Searching your own brand name in ChatGPT is the common first check. It tells you whether ChatGPT can retrieve your brand when you ask for it, not whether a buyer will discover it. Ask the question a buyer would ask instead. Then record which products the answer names, in what order, and why.
You can make a useful first pass in 30 minutes. Treat it as a snapshot of a small set of answers, not a verdict on your whole market.
Step 1: Write 10 questions buyers actually ask
Start with real customer language. Ask recent customers what they searched before they found you, what they compared you with, and what almost stopped them from buying. Use the questions in sales call notes, support conversations and onboarding forms. Keep a few close to the category, such as “best project tool for a small product team,” but avoid ten versions of the same query.
Group the list by job, comparison and constraint. A useful set gives ChatGPT more than one chance to describe where your product fits.
- Best project management tool for a small startup
- Best kanban board for a remote product team
- How to plan a product launch with a small team
- Affordable task app for a founder managing one product
- Project tracker that connects timelines and tasks
- How to replace a spreadsheet for operations planning
- Best project tool for getting client approvals
- How to track sprint goals across design and engineering
- Simple project dashboard for a solo founder
- Best offline task manager for a distributed team
Replace these with the wording your buyers use. If you sell to a narrow market, include its terms and the alternatives your customers actually consider.
Step 2: Run each question in a new ChatGPT chat
Paste one question into a new chat. Do not put your company name into the prompt unless the question naturally includes it. Record the full names in the answer, whether your product appears, and its position in the list. Save a copy of the answer and the date so you can compare it later.
The table below is a labelled example, not output from a live ChatGPT test. Taskroom and the other product names are fictional. Your answers will be different.
| Question | Named brands | Taskroom named? | Position |
|---|---|---|---|
| Best project management tool for a small startup | Taskroom, Boardly, Gridwork | Yes | 1 |
| Best kanban board for a remote product team | Boardly, Taskroom, Tasklane | Yes | 2 |
| How to plan a product launch with a small team | Tasklane, Workboard | No | Not named |
| Affordable task app for a founder managing one product | Sprintbox, Boardly, Planit | No | Not named |
| Project tracker that connects timelines and tasks | Taskroom, Gridwork, Boardly | Yes | 1 |
| How to replace a spreadsheet for operations planning | Rowbase, Taskroom | Yes | 2 |
| Best project tool for getting client approvals | Cliently, Boardly | No | Not named |
| How to track sprint goals across design and engineering | Sprintbox, Taskroom, Cyclewise | Yes | 2 |
| Simple project dashboard for a solo founder | Taskroom, Boardly | Yes | 1 |
| Best offline task manager for a distributed team | Sprintbox, Workboard, Tasklane | No | Not named |
Keep the question exactly as written. If you change it halfway through, mark it as a different prompt rather than treating the answers as directly comparable.
Step 3: Count the result without overreading it
In this example, Taskroom appears in 6 of 10 answers. It appears first in 3 answers. Those counts describe this example prompt set and these ten responses only. They do not mean 60% of buyers see Taskroom, or that it will be named in the next conversation.
Ten questions are a starting sample. One answer can change, and a small sample can swing sharply when one response changes. Record the numerator and denominator, keep the prompts stable, and avoid calling a small change a market trend.
Step 4: Read the answer for why a brand was named
Do not stop at the list. Read the sentence around each recommendation. Does it say the product is easy to set up, made for a certain team, or strong at a specific job? Then compare that wording with your homepage and product pages. What does the competitor explain clearly that your site leaves implicit?
For example, if an answer describes Boardly as a kanban tool for remote teams but calls Taskroom only “a productivity app,” look at the language available to ChatGPT. That is a clue to investigate, not proof that one sentence caused the answer. Check whether your own pages name the audience, the task and the product capability in plain language.
Step 5: Pick one change
The first useful change is often one clear sentence on your homepage that says what you do and who it is for. Do not rewrite the whole site based on one answer. Make the positioning more specific while keeping it accurate.
Taskroom helps teams get more done.
Taskroom gives startup product teams boards, timelines and task tracking in one workspace, so a small team can plan a launch without scattering work across tools.
This is an example, not recommended copy for every product. Use words your customers recognize and claims your product can support.
Step 6: Run the same check next week
Ask the same ten questions in new chats and record the date. Compare which brands appeared, their positions and the reasons in the answer. Note any site changes or product changes alongside the result.
AI answers change. If the result is different, you will have a record to inspect. You still will not know from one week alone whether a page edit caused the difference.
The limits of a manual check
ChatGPT, Claude and Perplexity can answer the same question differently. ChatGPT can also vary its answer between chats, and its responses may reflect your conversation or available search context. One person's manual session is not the same as a repeated, tracked sample.
A spreadsheet is enough to learn the basics. It is harder to keep prompts, dates, engines and answers consistent as the check grows. Do not treat a handful of manual answers as a stable visibility score.
If you want this every week
You can automate the repeated questions, keep their answer history and review named competitors in one place. Rankdawn checks ChatGPT, Claude and Perplexity on a weekly schedule and shows the answers and suggested fixes. That saves the repeated recording work, but it does not make the answers predictable or guarantee a mention.
Start with your first Rankdawn scan without a card. After that, Pro has a 7-day trial that requires a card. Review the Pro plan for the current limits and billing terms.