Measurement
Why Your AI Visibility Score Moves
AI visibility scores move week to week even when nothing you control has changed. Here are the 6 sources of normal variance, the patterns that signal real movement, and what to watch.
07 Verify · 7 min
Your second run
Re-measure against a frozen prompt set, and read the per-prompt movement rather than the headline.
Conditions for a valid comparison
Three things must be true. The prompt set is unchanged from the baseline. Enough time has passed for the change to propagate — at least two weeks for a page edit, longer for anything off-domain. And you have written down in advance what would count as success.
If any is false, run anyway if you like, but be clear that you are collecting an observation rather than testing a hypothesis.
Read the prompt you targeted first
Your hypothesis named a specific prompt. Look at that one before the aggregate, because the aggregate blends the prompt you targeted with nine you did not touch, and that blend can easily hide or manufacture the effect.
Then read the actual response, not just whether the flag flipped. Sometimes the informative result is that you are now cited but still not recommended — which is not failure, it is progress to the next failure mode, and it tells you what the following cycle is about.
Where this lives in SolCrys
- Trigger the run the same way as the baseline.
- The runs surface lets you compare against the previous run.
- Read the targeted prompt's response in full before looking at deltas.
Lab: Run 2 of 3
Needs a free SolCrys account (no credit card). Spends run 2 of 3.
- Confirm the prompt set is unchanged and your waiting period has elapsed.
- Re-read the success criterion you wrote before shipping.
- Trigger the run.
- Read the targeted prompt first, in full, then the aggregate.
- Record what happened without interpreting it yet.
If quotas reset before you are ready, wait. A premature second run is a wasted third of your budget.
Start Free — free plan, no credit card.
Check yourself
- Was the prompt set genuinely unchanged?
- Did the prompt you targeted move, or only the aggregate?
- Did you move from one failure mode to the next rather than to success?
Go deeper
This lesson is the spine. These guides are the depth.
Measurement
Why You're Not Cited in AI Search: the Three Failure Modes a Prompt Re-Run Tells Apart (and the Fix for Each)
A prompt re-run sorts a citation failure into one of three modes - Absent, Mis-described, or Out-competed - and each routes to a different AEO fix.
How SolCrys Works
AI Visibility Measurement Methodology
How we capture your AI visibility data across supported engines, with each response traceable to a prompt, engine, capture method, available model or surface signal, and timestamp. Consumer-surface and retail-assistant validation are scoped where technically reliable.
Continue
- Previous: Why one run proves nothing
- Next: Signal or noise: the third run
Free · No credit card
Turn AI answer gaps into governed marketing execution.
Start free with a ChatGPT visibility read, then add multi-engine tracking, Corporate Context governance, and the action-to-result loop when you are ready.