Prescy does not give you a 0–100 visibility score. It shows a fraction: named in X of Y.
That line is the whole measurement. Once you can read it without translating it into a fake percentage, every other screen — Home, Prompts, Findings, the Monday recheck — makes sense.
Key takeaways
- X is how many sampled answers named your business.
- Y is how many answers were in that sample — the denominator you must never hide.
- ChatGPT and Perplexity are reported as counts, not blended into one proprietary score.
- A move of a few answers after no work is usually model wobble, not a win.
What the fraction means
Take Named in 3 of 5 on a tracked buyer question.
- 3 — ChatGPT named your brand (or a saved alias) in three of the weekly ChatGPT samples for that question.
- 5 — Monitor asks each core question up to five times on ChatGPT in the weekly run, because ChatGPT varies run to run. Perplexity is sampled five times alongside it.
So “3 of 5” is not “60% of the market.” It is “three of five ChatGPT answers this week named you on this exact question.” The denominator is what makes the claim auditable.
Operator rule: never paste “you’re at 60%” into a client Slack. Paste “Named in 3 of 5 ChatGPT answers on [question].”
Why not a percentage
Percentages flatten thin samples. “33%” sounds precise when the truth was “2 of 6.” Stakeholders argue about the number instead of the buyer question and the cited URL.
Named-in-X-of-Y keeps three things visible at once:
- Whether you appeared at all.
- How often, inside a known sample.
- Whether the sample was small enough that you should wait for another Monday before celebrating.
How to read it on Home
On a Monitor home screen, the headline fraction is your current mission context — usually the gap you are working this week, not a vanity average across every prompt.
- You lose on a question — X is low or zero; a rival may show a higher named-in count on the same question.
- Waiting for Monday — you marked a fix done; the baseline X/Y is locked until the next weekly run.
- Mission result — compare the new X/Y to the locked baseline. Improved, declined, mixed, or no clear change each suggest a different next tactic.
If X moves several points and nobody marked work done between runs, treat it as model wobble. ChatGPT does not answer the same way twice; that is why the sample exists.
How to read it on Prompts
Each tracked question shows its own fraction for the ChatGPT run. Anything under the full weekly sample means the answer changed between runs — useful variance, not a broken scan.
Intent chips (Compare, Best of, Alternatives, Category, Price / risk) tell you why the question is in the set. Quiet, off-intent questions that never name you are candidates to retire so a real buyer question can take the slot.
What Prescy measures today
Live Monitor measures ChatGPT and Perplexity. Google AI Overviews is a separate Search Console track for now (in testing inside Prescy). Do not tell a client that five engines are scored every week — they are not.
The free snapshot asks eight ChatGPT questions once. Monitor rechecks sixty core buyer questions weekly with the deeper ChatGPT (up to ×5) and Perplexity ×5 sample.