Key Takeaways
- Write down the number you are trying to beat before you start. The full cost, not the monthly one. An audit plus a year of watching is a figure you can name, and a return you cannot name is not a return you can check.
- Count enquiries, not mentions. A rise in how often an assistant names you is evidence the work is landing. It is not money. Only the enquiry line tells you whether the naming reached anybody who buys.
- Ask every new enquiry one question. How did you come across me. It is the cheapest measurement available, it works when nothing else does, and assistant referrals are precisely the kind that analytics tends to miss.
- Expect the crossing point to sit months out, not weeks. Work published today has to be found, read and then repeated before it turns into an enquiry. Judging it at week three tells you nothing except that three weeks is short.
- Stop paying if the enquiry line stays flat for two quarters. Mentions climbing while enquiries do not is a real finding, and the honest response is to change what you are doing or stop, not to buy more of it.
Somebody sold you on being visible in AI answers, or you did the work yourself over a few weekends. Six months later the fair question is whether it was worth it.
Most of the answers offered to that question are not answers. A screenshot of an assistant saying your name is not a return. Neither is a score that went up.
Here is the short version, the three lines worth tracking, where the break-even point tends to sit, what each method of measuring can and cannot prove, and the version you can set up this afternoon.
How do you measure whether AI visibility work paid for itself?
Keep two running totals and find the month one passes the other. Total spent since you started, and total value of enquiries you can trace back to an assistant. That crossing month is the answer.
It sounds too simple to be worth writing down, and it is the step almost nobody does. The usual failure is measuring only the half that is easy: mentions went from four to eleven, so it is working. Mentions are an input. They belong in the middle of the chain, not at the end of it.
The other half is harder because assistant referrals are bad at identifying themselves. A buyer reads an answer naming you, then opens a new tab and searches your name. Your analytics records a branded search, and the assistant that caused it leaves no trace at all.
The three lines worth tracking
Three, in order: how often you are named, how many enquiries arrive, and what those enquiries were worth. Each one only means something in the company of the next.
Naming first. Re-run the same set of buyer questions on a fixed schedule and count. The rule is that the questions cannot change between runs, because a different question set produces a different number for reasons that have nothing to do with you. Reading that number honestly is covered in how to read an AI visibility report.
Enquiries second. Add one question to whatever intake you already have: how did you come across me. Free text, no dropdown. A dropdown tells you which of your guesses people picked, and the whole point is to catch the answer you did not guess.
Value third, and keep it crude. Average engagement value times the number that converted is enough. Precision here is false comfort, because the numbers that decide the outcome are the enquiry count and the conversion rate, and both of those are lumpy for a small practice.
Where the crossing point actually sits
Later than people expect, and then it moves quickly. Spending starts on day one and accumulates in a straight line. Returns start at zero, stay there for a while, and then accumulate faster than the spending does.
- Paid out, running total
- Value back, running total
Month
By month 4 this practice has paid $2,096 and seen $2,000 come back, so it is still behind. By month 5 the totals are $2,495 paid and $3,000 back, and it is ahead. By month 12 it is $5,288 paid against $10,000 back. The crossing is a month, not a feeling.
Everything in that figure except the prices is illustrative. The $500 audit and the $399 a month are real published prices and are covered in what AI visibility actually costs, while what a month of watching involves is in what to expect for $399 a month. The enquiry rate, the conversion rate and the client value are assumptions, and they are the three numbers that decide the outcome.
Run it with your own figures before you believe anyone's version, including this one. If one in five converts rather than one in three, the crossing moves out past month 8. If your average engagement is worth $900 rather than $3,000, it does not cross inside a year at all. That is not a reason to distrust the method. It is the method telling you something.
Four ways to measure, and what each proves
Each method answers a different question, and each one is blind to something. Running two of them beats running the best one.
| What you measure | What it proves | What it cannot tell you | Effort |
|---|---|---|---|
| Mentions in a fixed re-run | Whether assistants are naming you more often | Whether any buyer saw it | Low, if the questions stay fixed |
| Referral visits in analytics | That some people clicked through from an assistant | The larger group who read your name and searched it | Low, already in your analytics |
| Asking every enquiry how they found you | Which channel a real buyer says caused the contact | Anything about people who never got in touch | Low, one field on a form |
| Enquiry value against total spend | Whether the money came back, and in which month | Whether something else caused the enquiries | Medium, needs a spreadsheet kept up |
The last row's blind spot deserves saying out loud. None of this is a controlled experiment, and a small practice will never have enough enquiries to run one. You are looking for a pattern strong enough to act on, not proof.
The measurement you can set up this afternoon
A spreadsheet with four columns and one extra question on your enquiry form. That is the whole system, and it will outperform any tool you do not actually read.
Columns: month, cumulative spend, enquiries you can attribute to an assistant, and their value. Add the row once a month, on a date you will remember. The discipline is keeping the question set for your re-runs identical, so the mention count means the same thing in month 9 as it did in month 1.
Two outside sources are worth knowing here. Google Search Central's page on AI features in Search explains how appearances in AI Overviews and AI Mode are reported alongside ordinary search performance, which is the closest thing to a first-party count on the Google side. And the generative engine optimization study from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi is the published measurement worth reading before you set expectations: it reports visibility gains of up to 40 percent from specific changes to source pages, and it also reports that how well those changes work varies by field. A number from one field is not a forecast for yours.
The damaging admission: for a lot of small practices the honest answer at month 12 will be that it has not paid for itself yet, and the spreadsheet is what tells you. That is the reason to keep it. A measurement you only trust when it agrees with you is a decoration.
If you have not got a baseline to measure from, start with the free check. It asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it, and the report lands within the hour between 7am and 5pm Central, or by 10am the next morning outside those hours.
Questions we hear the most
How do you measure whether AI visibility work paid for itself?
Compare two running totals: everything you have spent, and the value of enquiries you can trace to an assistant. The work has paid for itself on the month the second total passes the first, and not before.
Why not just track traffic from ChatGPT and Perplexity?
Track it, but do not rely on it alone. Plenty of assistant referrals arrive with no usable referrer, and plenty of buyers read an answer, then search your name separately, which lands in your analytics as a direct or branded visit.
How long before I should expect to see anything?
Months rather than weeks. A page has to be published, found, read and then repeated in an answer before a buyer acts on it, and each of those steps has its own lag. Judging the work at week three measures the lag, not the work.
Is a rise in mentions worth anything on its own?
It is worth something as an early signal that the work is landing. It is not proof of return, and treating it as proof is the most common way people talk themselves into spending another year on something that is not working.
What is the cheapest measurement that actually works?
Asking every new enquiry how they came across you, and writing the answer down. It costs nothing, it survives missing referrer data, and after two quarters it is the most reliable record you have.
What counts as an honest reason to stop?
Two quarters of flat enquiries while mentions climb. That combination means you are being named and it is not reaching buyers, which is a positioning or offer problem that more visibility work will not solve.
Does the free check give me a baseline to measure from?
Yes, and that is what it is for. It asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it.
What does a re-run cost after the first check?
Nothing for the free check, which you can run again. The $500 audit covers 125 checks across five assistants, and $399 a month keeps the board re-measured, or $300 a month if you add it with the audit.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →