AI recommendation rank

What « rank » means here

Share of voice tells you how often you are cited. It does not tell you whether you come first or seventh. Rank is computed from the order in which entities appear in the answer: #1 means your brand is the first one named. That is the phrasing a client understands: « on Perplexity you are the third recommendation, behind these two ».

The four counters

• Tracked prompts: every prompt in the project with at least one reading.
• Mentioned (at least 1 engine): the ones where your brand appears somewhere.
• Top 3: prompts where you are among the first three brands named — the only threshold that really matters in a conversational answer.
• Top 10: the wider count, useful for spotting prompts about to tip over.

Reading a prompt row

• The Best badge gives your best rank across all engines; medals mark the top three places.
• The « *n*/*n* engines mention you » line tells you immediately whether the problem is general or specific to one engine.
• On the right, one pill per engine shows your rank there, or a dash if you are absent.
• Unfolding the row details each engine: Best, Latest when it differs, average rank, number of scans and the date of the last one.
• Competitors ahead of you names who gets there first, and how many times. It is the most useful line on the page: it turns a score into a list of rivals to displace.

Average rank trend by engine

Once at least two months of readings exist, a chart shows the average rank per engine, month by month. The axis is inverted: a curve going down is good news. This is the chart to put in a client report — far more eloquent than a percentage.

Checking for yourself

• Copy prompt puts the exact wording on your clipboard.
• Test live opens the prompt pre-filled in ChatGPT, Perplexity, Claude or Google's AI mode — handy for comparing the reading with what a real user sees, on the spot.