Why your scores fluctuate

You didn't change anything, and your visibility score still moved a few points overnight. That's normal, and here's exactly why it happens.

AI answers aren't identical every time

Ask ChatGPT the same question twice and you can get two different answers, in a different order, naming a different mix of companies, with different sources cited. That's just how these models work, they aren't deterministic. So when Peekaboo re-runs the same prompt on your next scheduled check, a small amount of run-to-run movement in who gets named and in what order is expected, not a sign something's wrong.

Your score is built to absorb this: it's an average across every run in your selected time window, not a single snapshot. Averaging smooths out the noise from any one AI response over time.

How often your data actually refreshes

Your refresh cadence depends on your plan:

• Starter: every 2 days
• Peek and Grow: daily

Refreshes run on a rolling schedule counted from your last completed run, not a fixed calendar time. That means if one run happens a bit later than usual, your next one shifts by roughly the same amount, so your exact refresh time can drift slightly cycle to cycle rather than always landing at the same hour.

Behind the scenes, the system checks every hour for brands whose next run is due, so a run can start up to roughly an hour after its scheduled time. If you have a large number of tracked prompts, a full analysis cycle processes in short chunks rather than all at once, so it can take anywhere from a few minutes to about an hour for every prompt and model combination to finish once a run kicks off. During that window, your dashboard may show partial results.

A few other things that can move your numbers

• Google Search Console data lags a few days. GSC itself doesn't finalize the most recent 1-3 days instantly, so if you have Search Console connected, the very latest activity in that view may look partial until Google backfills it. Peekaboo pulls fresh GSC data once a day, and the dashboard's own summary tables refresh nightly, so today's numbers may not be fully caught up until the next early-morning refresh.
• Occasional fallback runs mean fewer citations. If a model's usual data path is briefly unavailable, Peekaboo automatically falls back to an alternate method to still get a valid answer. That fallback path can come back with fewer or no cited sources for that particular run, while the mention and score data itself is still valid. This is uncommon and self-corrects on the next run.

The bottom line

Expect small movement between individual runs, that's the models being models, not the tracking being unreliable. Watch the trend over a week or more, not the number on any single day, and you'll get a much more accurate read on whether your AI visibility is actually improving.