Guide
Revenue and ROI from AI answers: how the numbers are worked out.
GEO Style starts from the visits AI assistants already send you, then models what winning more answers could add, as a range, with the source of every input shown.
Updated , 3 min read
On this page6 sections
Measured first
Connect Google Analytics and Attribution shows the visits AI assistants send to your site, by assistant and over time, with the conversions and revenue Analytics records for them. These are measured figures, and the revenue model starts from them.
Modeled second, as a range
The model assumes your AI visits grow with the share of answers that name you. If you are named in a third of the answers to your buying questions today, being named in all of them would bring about twice as many visits again. That is the high end of the range; the low end assumes half of it.
- Extra visits, high end: your AI visits a month × (1 ÷ the share of answers that name you − 1).
- Extra visits, low end: half the high end.
- Money: extra visits × conversion rate × average order value.
A brand named in almost no answers would get an unbounded figure, so the share is counted as at least 10% and the page says when that floor applies. The result is labelled Modeled and shown next to the Measured figures it starts from.
Where each input comes from
Every input shows its source, in this order:
- Entered by your team. The conversion rate, order value and currency you set are stored for the brand with who set them and when.
- From Google Analytics. Used once it rests on at least 10 conversions: first those from AI visits, otherwise those from all visits.
- A default. A 2% conversion rate and an order value of 100 in your currency, marked as defaults until you replace them.
ROI: when the work pays for itself
ROI takes the modeled range and the share of it you expect to win back (half, unless you change it), subtracts your monthly cost (your plan's price unless you enter your own) and any setup cost paid up front, over the number of months you choose (12 by default). It shows a low and a high return and the month the running total turns positive, even when that falls after the period you chose. It is worked out on the server from the same inputs as Revenue, so the two pages agree.
What the model does not do
It does not put an amount on each question. Nothing yet says how much demand each question carries, so a figure per question would be invented.
It shows no range until there is something to scale from: Google Analytics connected, answers to your buying questions, and AI visits in the period. The page says which of these is missing.
It is not a forecast of what your work will achieve. Whether a finished action changed an answer is measured separately, by Impact, against similar questions nobody acted on.
A worked example
Example inputs, not a customer's data:
| AI visits a month (measured) | 300 |
|---|---|
| Share of answers that name you | 40% |
| Extra visits a month | 225 to 450 (300 × (1 ÷ 0.4 − 1) = 450; half is 225) |
| Conversion rate × order value | 2% × $500 = $10 a visit |
| Modeled range | $2,250 to $4,500 a month |
Revenue and ROI come with every plan. See them on the Revenue feature page, or compare plans.