How to check if ChatGPT mentions your brand

Last updated: August 15, 2026

You can check by hand in about ten minutes: open ChatGPT in a logged-out or temporary chat, ask the five questions your customers actually ask, and write down which brands each answer names. The catch is that one reading tells you almost nothing — answers vary between sessions and shift week to week, so the useful signal is the pattern over time, not a single result.

The manual method, step by step

1. Write down the questions a shopper would really ask

Not your brand name. Shoppers who already know you are not the ones you are losing. Write category questions — the ones someone asks before they know who you are:

• “best vitamin C serum for sensitive skin”
• “affordable retinol alternative”
• “what moisturizer should I use for dry skin in winter”
• And one brand-name check: “is [your brand] legit?” — different question, also worth knowing.

2. Ask them in a clean session

Use a temporary or logged-out chat. In a normal account the assistant has memory and prior context, so it may name your brand simply because you have been talking about it — which tells you nothing about what a stranger sees. Turn web search on, since that is what a shopper asking a live shopping question gets.

3. Record three things per answer

• Were you named at all? This is the only binary that matters.
• Where in the list? First recommendation and fourth are very different outcomes.
• Which pages did it cite? Click the source chips. Those pages are what actually decided the answer.

4. Repeat on Gemini

The two disagree more often than you would expect, because they search differently and read different pages. Checking only one gives you half a picture.

Why one check isn’t enough

AI answers are not stable the way a search ranking is. Ask the same question twice and you can get two different brand lists. That makes a single manual check genuinely misleading in both directions — it can show you being named on a lucky run, or missed on an unlucky one.

There is also a lag. Research on AI search consistently points to a multi-week delay between something changing on the web and it showing up in AI answers. So if you make a change today and check tomorrow, you will learn nothing. You need a before-and-after separated by weeks, which means you need the “before” recorded — and that is the part people never have.

What the answer tells you

• You’re named often — good. Now find out which pages are doing it, so you can protect them.
• You’re named sometimes — the most common result, and the most workable. Look at which questions you win and which you lose.
• You’re never named— also the common starting point, and not a verdict on your product. It means the pages AI reads in your category don’t discuss you yet. Those pages are the citation gaps, and they are a finite, addressable list.

When it’s worth tracking properly

Doing this by hand across two assistants, ten questions and enough weeks to see a trend is a few hours a month of fairly dull work, and the record only exists if you keep it. Automating it is worth it once you actually intend to change something and want to know whether it worked.

That is what SeenScan does: the same questions, both assistants, every day, with the cited pages recorded so you get the gap list rather than just a number. Understanding how AI picks brands in the first place is the part that makes the numbers actionable.