Why there is no single benchmark
Lead volume from AI varies more by category than almost any channel, because two categories can sit at completely different points of the shift. A home care agency, where an unknowing adult child describes a crisis and gets introduced to the category, may see a large share of its total demand form in AI conversations. A commodity supplier whose buyers already know exactly what they want and price-shop directly may see almost none, yet. Same effort, wildly different volume, because the underlying exposure differs. Publishing a leads-per-week number without your category attached would be marketing, not math, which is exactly the kind of invented benchmark we refuse to put on this site. What travels is the method, not the number.
The estimate, in four inputs
Run these four numbers and you will have a defensible range. One, exposure: ask the ten questions your buyers actually ask across ChatGPT, Gemini, Claude, and Perplexity and count how many produce brand recommendations at all, that fraction is how far the shift has reached your category, and the free GEO Grader gives it to you in a minute. Two, moment volume: roughly how many of those buying conversations happen monthly in your market, estimable from your total addressable demand. Three, winnable share: what fraction of those moments you can realistically get named in, which starts small and climbs a few points a month as you build presence, the trajectory in does good GEO mean you show up every time. Four, conversion: AI-referred visitors tend to convert like referrals because they arrive pre-sold, covered in are AI-referred leads warmer. Multiply exposure by moment volume by winnable share by conversion, and you have a weekly range grounded in your business, not a slide.
Why the early number is small, and why that is fine
Set expectations correctly or you will kill a working program early. In the first weeks your winnable share is low, you are new to the moments, so the leads-per-week number starts small even in a high-exposure category. It grows as presence compounds: each covered moment, earned citation, and consistent fact raises your share, and share is what the lead count scales with. This is why the honest framing is a trajectory, not a launch number, the same compounding logic in why it is so hard to show up in AI. Teams that expect a big week-one number and quit are the ones who never see the week-twelve number, which is usually several times larger. Judge the slope, not the first data point.
How to count the leads you do get
You cannot estimate what you cannot see, and most AI leads arrive invisibly, as direct or branded traffic, so instrument before you judge volume. A how-did-you-hear field with an AI assistant option on every form is the single most important instrument, backed by branded-search and direct-traffic trends, per AI traffic analytics. Tag those leads in your CRM and watch their close rate, because a smaller number of warmer, higher-converting leads can outperform a larger number from a colder channel, which changes how you value the weekly count. Without this instrumentation, teams routinely undercount their AI leads by attributing them to direct traffic, then conclude the channel is not working when it quietly is, the reconciliation method in tying GEO results to revenue.
Turning the estimate into a plan
Use the estimate to size effort, not to make a promise. If your exposure test shows competitors named in most of your buying moments, the winnable-share upside is large and worth real investment, weigh it with how much to spend on GEO and AEO. If your category shows few named brands yet, expected volume is low today but rising, so the move is the cheap free-first playbook in how to fix AEO and GEO without spending a lot plus a quarterly re-check. Either way, the deliverable is a range you can defend and a slope you can watch, which is far more useful than a fabricated leads-per-week figure. The most honest thing we can offer is the method and the free baseline to run it on your own numbers.
