GA4 Says Claude Sent Almost Nobody. The Signup Survey Said 10.6%.
Otterly put a number on a problem a lot of founders can feel but can't prove. In Otterly's post, Claude looked like 0.1% of signups in GA4 and 10.6% in the signup survey.
That kind of gap matters more than it first seems. If you only trust last-click analytics, you'll cut the channels already putting your name in people's heads. That's the same kind of mistake people make with bmi calculation: they treat one clean number as the whole truth, even when it only shows one slice of what's going on.
How does bmi calculation relate to AI traffic attribution?
A single score can hide the real story when the thing you're measuring has more than one cause. Otterly measured signups two ways: GA4 tracked the last click before signup, and a post-signup survey asked people how they first heard about the product. Those two methods gave very different answers.
GA4 said Google Search drove 35% of signups, ChatGPT 7%, and Claude 0.1%. The survey said Google Search was 15.7%, ChatGPT 11%, and Claude 10.6%. Attrifast called that Claude gap 106× and made the obvious point: both numbers are honest inside their own method, but neither is the full picture.
That's why the bmi calculation analogy works. BMI can give you a rough read, but it can't tell muscle from fat, or fitness from illness. GA4 can give you a rough read on the final click, but it can't tell you who first heard about you in ChatGPT or Claude, then came back later through branded search, direct traffic, or a bookmarked tab.
If you're running a small SaaS, don't confuse a neat dashboard with ground truth.
GA4 misses AI demand when ChatGPT or Claude create the visit but don't get the click
AI assistants often act more like a recommendation layer than a referral link. Otterly says AI search behaves like a brand-awareness channel, which is why click-based analytics misses so much of it.
The path is simple. Someone asks ChatGPT, Claude, or Perplexity for a tool. Your product shows up. They remember the name, open a tab later, search your brand on Google, or type the URL direct. GA4 then gives credit to the last touch, not the source that created demand in the first place.
Otterly's own data shows why you shouldn't wave this off as edge-case noise. ChatGPT and Claude together accounted for more than 21% of self-reported signups. GA4 last-touch made Claude look nearly irrelevant. That's not a small miss. That's the kind of miss that leads a founder to stop writing, stop distributing, or stop testing AI discovery because the dashboard says it isn't working.
Attrifast pushes the same point from another angle. It says the problem is not bad intent or sloppy setup. The measurement layer itself breaks when discovery and conversion happen in different sessions or different channels.
A simple signup question will teach you more than another week inside GA4
You do not need a giant analytics rebuild to fix this. Otterly got its best signal from a post-signup survey that asked users how they first heard about the product.
For a solo founder or tiny SaaS team, the move is boring and effective:
- Add a required or near-required field after signup
- Ask: "How did you first hear about us?"
- Normalize answers into a few buckets: ChatGPT, Claude, Perplexity, Google Search, X, friend, review site
- Review those answers every week
- Compare them against GA4, not instead of GA4
Attrifast goes further and says even self-report has limits, because memory is messy and people compress multi-step journeys into one answer. Fair. But for a small team, rough first-party signal still beats clean-looking bad data.
Waiting for perfect attribution is how you end up learning nothing.
Three views together beat one clean chart
One dashboard will lie to you faster than three messy signals side by side. Otterly shows what last-click sees. Attrifast explains why self-report and payment-verified joins will disagree with it.
The useful operator move is to line up three views:
- GA4 source and conversion data
- Self-reported attribution from signup or post-signup survey
- Branded search lift after AI visibility, mentions, or citations increase
If GA4 clicks stay flat but survey mentions of ChatGPT and Claude rise, something is happening. If branded search for your product name rises at the same time, that is even stronger evidence that AI assistants are creating demand upstream.
That's the real answer hidden inside the bmi calculation query. Don't bet the business on one summary number just because it looks official. Use the cheap signal you can collect now, then stack it against the click data you already have.
If your signup form keeps saying "ChatGPT" and GA4 keeps shrugging, believe the people who just gave you money.