2026-08). Raising the number in ChatGPT is the subject of how to get a brand cited in ChatGPT answers.
Sources
- docs.peec.ai/metrics/…/visibility
- docs.tryprofound.com/getting-started/glossary
- docs.peec.ai/metrics-overview
- docs.peec.ai/understanding-chats
- ahrefs.com/blog/ai-share-of-voice
- developers.google.com/search/…/ai-features
- support.google.com/analytics/…/9756891
- peec.ai/pricing
- agenzy.lt/blog/ai-search-visibility-study-2026
The week in AI search, in one email
One email a week with what changed in ChatGPT, Google AI Mode, Perplexity and Gemini, and the week's data piece.