The self-fulfilling prophecy problem
Imagine asking ChatGPT "What are the features of [YourBrand]?" Of course ChatGPT is going to talk about your brand. The prompt told it to. A green "Cited" badge on that prompt tells you nothing about whether real buyers would ever discover you.
Compare that to "What is the best way to [problem your brand solves]?" Now the prompt is agnostic. If ChatGPT cites you, it chose to — because it judged you as a relevant source for that buyer question, on its own. That result is signal.
What an unbranded query looks like
| ✗ Branded (we reject) | ✓ Unbranded (we use) |
|---|---|
| What does Findabl cost? | How much does AI citation monitoring cost? |
| Is HubSpot better than Salesforce? | What is the best CRM for mid-market B2B? |
| Who makes the best Notion alternative? | What are the top collaborative document tools? |
| How do I set up Zapier for Slack? | How do I automate Slack notifications? |
The unbranded version simulates a real buying moment. The branded version simulates asking a search engine for a spec sheet.
What this means for you
When Findabl says "ChatGPT cited you on 3 of your 5 tracked prompts," it means ChatGPT picked you without prompting. That is the metric worth optimizing for. Nothing else is.
It also means an AI engine mentioning you by name in response to a branded query (e.g. a user typing "tell me about [your brand]") is not a Findabl signal. That is just the engine repeating a name it was given. Useful for reputation reasons; not a citation win.
Why this is enforced in code
Unbranded queries are an architectural rule (ADR-005). The generateCitationQueries function actively filters brand names out of generated prompts, and the citation runner rejects user-submitted prompts that contain the tracked brand. There is no toggle to turn this off.
We encode this in code (not just policy) because it protects you from yourself. If we let a user add "Best CRM like [their brand]" as a tracked prompt, their citation rate would look great and mean nothing. The code refuses, so the numbers stay honest.
But what about branded search?
Branded AI search (users asking directly about you) is a real use case — it is just a different product question. Reputation management, correcting misinformation, verifying that an AI engine knows you exist — those all matter, and most teams handle them with a separate monitoring workflow.
Findabl focuses on the unbranded competitive game because that is where most of the acquisition opportunity lives and where most brands have the biggest visibility gap.