Step 1. Pick one recommendation
Open Recommendations, filter by area (On-page, Off-page or General) and sort by Impact: high to low. This guide's example is a question-type recommendation, filed under General: Prompt Strategy, Branded Prompts or Awareness-stage Prompt Coverage. It changes the questions you track, so every engine you track reflects it from the next scan.
Step 2. Read why it was flagged
Open the recommendation and read Why this was flagged, What to do and Evidence. The evidence says what aSERP measured in the answers, and the product is plain about its limits: “It shows what was measured, not what caused it.”
Step 3. Add it to To-Do before you change anything
Choose Add to To-Do. As the recommendation itself says, “The baseline is captured the moment this is added to To-Do — so before and after can be compared.” Work you do before this step has no baseline to be compared with.
Step 4. Give it an owner and do the work
Choose Assign owner, then Start work. Do the task with the guide that fits it; for a question-type recommendation, that is Write the questions your buyers actually ask AI. Add new questions as new questions rather than rewording the ones you track, so your tracked questions go on being asked exactly as before.
Step 5. Mark it ready, then complete it
When the work is done, choose Ready for verification, then Complete. aSERP then compares the figures from before and after completion.
Step 6. Read the result
The result appears under Did it work?. It reads Observed improvement, No material change, Declined or Insufficient support, with an attribution level from No attribution to Strong attribution. Before the check period ends there is no result yet, so read nothing into the wait.
Check it worked
Right away
The moment you choose Add to To-Do, the recommendation leaves the suggestions and joins your To-Do work, with its baseline captured.
In aSERP
The check is the result. aSERP calls it an observed improvement only when the change is larger than the metric's minimum or the uncertainty of the two periods added together, whichever is greater. Even then, Observed improvement is an observation, not proof that your change was the cause. For this guide's example, the new questions are asked of every engine you track from the next scan: read them on the Topics page, as Write the questions your buyers actually ask AI describes.
With the free audit
The free audit cannot check a recommendation: it runs once per visitor and asks each question once.
If nothing moved
No material change means the check found no change larger than its uncertainty; Insufficient support means there were too few answers to judge. Keep the new questions running and read their topic over more measured days before you decide the work did nothing.
What not to do
- Don't do the work before you choose Add to To-Do: no baseline is captured.
- Don't read Observed improvement as proof that your change was the cause.
- Don't reword tracked questions to act on a question-type recommendation; add new ones.