To check whether ChatGPT knows your brand, open a temporary chat and ask it what it knows about your company, then ask the questions your buyers ask about your category without naming you. The first test shows whether ChatGPT knows you; the second shows whether it recommends you. Most brands pass the first and fail the second. Here's a 15-minute method, a scoring sheet, and what to fix for each result.
Before you start: set up a fair test
ChatGPT's answers depend on memory, past chats and whether it searches the web. To see what a new buyer sees:
- Use a temporary chat (or a logged-out browser) so memory and past conversations don't colour the answer. See ChatGPT memory and personalization.
- Ask each question at least three times. Answers vary between runs; one screenshot proves little. See why AI answers vary.
- Note whether ChatGPT searched. If it shows sources, the answer came from live web results; if not, from its trained knowledge. The two can differ.
Test 1: Does ChatGPT know your brand? (5 minutes)
Ask these by name:
- "What do you know about [Brand]?"
- "What does [Brand] do, and who is it for?"
- "How much does [Brand] cost?"
- "What are the pros and cons of [Brand]?"
- "Who are [Brand]'s main competitors?"
Score each answer:
| Check | What to look for |
|---|---|
| Recognised | Does it know you exist, or confuse you with a similar name? |
| Category | Does it put you in the right category and market? |
| Facts | Are products, pricing, locations and founders correct and current? |
| Sentiment | Is the description positive, neutral or negative? |
| Competitors | Are the competitors it lists the ones you actually compete with? |
Brand-name confusion is common. If another company or product shares or resembles your name, ChatGPT may merge the two. Clear Organization schema, a consistent name everywhere and an explicit "not affiliated with" line help. See brand consistency and entity signals.
Test 2: Does ChatGPT recommend your brand? (10 minutes)
This is the test that matters for revenue. Ask the questions buyers ask without your name:
- "What is the best [category] for [audience]?"
- "Which [category] should a [company type] in [city/country] use?"
- "[Competitor] alternatives"
- "[Your brand] vs [competitor]" (this one includes your name, but tests positioning)
- A problem-led question: "How do I [job your product does]?"
For each, record:
| Prompt | Run 1 | Run 2 | Run 3 | Your position | Competitors named | Sources cited |
|---|---|---|---|---|---|---|
| best [category] for [audience] | ✅ | ❌ | ✅ | #3 | A, B, C | g2.com, reddit.com |
Your mention rate is the share of runs that name you. Three runs on five prompts gives 15 data points: enough for a first read, not enough for a trend. See AI visibility metrics and building a prompt tracking set.
Reading your results
| Result | What it means | What to fix first |
|---|---|---|
| Not recognised | ChatGPT has little or no information about you | Crawlable site, Organization schema, consistent profiles, coverage on independent sites |
| Recognised but wrong facts | Old or conflicting sources | Update your own pages and third-party profiles; publish facts plainly |
| Recognised but not recommended | You lack third-party consensus for the category | Reviews, best-of lists, Reddit and Quora, digital PR |
| Recommended, but low and inconsistent | You're in the consideration set but not the default | Comparison pages, specific proof, more cited sources |
| Recommended near the top | Defend and expand | Track monthly, extend to more prompts and engines |
Why ChatGPT gets brands wrong
Wrong answers usually trace back to a source, not to the model inventing things at random:
- Your own outdated pages: old pricing pages, retired product pages, stale "about" copy.
- Third-party profiles: directories, review sites, LinkedIn and old press coverage.
- Name collisions: similarly named companies, apps or open-source projects.
- Missing information: if your pricing or locations aren't public, ChatGPT guesses or relies on others.
Find the cited sources in the answer, fix the ones you control and request updates to the rest. See fixing AI hallucinations about your brand and fixing negative or outdated AI answers.
How to get ChatGPT to recommend you
- Let OpenAI's crawlers in. Allow OAI-SearchBot and ChatGPT-User in robots.txt and your CDN, and serve content in HTML. See AI crawlers and robots.txt.
- Say clearly what you do and for whom. "Payroll software for Indian startups" matches prompts; vague slogans don't. See answer-first writing.
- Publish comparison and alternatives pages that are fair and specific. See comparison pages.
- Earn independent mentions on the review sites, lists, communities and publications ChatGPT cites for your category. See reviews and AI recommendations, getting included in best-of lists and digital PR for AEO.
- Re-test monthly with the same prompts and compare. See ChatGPT ranking factors.
What doesn't work
- Telling ChatGPT about your brand in a chat. It doesn't change answers for anyone else.
- Hidden text or instructions aimed at AI. Unreliable and manipulative.
- Fake reviews or sockpuppet Reddit accounts. Against platform rules, and they create the negative threads you want to avoid.
Run the same test in other AI engines
Gemini, Google AI Overviews, Perplexity and Claude use different sources, so results often differ. A brand can be recommended by Perplexity, which leans on live web results, and missing from ChatGPT's trained knowledge. Run the same five category prompts in each and compare. See how each AI engine finds sources.
Want it done for you?
The 15-minute check gives you a first read. For a fuller picture across every major engine, more prompts and more runs, with the sources and fixes laid out, request a free AI visibility audit from our team, or follow the manual AI visibility audit method yourself.
