How to check if your brand shows up in ChatGPT™

To check whether your brand shows up in ChatGPT™, ask it the questions your buyers actually ask (best-of lists, alternatives, comparison questions) with search enabled and in a fresh session, then record four things per answer: whether your brand is mentioned, where it appears in any list, how it is framed, and which sources are cited. Repeat each prompt several times and across other assistants before drawing conclusions, because AI answers vary from run to run and provider to provider. The six-step audit below takes an afternoon and tells you exactly where you stand.

Step 1: Build a prompt list from real buyer questions

Most teams start by typing their brand name into ChatGPT™ and reading the answer. That test tells you very little: buyers who already know your name were never the visibility problem. The prompts worth auditing are the ones where a buyer describes a need and the assistant decides which brands deserve a mention.

Build a list of 20 to 40 prompts across four categories:

• Best-of prompts: "What are the best [category] tools for [use case]?" These are the AI equivalent of a high-intent search results page, and they produce ranked lists you can score.
• Alternatives prompts: "What are good alternatives to [competitor]?" You want to appear in competitors' consideration sets, and you want to know who appears in yours.
• How-to-choose prompts: "How should I choose a [category] vendor? What should I look for?" Assistants frequently name example brands while explaining the criteria.
• Direct brand prompts: "What does [your brand] do? Is [your brand] good? [Your brand] vs [competitor]?" These measure how deeply and accurately the model knows you, not whether it discovers you.

Pull phrasing from real sources: sales call transcripts, support tickets, community threads, People Also Ask boxes, and your paid search query reports. Write prompts the way a buyer types, not the way a marketer writes.

Step 2: Ask with search enabled, in a clean session

Two settings decide whether your results mean anything.

First, turn search on. ChatGPT™ without browsing answers from training data that may be months or years old; ChatGPT™ with search retrieves live pages and cites them. Those are effectively two different products, and the search-enabled one is where the competitive fight over citations happens.

Second, remove contamination. ChatGPT™'s memory and custom instructions personalize answers, and a session where you have already discussed your brand will treat it more favorably than a stranger's session would. Use a temporary chat (or a clean account), turn memory off, clear custom instructions, and start fresh for every prompt. Skip this and you are measuring your own chat history, not your market visibility.

Step 3: Record four signals for every answer

A yes/no mention column is not enough. Capture four things per answer:

Signal	What to record	Why it matters
Mention	Was your brand named at all?	The baseline visibility measure
Position	Where it appeared in any list (1st, 3rd, 7th)	Early slots get read; trailing mentions get skimmed
Sentiment and framing	Recommended, listed neutrally, hedged, or criticized	A hedged mention ("some users report issues") can hurt more than absence
Citations	Which domains the answer cites, and whether any are yours	Citations reveal which sources the engine trusts for your category

Log results in a spreadsheet with one row per prompt, per provider, per run, and record your competitors' mentions in the same pass. Half the value of the audit is seeing who wins the prompts you lose.

Step 4: Repeat each prompt before concluding anything

Run every prompt at least three times before you write a single conclusion. AI answers are generated, not retrieved from a fixed database: the same prompt can name your brand in one run and omit it in the next, because of both model sampling and variation in live search results. A single run is one draw from a distribution. What you actually want is a mention rate ("named in 2 of 3 runs"), not a binary. See why AI answers change between runs for why this happens and how much variance to expect.

Step 5: Repeat across providers, because they disagree

ChatGPT™, Claude®, Gemini™, and Perplexity™ run different models, different retrieval systems, and different underlying indexes. It is normal for a brand to lead the list on one provider and be invisible on another, and each provider maps to a real slice of your buyers. Run the same prompt panel across all four and score them separately. Per-provider mention rates are the raw material for AI share of voice: how to measure AI share of voice covers turning them into a competitive metric.

Step 6: Automate once the manual audit proves the gaps

The manual audit is a diagnosis, and it is worth doing by hand once. You will read full answers, notice framing problems, and build intuition that no dashboard gives you. But it does not scale: thirty prompts, four providers, and three runs is 360 responses to collect and score, and trends only mean something if you re-run the panel on a schedule.

This is the job AI visibility tools automate. They run a fixed prompt panel across providers on a schedule, extract mentions, positions, sentiment, and citations from each response, and trend the rates over time. Real options in the category include Gen3 AI Visibility, Profound, Otterly.AI, and the Semrush AI Visibility Toolkit, among others. When evaluating any of them, check three things: which providers they actually query, whether they run each prompt multiple times (single-run scores are noise), and whether they capture citations, since citations connect visibility back to content work you can act on. Gen3 AI Visibility is built around the second check in particular: scans repeat each prompt and report the spread across runs.

Why might ChatGPT™ not see your site at all?

If the audit shows you absent from ChatGPT™ while present on other providers, suspect crawlability before strategy.

OpenAI's crawlers fetch raw HTML and do not execute JavaScript. If your site is a client-rendered single-page app (React, Vue, or similar without server-side rendering), the crawler receives a nearly empty document: no product copy, no pricing, no comparison pages. Google renders JavaScript; ChatGPT™'s crawlers do not, so a site can rank well in Google while remaining effectively invisible to ChatGPT™ search.

The slots you are competing for are also scarce. Across a sample of scans we ran between April and August 2026 (33,000+ responses, web search enabled on every request), ChatGPT™ named the fewest brands per answer of the four major providers: under 2 tracked category brands on average for category prompts, against up to 2.6 for the most generous provider. A short answer means every slot is contested, and a page ChatGPT™ cannot render is competing for slots it cannot even enter.

The fixes are mechanical:

• Check what crawlers actually see. View the raw page source (not the rendered inspector) or fetch key pages with curl. If your core copy is not in the raw HTML, ChatGPT™ is not reading it.
• Server-render or prerender the pages that matter: product pages, comparison pages, pricing, and docs.
• Allow OpenAI's crawlers in robots.txt. OAI-SearchBot powers ChatGPT™ search retrieval; blocking it removes you from consideration entirely.
• Register with Bing Webmaster Tools and submit via IndexNow. ChatGPT™ search draws on Bing's index alongside OpenAI's own crawling, so a page absent from Bing starts from behind.
• Maintain an XML sitemap so new and updated pages get discovered quickly.

Measure it

If you want this audit run for your brand without building the spreadsheet, the free Pulse visibility check on our homepage shows how your brand appears across the major AI assistants. For the deeper mechanics of scoring and trending, explore the rest of the guides at /learn/.