Track What ChatGPT Says About Your Brand

You track brand mentions in AI by regularly asking the same set of buyer-intent prompts across ChatGPT, Claude and Perplexity, and recording whether your brand is mentioned, where in the answer, and how it's described. Do it weekly with a spreadsheet and ten prompts, or automate it with a tracker that samples those prompts on a schedule. Either way, the unit of measurement is simple: when a buyer asks, are you in the answer?

Why AI mentions matter more than social mentions now

AI mentions matter more than social mentions because they happen at decision time. A social mention is ambient noise — someone scrolling past your name. An AI mention is different: the person asking "what's the best tool for X" or "is Y worth it" is usually about to spend money, and the assistant's answer is the last thing they read before they act.

There's also no page 2. Google gives you ten blue links and you can claw your way from position 8 to position 3. An AI assistant gives one synthesized answer. You're either named in it or you don't exist for that buyer. That makes the difference between GEO vs SEO more than academic — it's the difference between being shortlisted and being invisible.

The uncomfortable part: this is already happening whether you track it or not. Buyers are asking AI about your category, your competitors, and you. The only question is whether you know what the answer says.

The manual method: 10 prompts, one spreadsheet, once a week

The manual method is to write down ten prompts a real buyer would type, run them across the major assistants once a week, and log the results in a spreadsheet. It takes about 30 minutes per week and costs nothing.

Here's how to build your prompt spread:

1. Four category queries. "Best [your category] tools", "top [category] for [your audience]", "what should I use for [the problem you solve]". These reveal whether you're in the consideration set at all.
2. Three comparison queries. "[You] vs [main competitor]", "[Competitor A] vs [Competitor B]", "is [you] better than [competitor]". These show how you're positioned relative to alternatives.
3. Two alternatives queries. "Alternatives to [competitor]", "cheaper/simpler alternatives to [incumbent]". Buyers switching tools ask these, and you want to be the answer.
4. One direct query. "Is [your brand] any good?" or "What is [your brand]?" This is what the assistant says when it already knows your name — check that it's accurate.

Then run each prompt in ChatGPT, Claude and Perplexity — Perplexity citations are especially useful because you can see which sources the answer leans on. Log every run in a spreadsheet. A simple layout works:

| Prompt | Model | Mentioned? | Position | Sentiment | Competitors named | Sources cited | | --- | --- | --- | --- | --- | --- | --- | | best SEO tools for solo founders | ChatGPT | No | — | — | two incumbents | — | | best SEO tools for solo founders | Perplexity | Yes | 2nd | Positive | an incumbent | indiehackers.com thread | | is SuperSEO.sh any good | Claude | Yes | 1st | Neutral | — | superseo.sh |

Keep the wording of each prompt identical week to week. The value isn't any single answer — assistants are non-deterministic and one run proves nothing. The value is the trend: did you appear in 3 of 10 prompts last month and 6 of 10 this month?

What to record

Record five things for every prompt run: whether you were mentioned, your position in the answer, the sentiment, which competitors appeared, and which sources were cited.

• Mentioned (yes/no). The core metric. Track your mention rate as a percentage of your ten prompts.
• Position. First name in the list or buried fifth. Being mentioned fourth in a list of five is only slightly better than not being mentioned.
• Sentiment. Positive, neutral, or negative — and why. "Affordable but limited" is a different problem than "no one has heard of it."
• Competitors named. If the same two competitors appear in every answer and you never do, that's your real target list — not whoever you assume your competitors are.
• Cited sources. When the assistant links or names a source (a review site, a Reddit thread, a comparison article), write it down. These tell you exactly which pages shape the answer.

What moves the needle

What moves the needle is web consensus: AI assistants repeat what the open web collectively says about you, so the way to change your mentions is to change what the web says. Models don't invent opinions about brands — they compress reviews, forum threads, comparison articles, and your own pages into an answer. Exactly how each model weights its sources isn't public and varies by model, but the pattern practitioners observe is consistent: mentions follow citations.

That gives you four concrete levers:

1. Reviews. Presence on the review sites and marketplaces your buyers trust. An assistant asked "is X any good" leans heavily on these.
2. Forum threads. Reddit, Hacker News, Indie Hackers, niche communities. A single detailed thread where real users compare you to a competitor can show up as a cited source within weeks.
3. Comparison articles. "Best X tools" listicles and "X vs Y" posts — including ones you write yourself, as long as they're honest and specific.
4. Your own citable pages. A clear page that says who you are, what you do, and who you're for, in plain extractable text. If the assistant can't find a clean definition of your product, it fills the gap with someone else's description.

One warning: don't try to spam your way in. Mass-produced "best tools" articles with your name stuffed in are exactly the kind of content models and their search layers are getting better at discounting. Consensus means other people saying it too.

Automating it

Automating AI mention tracking means a tool runs your prompt set on a schedule across multiple models, records the answers, and charts your mention rate and position over time — so you see trends and get alerted to changes instead of spending 30 minutes a week copy-pasting into a spreadsheet. Automation also fixes the manual method's biggest flaw: it samples consistently, so a bad week doesn't get skipped because you were busy.

This is what SuperSEO.sh does: it samples real buyer-intent prompts against live AI models as part of every scan and shows you whether and how you're being mentioned, alongside the technical GEO audit. If you'd rather skip the spreadsheet, scan your site free and see your first baseline in a couple of minutes.

Manual or automated doesn't matter much at the start. What matters is that you have a baseline at all — most founders have never once asked an AI assistant what it says about them.

FAQ

How often should I check my AI mentions?

Weekly is the right cadence for most solo founders. AI answers don't change overnight, but they do drift as models update and new web content gets indexed — weekly sampling catches a shift within days without turning tracking into a second job.

Can I change what AI says about my brand?

Yes, but indirectly. You can't edit a model's answer, but you can change the web consensus it compresses: publish clear citable pages, earn reviews, and participate in the forum threads and comparison articles the models cite. Expect weeks to months, not days.

What if the AI says something negative about me?

First verify whether it's true, because the assistant is usually summarizing real criticism from reviews or forums. Fix the underlying issue, respond on the source platform where appropriate, and publish your own clear account — the negative mention fades as the consensus shifts.

Does this work for local businesses?

Yes, and the stakes are often higher because buyers ask things like "best plumber in Austin" with immediate intent. Use the same method with location-specific prompts, and pay extra attention to Google reviews and local directories, which local answers lean on heavily.

How many prompts is enough?

Ten well-chosen prompts is enough for a solo founder. Cover your category, your top two competitors, and one direct brand query — adding more prompts adds time faster than it adds insight. Consistency of the same prompts over time beats breadth.

If you've never asked ChatGPT what it thinks of your product, do it today — then scan your site free to get a tracked baseline and find out whether buyers hear your name when it counts.