How to Check If ChatGPT Recommends Your Brand (Free 30-Minute Test)

Here's a five-second experiment that ruins most marketing teams' afternoon: open ChatGPT and ask it to recommend the best products in your category. Read the answer. Now the real question — was your brand in it?

If it wasn't, you're not alone, and you're not doomed. But you are making decisions blind. This guide gives you a free, rigorous 30-minute protocol to measure exactly where your brand stands across the five AI platforms your buyers use — the same protocol our full audit system uses, distilled to its core.

What you're measuring (and why one test isn't enough)

AI answers are probabilistic and personalized. Ask once and you've sampled one answer from a distribution. That's why the protocol below uses 10 prompts across 5 platforms — 50 data points, enough to see your real position without turning this into a research project.

You'll track four things for every answer:

• Mentioned — is your brand named at all?
• Cited — is your domain linked as a source?
• Position — first brand named, or fifth?
• Competitor — who got named when you didn't?

Step 1 (5 min): Write your 10 test prompts

Write prompts the way your buyers actually ask. Cover these five intents, two prompts each:

1. Best-of: "What are the best [your category] tools for [your audience]?"
2. Comparison: "[Your brand] vs [main competitor] — which is better?"
3. Problem: "How do I solve [the problem you solve] without hiring?"
4. Pricing: "How much does [your category] software cost?"
5. Alternatives: "Best [competitor] alternatives in 2026"

Step 2 (15 min): Run all 5 platforms

For each platform, open a fresh, logged-out session (or a new account with no history), and ask your 10 prompts one at a time in separate conversations:

• ChatGPT — chat.openai.com (with search enabled)
• Perplexity — perplexity.ai
• Gemini — gemini.google.com
• Microsoft Copilot — copilot.microsoft.com
• Google AI Overviews — google.com (search the prompt; note the AI answer on top)

Log each result as a row: platform, prompt, mentioned (Y/N), cited with link (Y/N), position, top competitor named, and a screenshot. Verbatim logging matters — you'll want the actual wording later when you fix sentiment problems.

Step 3 (5 min): Score yourself

Compute two rates per platform: Mention Rate (named ÷ 10) and Citation Rate (linked ÷ 10). Then your totals across all 50 tests. Interpretation:

• 0–20% mentioned: you're invisible — and you likely have an access problem (see our robots.txt guide)
• 20–50%: you're in the conversation but losing shortlists — consensus problem (reviews, listicles, mentions)
• 50%+: competitive — now it's about position and sentiment
• Mentioned ≫ cited: engines know you but don't trust your pages as sources — content structure problem

Step 4 (5 min): Find your #1 gap

Look at the rows where competitors appear and you don't. Patterns tell you the fix: absent everywhere → crawler access; absent in commercial prompts but present in informational ones → reviews/listicles; present but described wrongly → entity/consensus; cited but ranked below a rival → authority.

What most brands discover

Running this test with real companies, three findings repeat: (1) at least one platform can't reach their site at all — a firewall or robots.txt rule is silently blocking AI crawlers; (2) competitors dominate 'best X' prompts purely through listicles and review platforms, not better products; (3) their own descriptions in AI answers are outdated, because the web's consensus about them is stale. All three are fixable — but only once you can see them.

From 30 minutes to a full system

This quick test is a flashlight. The full CiteSight audit is the floodlight: 30 prompts instead of 10, a 120-point weighted checklist that diagnoses why you're losing, a competitor Share-of-Voice benchmark, and a 90-day prioritized fix plan. Same protocol, industrial strength.