Last verified: 3 November 2026
To track brand mentions across ChatGPT, Gemini and Perplexity, run a fixed set of buyer prompts through each engine on a regular schedule and record three things every time: whether your brand is named, which competitors are named, and which URLs are cited. Analytics referral data alone cannot do this, because most AI answers are never clicked and much of the traffic that is clicked arrives unlabelled.
Key takeaways
- Answers vary between runs and modes: only 25.6% of cited domains appeared in both of two ChatGPT reasoning modes for the same prompt (Semrush, June 2026).
- Referral data is unreliable: from 9 October 2026 many ChatGPT links arrived without their tracking tag, and LLM Pulse measured a 46% fall in ChatGPT sessions in GA4 over three days while total traffic stayed flat (LLM Pulse, October 2026).
- The cited URL is the most useful thing to record. It tells you which page you need to out-answer.
- Platform behaviour shifts quickly: Reddit’s share of ChatGPT citations fell from 3.8% to 0.5% in weeks in August 2026 (Semrush).
Why can’t Google Analytics measure this?
Because the outcome you care about usually happens without a click. A buyer asks an assistant, reads a short answer that names three brands, and moves on. Nothing reaches your analytics. When someone does click, the referral may arrive without a source, as the October 2026 ChatGPT change showed. Analytics tells you about a fraction of the visits AI answers send, and nothing about the answers where you were left out.
What should you track?
- Mentions: how often your brand is named in answers to your prompts.
- Share of voice: your mentions as a share of all brand mentions in those answers, against named competitors.
- Cited pages: which URLs the engines quote, yours and everyone else’s.
- Engine and mode: results for ChatGPT, Perplexity, Gemini, Google AI Overviews and AI Mode kept separate, because each retrieves differently.
- Trend: the same prompts on the same schedule, so changes reflect the answers rather than the questions.
How do you choose the prompts?
Start from the questions buyers ask before they choose a provider in your category, not from your keyword list. Include generic questions (“best agency for…”), comparisons, cost and process questions, and questions about trust and regulation. Tag each prompt by topic and intent so you can see where you win and where you are absent. Keep the set stable once it is running; rewriting prompts every month breaks the trend.
Which tools do we use?
We measure AI visibility with dedicated tracking tools, including Ahrefs Brand Radar and Peec AI, which run prompt sets across the major engines and record mentions, share of voice and citations. For clients, the results appear in the AI Brand Radar inside AgencyMindHub, on the same period picker as organic search, so AI visibility is reported next to the rest of your performance rather than in a separate tool.
The Stelrad figures in our GEO case study, 2,389 AI citations across 405 pages and a 62% share of voice in AI Overviews, were measured in Ahrefs Brand Radar in exactly this way.
What do you do with the results?
- Find the prompts where competitors are named and you are not.
- Look at the cited URLs for those prompts and decide whether to out-answer them on your own site or correct what they say about you.
- Change the pages, then watch whether the cited URLs move to yours over the following runs.
The method for becoming the cited page is in how to get your brand mentioned in ChatGPT. If you want a baseline for your own brand, talk to our AI visibility team.
