To track your brand in AI answers, write the questions your buyers ask, run each one on every AI engine every day, and record whether you are named, where, how you are described and which sources the engine cites. Then watch the trend over weeks, not any single answer. That loop is the whole method; the rest of this guide shows how to set it up.
It applies to ChatGPT, Gemini, Google AI Overviews and Google AI Mode, and it is the practical side of answer engine optimization (AEO).
What does it mean to track a brand in AI answers?
Tracking means asking the same set of questions on a schedule and keeping the history. For each question, on each engine, you record:
- Whether your brand is named at all
- Where it appears in the answer, if it does
- How it is described, including whether the tone is positive, neutral or negative
- Which competitors are named beside you
- Which sources the answer cites
A single answer is a snapshot. A daily history is a measurement.
Step 1: Write prompts the way buyers ask
A prompt is a full question in a buyer's words, not a keyword. "Payroll software" is a keyword. "What's the cheapest payroll software for a café with 12 staff?" is a prompt.
- Mix stages: discovery ("best X for Y"), comparison ("A vs B") and brand checks ("is A worth it?").
- Take wording from sales calls, support tickets and reviews.
- Leave your own brand name out of most prompts. If every prompt names you, you only measure people who already know you.
How to choose the prompts to track goes deeper on sourcing, volume and overlap.
Step 2: Group prompts into subcategories
Averages across everything hide where you actually win and lose. Group prompts into subcategories such as "pricing and value" or "multi-site groups", so you can see which topics you own and which ones rivals dominate.
Step 3: Check every engine, every day
The four engines to cover are ChatGPT, Gemini, Google AI Overviews and Google AI Mode. They do not always name the same brands, so keep their results separate.
Daily checks matter because answers vary. OpenAI's documentation describes model output as non-deterministic (OpenAI, text generation guide), and a January 2026 study from SparkToro and Gumshoe.ai found that repeated runs of the same brand-recommendation prompt almost never returned the same list (SparkToro). The study is not peer reviewed, but it points the same way as OpenAI's own note: one run is not enough.
Step 4: Take a baseline before you change anything
Record your numbers on day one. Without a baseline you cannot tell a real gain from a lucky day. Keep every later check, so you can compare a week or a month against where you started.
Step 5: Read the numbers
Start with three, per engine and per subcategory:
- Visibility: the share of checks where you are named at all
- Share of voice: your mentions as a share of all brand mentions
- Average position: where you appear when you do appear
Always keep the count beside the percentage, such as "named in 18 of 60 checks", so you know how much data is behind it. AI visibility metrics explains how to calculate each one and which to trust most.
Step 6: Read the sources to find out why
When a rival wins a prompt, open the sources the engine cited. Usually it is a comparison page, a review site or a forum thread that says something your own site does not. A source that cites your rivals and never you is a citation gap, and it is often the shortest route to being named. See AI citation gaps.
Step 7: Test wording, then act
The exact words in a prompt can change who gets named. Run a few variations of your most important questions to see how sensitive you are. Why AI answers change with wording shows how. Then fix your own pages first, close citation gaps next, and compare against the baseline.
Can you track a brand by hand?
Yes, for a handful of prompts. A spreadsheet works for a first look. It breaks down as the number of prompts, engines and days grows.
| Item | By hand | With Verte |
|---|---|---|
| Engines | One at a time, copied and pasted | ChatGPT, Gemini, Google AI Overviews and Google AI Mode |
| Prompts | A dozen or two before it becomes painful | Written for you from your name and website, editable, grouped by subcategory |
| Frequency | Whenever you remember | Daily, with results each morning |
| History | Whatever you saved | Kept automatically, from your first checks |
| Sources | Read every answer | Cited sources and citation gaps shown for you |
How does Verte track your brand?
Verte writes the prompts for your business from its name and website, and you can edit them. It checks them every day on the four engines and shows visibility, share of voice, position, sentiment and sources on one dashboard. Your first checks run when onboarding ends, so the baseline starts on day one.
- AI visibility tracking shows each number with its count.
- Citations and sources show which pages the engines cite.
- The Verte Engine answers questions about your results, such as which subcategory is weakest, and proposes changes. Nothing changes until you confirm.
Start tracking your business, or start as an agency to track every client. See current plans.