Guide
How to check what ChatGPT says about your brand.
Last updated August 5, 2026
The short answer
The quick manual check: open ChatGPT (ideally with web search enabled), ask the questions your buyers actually ask — "best [your category] tool", "[competitor] alternatives", "is [your brand] any good" — and see whether you're named. Repeat each question a few times in fresh chats, because answers genuinely vary run to run. Then repeat the same questions in Perplexity and Gemini, which use different sources and frequently name different brands.
The manual check is worth doing today, but it has three blind spots: one run per question is statistically meaningless, your own chat history biases what you see, and you can't watch answers change over time. A proper measurement asks each question many times per engine, in clean sessions, across all five major engines, and reports the mention rate with a confidence interval — which is exactly what a Visitd teardown does for free in about sixty seconds.
Step 1: ask like a buyer, not like a founder
Don't ask "what do you know about Acme?" — ChatGPT will politely summarize your website back at you and you'll learn nothing. Buyers ask category questions: "best CRM for a 10-person startup", "HubSpot alternatives", "most affordable helpdesk software". Ask those. The uncomfortable version is the comparison question your sales team hears on every call — that's the answer that's costing you deals.
Cover the three question types that decide purchases: discovery ("best X for Y"), comparison ("[rival] alternatives", "A vs B"), and validation ("is [brand] legit", "[brand] reviews").
Step 2: repeat everything, in fresh sessions
AI engines sample their output — the same question can produce different brand lists an hour apart. One run tells you almost nothing. Ask each question at least three to five times in new chats (memory off, or a private window), and count how often you're named rather than whether you're named. If you appear in one answer out of five, you don't 'show up in ChatGPT' — you show up 20% of the time, and a competitor owns the rest.
Step 3: check the other engines — they disagree
ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews are built on different indexes and cite different sources: Perplexity leans heavily on Reddit and YouTube, Gemini inherits Google's index, ChatGPT Search runs on Bing. It's entirely normal to be visible in one engine and absent from another — which also means each engine needs its own fix strategy.
Step 4: check whether AI can even read your site
A surprising share of sites accidentally block AI crawlers in robots.txt — GPTBot, ClaudeBot, PerplexityBot — often from a years-old "block all bots" rule. If an engine's crawler can't read your site, no amount of content work will get you cited. Check your robots.txt, or run a scan that includes the crawler check.
The measured version of all of the above
A Visitd teardown automates this: it generates your buyers' real questions, asks them across five engines multiple times in clean sessions, and returns your visibility score with a 90% confidence band, the competitor getting named instead of you, one verbatim proof answer, and the crawler check. It's free with an account and takes about a minute — and monitoring re-runs it weekly so you see when answers flip.
Questions, answered
Why does ChatGPT give different answers each time?
Because generation is sampled, not looked up — the model draws from a distribution, so brand lists shift run to run. That's why mention rate across many runs (with a confidence interval) is the honest measurement, and any single run — or any tool selling a single clean number — is a guess.
Does asking ChatGPT about my own brand bias the answer?
Your account's chat history and memory can color responses. Use fresh chats with memory off, a private window, or ideally an API-based measurement in clean sessions — which is how Visitd runs its scans.
Can I just ask ChatGPT why it doesn't recommend me?
You can, but treat the answer as a plausible story, not ground truth — the model doesn't have reliable introspection into its own retrieval. The measurable causes are usually crawler access, weak third-party citations, and missing comparison content.
How often should I re-check?
AI answers shift constantly — engines update models, indexes refresh, and competitors ship content. Monthly manual checks are the floor; weekly automated monitoring is what catches an answer flipping before the quarter's numbers do.