How many prompts do I need to track?

Twenty is enough to see a pattern; fifty gives you stable percentages. Beyond a hundred the extra prompts rarely change the picture, and the checking cost grows faster than the insight.

Why do I get a different answer when I ask the same question twice?

ChatGPT generates each answer fresh, and small changes in wording, history or the model's sampling change the result. That is why you track rates over repeated runs rather than trusting any single answer.

Should I use a logged-in account or a clean session?

A clean session with memory and custom instructions off. A logged-in account with history reflects your own past conversations, which your buyers do not share.

Does ChatGPT cite sources?

When it searches the web for an answer it shows the pages it used. Whether it searches depends on the question; for well-known topics it may answer from training data with no sources at all. Track citations as a separate outcome from mentions for exactly this reason.

Can I track Gemini and Perplexity the same way?

Yes, with the same prompt list and the same three outcomes. Expect the results to differ by engine; that difference is part of what you are measuring.

How to Track Whether Your Brand Appears in ChatGPT

A practical walkthrough for monitoring your brand's presence in ChatGPT answers: what to measure, how often to check, and what to do with the data.

If buyers are asking ChatGPT about your category, you need to know whether your brand appears in the answer. Guessing does not work, and neither does asking once and screenshotting the result. Here is how to track your brand's presence in ChatGPT systematically: what to record, how often to check, and how to turn the data into work.

Step 1: Build your prompt list

Everything downstream depends on this list, so it is worth an hour.

Start from buyer questions, not keywords

A rank tracker takes keywords. ChatGPT takes questions, and the questions your buyers ask are longer, more specific and more conversational than anything in your keyword research. Good sources:

Sales calls and demo requests: the exact phrasing people use to describe what they need.
Support tickets and onboarding questions from your first weeks with a customer.
Your own "alternatives to" and "vs" pages, if you have them.
The "People also ask" boxes under your main Google keywords, rewritten as full sentences.

Cover the funnel

Aim for a mix across three stages:

Stage	Example prompt	What a mention means
Category	"best invoicing tools for freelancers"	You are on the shortlist
Comparison	"FreshBooks vs Zoho Invoice for a solo designer"	You are a known option
Brand	"is [your brand] good for small agencies"	The engine has an opinion about you

Twenty to fifty prompts is plenty. Write them once, save them, and resist the urge to tweak the wording later; a changed prompt is a new prompt, and it breaks your trend line.

Step 2: Check for three outcomes

"Did we show up?" is one column. You need three.

Mention

Is your brand named at all, anywhere in the answer? Record yes or no, and if yes, the position: first, second, third, or later. Being named first in a shortlist is a different outcome from being the fifth name in a "you could also consider" sentence.

Citation

Is your website linked as a source? ChatGPT shows sources when it searches the web for the answer (OpenAI describes the behaviour in its ChatGPT search announcement), and whether it searches depends on the question. Citation is the outcome your own content can most directly influence, which makes it worth tracking separately from mention. A brand can be mentioned from training data without a single link to its site, and a brand can be cited for a definition without being recommended.

Competitors named

Who appears instead of you, or alongside you? This column becomes your competitor benchmark in Step 4, and it is also the fastest way to see what the engine values. If the same three rivals show up for every category prompt, look at what those three have in common: review coverage, directory listings, a clear one-line description, comparison articles about them.

Worth knowing: when an answer cites sources, write the source domains down too. Those are the sites the engine trusts for your category, and each one is a place where a listing or a review changes your visibility.

Step 3: Run checks on a schedule

AI answers change as models update, as the web changes, and simply because each answer is generated fresh. A one-time check is a snapshot. A weekly check is a trend, and the trend is the only thing that tells you whether your AEO work is doing anything.

Keep the conditions the same

Use a clean session: memory off, no custom instructions, no chat history. Your buyers do not share your history.
Ask each prompt in a fresh conversation, not as a follow-up.
Note the model and whether web search was on. Both change the answer.

Pick a cadence and stick to it

Weekly is the right default. Daily produces noise you will be tempted to react to; monthly hides changes until they are old news. Same day, same time, same prompts. If you run the checks by hand, batch them into one sitting so the conditions stay comparable.

Record it somewhere that keeps history

A spreadsheet with one row per prompt per week works for a baseline. The moment you add a second engine or a second brand, the sheet turns into a chore; that is the point where a tool earns its keep. Presence Scout runs the same prompts on ChatGPT, Gemini and Perplexity on a schedule and keeps every answer, so the history is there when you need it. The ChatGPT visibility tracker page shows what that looks like for ChatGPT specifically.

Step 4: Benchmark competitors

The most actionable metric in the whole exercise is share of voice: how often you are named compared with the competitors named for the same prompts.

Suppose you track thirty category prompts. A competitor appears in twenty-four of the answers and you appear in six. That is not a vague sense of being behind; it is a gap of eighteen prompts, each of which you can open, read, and learn from. Which sources did the engine cite when it named them? Do those sources mention you at all? Is your description on those sources the same as your description on your own site?

Share of voice also protects you from a common mistake: celebrating an improvement in your own mention rate while a rival's grows faster. Both numbers move; the ratio is what matters.

Step 5: Turn the data into work

Tracking without a to-do list is just anxiety with a spreadsheet. Three patterns cover most of what you will find:

Named by competitors' sources, not by you. You are mentioned but not cited. Your own pages are not the ones the engine reads. Fix: write the page that answers the prompt directly, in the first sixty words, and keep it updated.
Absent from category prompts, present on brand prompts. The engine knows you but does not shortlist you. Fix: corroboration. Get listed and reviewed on the sources it cites for the category.
Described wrongly. Named, but with the wrong category or an outdated feature set. Fix: entity clarity. Make your About page, listings and structured data agree, then wait for the refresh.

Our Action Plan turns these patterns into tasks automatically, but the logic works just as well on paper.

Doing it across engines

Everything above applies to Gemini, Perplexity and Google's AI Overviews with the same prompt list and the same three outcomes. Expect the results to differ. An engine that names you for every prompt while another ignores you is not a contradiction; it is the measurement. Track them side by side, and read the Gemini tracker and Perplexity tracker pages for the habits each engine has.

If you want the wider picture first, What is Answer Engine Optimization explains what these numbers are for. If you would rather see your own baseline today, add your brand free and run your first check.