Typing a few questions into ChatGPT and noting whether your brand appears feels like tracking, but it's closer to reading tea leaves. Answers change between runs, depend on location and differ between engines. Here's how to set up tracking that holds up.
Why manual checks mislead
- Answers vary. Ask the same question twice and the list of brands can change.
- Apps personalize. Consumer apps can use your history, memory and location, so what you see isn't what a prospect sees.
- Engines disagree. Perplexity, Gemini and ChatGPT retrieve different sources and often recommend different brands.
1. Choose the right prompts
Track questions buyers ask before they know your name, not searches for your brand. Cover the stages of a purchase:
- Discovery: “best project management tool for remote teams”
- Comparison: “Asana vs Monday for a 10-person agency”
- Alternatives: “cheaper alternatives to Salesforce”
- Use case: “CRM that integrates with Gmail and invoicing”
Twenty to fifty prompts is a good start. Group them by topic or intent so you can see where you're strong and where you're missing.
2. Pick engines and how they're collected
Start with the assistants your buyers use. Then decide what you're measuring. Answers from a vendor's model API with web search are consistent and repeatable. Answers collected from a live consumer experience, like Google AI Overviews, show what people see but vary more. Both are useful; they shouldn't be averaged together, because they answer different questions.
3. Collect enough samples
Visibility is a rate: the share of answers that mention you. A rate from ten answers is very uncertain. At 6 mentions out of 10 answers, the 95% confidence interval runs from about 31% to 83%. With 60 out of 100, it narrows to about 50% to 69%. Increase the number of answers by tracking more prompts, more engines or more samples per prompt. The confidence interval guide shows the maths.
4. Set a schedule and locations
Weekly tracking suits most brands; daily makes sense during launches or campaigns. Run prompts from the countries you sell in, since answers can differ by location.
5. Read the results
- Compare engines side by side. Strong in one and absent in another usually points to a source gap.
- Find competitor gaps. Prompts where rivals appear and you don't are your clearest opportunities.
- Study citations. The domains cited for those prompts show where to focus content and PR.
- Read the answers. How you're described matters as much as whether you're mentioned.
6. Get alerted when it matters
Rather than checking dashboards daily, set alerts for statistically significant changes in visibility, so ordinary run-to-run noise doesn't trigger false alarms.
Checklist
- 20–50 unbranded buyer questions, grouped by intent
- The engines your audience uses, with API and consumer results kept separate
- Enough answers per period for useful confidence intervals
- Weekly schedule from your key countries
- Competitor tracking for share of voice and gaps
- Alerts on significant changes only