How to Monitor Brand Mentions in ChatGPT (By Hand or With a Tool)

Looks simple: type your brand's category into ChatGPT, see whether your brand shows up, done. The catch is that ChatGPT doesn't give one fixed answer. Ask the same question twice and you may get two different lists of brands.

To monitor brand mentions in ChatGPT in a way you can put in front of a client, you need to repeat the question, count the results and keep a record. This guide shows how to do that by hand, where that stops working, what to look for in a tool, and how to do it with Aivting. By the end you can decide which route fits your workload and what to record either way.

What it means to monitor brand mentions in ChatGPT

To monitor brand mentions in ChatGPT is to run a fixed set of prompts on a regular schedule, record which brands each answer names, and compare the results over time. Three things get recorded: whether your brand is mentioned, where in the answer it appears, and which sources the answer cites.

The unit of measurement is the answer, not the day. One prompt run on one day produces one answer. Monitoring means collecting many of them, then counting. A figure like "named in 12 of 40 answers" can be checked. "ChatGPT recommends us" cannot.

Why a single ChatGPT check isn't enough

A single check is a sample of one, and ChatGPT answers vary from run to run. A brand named first in the morning can be missing in the afternoon. The way you phrase the prompt matters too: "best project tools for small agencies" and "what should an agency use to manage projects" can return different brands.

You've probably heard that ChatGPT can now search the web and cite sources, which the OpenAI Help Center describes in its guide to ChatGPT search. True, but that doesn't make answers fixed: the sources cited can also differ between runs.

This is the core reason AI search monitoring is a counting exercise. If you only look once, you can't tell a stable mention from a lucky one.

How to see brand mentions in ChatGPT by hand

If you have a handful of prompts and one brand, doing it by hand is a fair place to start. It costs nothing, and it teaches you what the answers look like. Here is the routine.

Step 1: Write the prompts your buyers ask

Write 10 to 20 prompts in the words a buyer would type. Mix two kinds. Non-branded prompts ask about the category ("what's the best accounting software for freelancers?"). Branded prompts include a brand name ("is Ostrova good for freelancers?"). Keep them in separate lists, because branded prompts name you far more often and will inflate any combined figure.

For example, an agency tracking a fictional invoicing brand called Ostrova might use five non-branded prompts, three comparison prompts against Corvane (also fictional), and two branded prompts.

Step 2: Run each prompt several times in fresh chats

Open a new chat for every run, and use a temporary chat or turn off memory, so earlier conversations don't shape the answer. Run each prompt several times. Five runs per prompt is a workable minimum for a manual check. Note the date, because answers drift over time.

Step 3: Log mentions, order, competitors and sources

Build a sheet with one row per answer. Useful columns: date, prompt, run number, brand named (yes/no), position (1 for first named among the brands you track), competitors named, and cited sources. Position matters: being named first in a short answer is a different outcome from being fourth in a long one.

Here is what a filled-in row set looks like for the fictional brand Ostrova and its fictional competitors, one prompt run five times:

Ostrova is named in 3 of 5 answers for this prompt, and its average position across those three is 2. Run it once and you'd report either "always" or "never", depending on the run you happened to pick. The five-row version is the first one you can defend.

Step 4: Repeat on a schedule

Run the whole set weekly at a minimum. A one-off check shows a snapshot. A schedule shows direction. Keep the prompt list fixed between runs, or the comparison between weeks stops meaning anything.

Where manual monitoring breaks down

Manual checks do three things well. They're free, they're flexible, and they show you the actual wording of the answers. They struggle with everything else.

Volume. Ten prompts, five runs each, weekly, is 50 answers a week to run and log, per brand. Add three competitors and the logging grows with every column.

History. A sheet only tells you what you wrote down. It can't tell you what an answer looked like on a day you skipped.

No count you can defend. When two people run the checks, they phrase prompts differently and log differently. The result is a number nobody can reproduce.

Once your sheet needs a count, a trend and a competitor comparison, you are building the AI search visibility metrics KPIs by hand: visibility, share of voice, position and sources. That is the point where a tool earns its place.

Side by side, the two routes compare like this:

Neither route is wrong. The question is how many brands, prompts and weeks you need to cover, and whether anyone will still be logging in month three.

What a ChatGPT brand tracker should record

A ChatGPT brand tracker is software that runs your prompts on a schedule, stores each answer, detects which brands each one names and keeps the history. Whatever you choose, check that it records the following.

• Every answer, kept. You should be able to open the answer behind any number.
• The brands named, and in what order. Position inside the answer, not just yes or no.
• Competitors alongside your brand. The same answers, scored for each brand you track.
• Cited sources. The pages and domains the answer relied on.
• Counts with n. Every rate shown with the number of answers it rests on.
• History by day. So you can compare periods.

If you are shopping for a ChatGPT mentions monitoring tool, these six points are also a fair filter. Anything that shows a score without the answers behind it is asking you to take the number on trust.

Questions to ask before you choose one

Put these to any vendor, ours included.

1. Does it run each prompt more than once, and does it show how many answers a figure is based on?
2. Can you see the raw answer behind each number?
3. How many competitors can you track, and are they part of the same counts?
4. Does it show which sources are cited, and where you are missing?
5. Can you export the data, or produce a report with your own logo?

To be fair to the manual route: if you track one brand with a few prompts and keeping costs at zero matters more than history, your sheet may be enough for now.

How to monitor brand mentions in ChatGPT with Aivting

Aivting is built to do the repeating, counting and record-keeping for you. It collects answers to your prompts in runs and turns them into the numbers above [CADENCE: how often collection runs automatically]. [ENGINE AVAILABILITY: e.g. "It covers Google AI Overviews too, so you can compare the two."] This is how you set it up:

1. Create a project for one brand, and enter its website.
2. Add competitors. Aivting suggests some, and you can track up to 10 per project.
3. Add prompts. Type your own, or accept suggestions. Aivting suggests [PROMPT LIMIT] at a time, using your brand, competitors and category.
4. Start a collection run. Aivting collects answers to your prompts, so the history builds run by run [CADENCE: how often collection runs automatically].
5. Read the dashboard. The main screen is a daily graph, with visibility, share of voice and average position, each shown with its sample size.
6. Open the answers behind each number, and the sources they cite.

The steps for the first project are on the help page, create your first project. If you want to understand how a mention gets recognized in an answer, read how Aivting detects brand mentions.

Three things worth knowing. Collection is not real time. AI answers vary between runs, and Aivting shows the number of answers behind each figure instead of hiding it. And it measures and advises: it doesn't guarantee that ChatGPT will name or cite any brand. For what each plan includes, see plans and pricing.

What to do when ChatGPT doesn't mention you

Not being named is data, not a verdict. Start by checking n: if you've counted only a few answers, wait for more before drawing a conclusion. If the gap holds across enough answers, three places to look.

Gaps. A gap is a source that is cited in answers naming a competitor and in no answer naming you; your own site and your competitors' own sites are never counted as gaps. If a comparison site keeps appearing next to your rivals and never next to you, that site is your first outreach target.

Listicles. Many cited pages are list articles, such as "best tools for..." or "alternatives to...". Aivting flags them, and with the Listicles only and Gaps only filters together it shows the ones that answers cite alongside competitors but never with you. Each is a page engines already cite when answering your buyers' questions.

Your own site as a source. Compare how often you're named in answers that cite your domain against answers that don't. If the difference is large, treat it as a pattern worth acting on, not proof of cause and effect: the citation may not be why you were named.

None of this guarantees a mention. It gives you a short list of specific pages to work on, which beats guessing.

When you bring this to a client, keep the report small. Show the mention count with its n, the competitors beside it, the two or three gaps you found, and what you plan to do about them. Say plainly that answers vary between runs and that a change over a few days may be noise. A client who hears that up front trusts the next report more, not less. Aivting can export a report with your agency's or your client's logo, so the same numbers go from the dashboard to the client without retyping.

FAQ about ChatGPT brand monitoring

Can you monitor real-time brand mentions in ChatGPT?

No. Aivting collects answers in runs, not in real time [CADENCE]. For brand monitoring that's usually enough, because answers vary between runs and trends over days and weeks tell you more than any one moment.

Does ChatGPT give every user the same answer?

No. Answers can differ between runs, even for the same prompt. That's why one check tells you little, and why Aivting shows a count of answers (n) next to every figure.

Can I monitor competitors' mentions in ChatGPT too?

Yes. Aivting tracks up to 10 competitors per project and suggests some when you set up. Their mentions are counted from the same answers as yours, so you can compare share of voice directly.

How many prompts should I monitor?

Start with a mix of branded and non-branded prompts that mirror what your buyers ask. Aivting suggests [PROMPT LIMIT] prompts at a time.

Does monitoring ChatGPT also cover Google AI Overviews?

They are separate engines, so each is measured on its own. [ENGINE AVAILABILITY: e.g. "Aivting collects both ChatGPT and Google AI Overviews,"] and you can filter the dashboard by engine.