Peekaboo
What Peekaboo does not do

What Peekaboo does not do

Personalization and history limits

Why a tracked AI answer is one observation, not a personalized answer, and why there is no history before you started tracking.

3 minute read · Last reviewed: August 14, 2026

Two limits are worth understanding before you lean too hard on any single number in the dashboard: what a tracked answer actually represents, and how far back your data goes.

An observation, not a personalized answer

When Peekaboo tracks a prompt, it's asking that question live, under a known set of conditions, and recording exactly what the AI said back. That's a real answer, but it's one answer, not the answer every person asking that question would see.

AI models don't give identical answers to everyone. The response can shift based on:

  • Personalization: a model's memory of that specific user's past conversations
  • Location: the same question can surface different local businesses or region-specific results
  • Account context: being logged in, subscription tier, or connected apps
  • Plain randomness: the same prompt asked twice can come back with a different brand order or a different set of citations, even with nothing else changed

None of this makes a tracked answer wrong. It means treat each result as a real, verified snapshot under controlled conditions, not as a claim about what every individual customer sees. Averaging across many runs over time is what makes the trend meaningful, a single day's number moving a few points is expected variance, not a signal.

No backfill before you started tracking

This matters most for two situations:

  • Proving improvement from a specific past event. If you want to show that a campaign, a rebrand, or a content push moved the needle, you need tracking running before that event happens, not after.
  • New accounts with no baseline yet. A brand new to Peekaboo has zero history by definition, and the visibility trend won't mean much until enough runs have accumulated.

What to do in your first 30 days

If you're newly tracking and don't have a trend line to show yet, don't wait on one. Lead with what's real right now:

  • Filter to Brand Mentioned in Recent Chats and show the actual AI conversations naming your brand today
  • Check the Sources tab for real, current citations, with exact URLs
  • Use these as your evidence while the trend line builds underneath them

Full guidance on this approach, including what to say to a client before you have real history, is in Reporting before you have a baseline.

Frequently asked questions

Can Peekaboo show me what a specific customer saw when they asked ChatGPT about us?

No. Peekaboo shows a real, controlled observation of that prompt, but personalization, location, and account context mean an individual customer could see a different answer.

Why did the same prompt show a different answer two days in a row with nothing changing on our end?

AI models are non-deterministic, so the same prompt can return a different brand order or set of citations even when nothing real changed. This is expected variance, which is why the score is averaged across many runs over time.

Can I see what AI said about my brand before I signed up?

No, there is no historical backfill. Measurement only begins once a prompt starts being tracked, so anything before that point was never recorded.

What should I show a client in the first 30 days when there is no trend yet?

Use Recent Chats filtered to Brand Mentioned and the Sources tab to show real, current evidence, then let the trend line build in behind it.