Method

An open method, with the maths.

Every number Massyl shows comes with a range, and every claim of progress is tested against questions left untouched. Here is exactly how.

1. Ask each question twenty times

AI engines are not deterministic. The same question, asked twice, can name different businesses. So Massyl asks each of your questions 20 times per engine every month, spread across the weeks, and counts how often you are named.

Spreading runs across the month captures real day-to-day changes instead of one lucky (or unlucky) moment.

Asked once

100%

Last check: named

One answer is a coin toss. Next week the same check may say 0%.

Asked twenty times

40%

range 22–61%, n = 20

Named in 8 of 20 answers. The true rate is very likely between 22% and 61%.

2. Show the range, not just the number

Each rate comes with a 95% range computed with the Wilson score interval, which stays honest with small samples and with rates near 0% or 100%. If Massyl says 40% (range 22–61%), the true rate at which that engine names you is very likely inside that range.

AnswersRange when the true rate is about 40%
10–79%
512–77%
1017–69%
2022–61%
5028–54%
10031–50%

One question alone is a rough read. Your report adds questions together: 24 questions × 20 answers = 480 answers, and the overall rate narrows to about 36–44%. That is precise enough to see whether a change moved things.

3. Say “Not enough data” when it is true

When a range is too wide to act on, Massyl shows Not enough data instead of a confident number. When the answer is already clear (for example, never named in the first 10 runs), it can stop early, so your plan spends runs where they matter.

4. Prove fixes against a control

AI answers drift on their own, so “before and after” is not proof. Massyl keeps some questions untouched as a control group. After you publish fixes, it reports the gain of the questions you worked on over the control, with its range. If the ranges overlap, it says the change is not proven yet.

Mention rate, 24 questions+14 pts over control (range +6 to +22)
0%20%40%60%W37W40W43W46Fixes published
Questions you fixedControl, left alone95% rangeMade-up example

5. Keep every answer

Every answer is stored with its date, engine, country and language, and has its own link you can send to a colleague or client. The words that name you, or a competitor instead, are highlighted. When an engine states a wrong fact (hours, prices, locations, features), you see its words next to yours.

Try the maths yourself

The margin of error calculator uses the same formula. For the long version, read One AI answer is a coin toss and How to prove an AI-visibility fix worked.

Get your free 20-run report.

Twenty answers per question, every answer, every range. Free when Massyl opens, no card.

Free report when Massyl opens. One email, no sales calls.