Citino teamTo track brand mentions across ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews, ask every assistant the same fixed buyer questions on a schedule and record what each answer says. The routine is the same everywhere; what differs is where each assistant gets its information, and that changes how you read the results.
This is the core of AI search analytics. The routine below works for any number of assistants.
Why ordinary brand monitoring misses this
Social listening and media monitoring read what is published: posts, articles, reviews. A conversation between a buyer and an assistant is not published anywhere. If ChatGPT recommends a competitor to a buyer, no tool sees it happen.
The only way to know what an assistant says about you is to ask it the questions your buyers ask, the way they ask them, and keep the answers.
Step 1: Fix one set of questions for every assistant
Use the same questions everywhere, so the assistants can be compared: category, problem, comparison, alternatives and fact-check questions, worded as buyers word them. Keep branded questions in their own group. How to choose the prompts to track covers the method.
Step 2: Ask every assistant the same way
- A clean session. No memory, saved instructions or earlier messages; personal history changes answers. See are AI answers personalised?
- Fixed wording. Language models are sensitive to small changes in a prompt (Sclar et al.).
- A fixed schedule. Daily if you can, at the same time.
- The same market. Location and language change answers, so ask from where your buyers are.
Step 3: Record each answer the same way
For every question, on every assistant, every run, keep the full answer and these fields:
| Field | Why |
|---|---|
| Assistant | Each one is reported separately |
| Brands named, in order | Mentions and position |
| How you were described | Tone and accuracy |
| Pages cited, if any | Why the answer said what it said |
| Searched the web? | Whether the answer can change when pages change |
Step 4: Know where each assistant gets its answers
The assistants do not read the same things, which is why your mentions differ between them.
| Assistant | What its documentation says about sources |
|---|---|
| ChatGPT | Searches the web for some questions and answers with sourced citations |
| Google AI Overviews | May run multiple related searches across subtopics to build a response |
| Gemini | Google describes grounding in Gemini apps as providing content from the Google Search index to the model |
| Claude | Can search the web and provides citations when it uses web information |
| Perplexity | Searches the web for each question and gives numbered citations |
So a brand well covered in Google’s index may do well in AI Overviews and Gemini and still be absent from an assistant that leans on other sources. The per-engine guides go deeper: ChatGPT, Google AI Overviews and Perplexity.
Step 5: Count mentions per assistant
From the records, report for each assistant:
- Visibility: the share of answers that mention you.
- Share of voice: your mentions as a share of all brand mentions.
- Position: how early you are named, over answers that mention you.
- Citation rate: the share of answers that cite one of your pages.
Each with a margin of error calculated over questions, because answers to the same question tend to agree from day to day. A change smaller than its margin is within noise.
Step 6: Compare the assistants
Put the assistants side by side. The differences are where the work is:
- Named by one, absent from another: look at what each one cites. The sources one assistant reads mention you; the other’s do not.
- Named everywhere, but late: the sources mention you without making a case. Look at the reasons answers give for the brands named first.
- Named nowhere: start with where competitors beat you.
One routine, every assistant, the same questions. Without that, the differences between assistants are just differences in how you asked.
Where Citino helps
Citino is an AI search analytics tool. It asks ChatGPT, Google AI Overviews and Perplexity your buyers’ questions every day, with Gemini, Google AI Mode and Claude as add-ons, from a clean session with fixed wording. It keeps every answer with the brands named and the pages cited, and reports each assistant separately with a 95% margin of error, as the methodology sets out. See how Citino measures.