Step 1. Read a rate, not an answer
The Visibility card describes itself as “Share of AI answers mentioning you”: the answers that name you, divided by the answers measured in the period. A single answer is one day, on one engine; a figure is a pattern across many, and only the pattern is worth acting on.
Step 2. Read the range and the label
Each card carries a confidence label: High confidence, Directional or Low confidence. The label folds in the number of answers behind the figure, the width of its 95% interval and the share of the period's days measured. The headline above the cards shows only High or Low, so a figure its card calls Directional shows as Low there.
Step 3. Check how many answers it rests on
Under each card, a line says how many of the planned answers the figure is based on; that is its coverage. An engine that kept failing to answer on a day is marked AI unavailable, and those answers are not counted as "not mentioned": a gap in the measurement is not a sign that you were left out.
Step 4. Read engine by engine
Set the Platform filter to one engine at a time: ChatGPT, Claude, Gemini, Perplexity, Grok or DeepSeek. Engines disagree with each other more often than most teams expect, so a figure for all six together can hide where you are actually missing.
Step 5. Compare periods only when aSERP does
aSERP compares a period with the equal period just before it, and declines a comparison that would not be fair, saying why: for example, “the method changed between the two periods” or “this period has too few measured days”. Don't rebuild by hand a comparison aSERP declined.
Step 6. Expect movement from one day to the next
In the consumer apps, ChatGPT and Google's AI Overviews Not tracked by aSERP each returned the same brand list less than 1% of the time across repeated runs of the same prompt (SparkToro and Gumshoe, reported by Search Engine Journal, 30 January 2026).[1] aSERP asks six answer engines the same questions, the same way, every day, and their answers still change. That is why a figure needs its range, and why one day's move is rarely news.
Step 7. Know which question each metric answers
Each card answers its own question, and each has a full entry in the glossary:
- Visibility: how often do answers name me?
- Prominence: when I am named, how early and how strongly am I placed?
- Share of answer: of the brands named, what share is mine, weighted by prominence?
- Citation rate: when answers name websites, how often is it mine?
- Answer stability: how steady is the mix of brands from day to day?
Check it worked
Right away
You can state, for each topic you care about, its range, its label and how many answers it rests on.
In aSERP
Read each topic again with the Platform filter set to each engine in turn. A direction you can trust shows on comparable periods, with intervals that no longer overlap.
With the free audit
The free audit is one capture of three questions: it has no range and no confidence label, so never compare its score with these figures.
If nothing moved
If a figure still reads Low confidence, wait for more measured days or track more questions. Answers to the same question tend to move together, so when a figure rests on few questions, its real uncertainty is wider than its range shows.
What not to do
- Don't screenshot one answer as proof.
- Don't average percentages across engines by hand.
- Don't treat AI unavailable as "not mentioned".
- Don't compare the free audit's score with Visibility.
Sources
- Search Engine Journal, reporting a SparkToro and Gumshoe study, “AI Recommendations Change With Nearly Every Query: SparkToro”, 30 January 2026.