I checked my analytics and found half my traffic comes from places I never marketed on
I built NotionLock — it adds password protection to public Notion pages (the missing middle between "public to anyone" and "invite-only Notion account required").
I barely marketed it. No ads, no launch. I shipped it and put the link in my IH profile. Then ~18 people signed up on their own.
So I finally dug into my analytics, and the breakdown surprised me:
- ~36% of my traffic comes from Indie Hackers — from a single passive profile link I almost forgot I'd added
- ~18% comes from AI assistants — ChatGPT and Copilot literally recommending the tool to people asking "how do I password protect a Notion page"
- The rest is organic search + direct
The AI part is what blew my mind. Nobody submitted it there. They just started suggesting it. And those visitors convert way better than search.
But here's the humbling part: of those signups, only ~17% actually protected a page. Turns out my setup modals were showing in Italian to English-speaking users 🤦 — a silent friction killing activation right at the critical step. Just fixed it, now watching if the new signups convert better.
Two takeaways I keep coming back to:
- The channels that work are rarely the ones you optimize for. I obsessed over SEO; my best traffic came from a link I forgot about.
- Check your source/medium in GA4 for AI referrals — you might be getting "AEO" traffic and not know it.
Building in public from here. Curious: is anyone else seeing AI-assistant traffic show up in their analytics yet?
The AI traffic part resonates. Put llms.txt on our site a few weeks ago mostly as an experiment — started seeing referrals from ChatGPT for queries I never optimized for. Nobody submitted us, they just started recommending.
The activation point stings though. We had the same thing — users creating QR codes but never scanning to test before printing. One prompt fixed it. Analytics hides this stuff until you go looking.
The llms.txt experiment is exactly the kind of thing I want to try next — thanks for sharing that it actually moved the needle for you. Did you do anything beyond just adding the file (structure, specific content), or was the plain version enough to start showing up? And yeah, the activation point stings precisely because analytics won't tell you — a real user finally emailed me describing where he got stuck, and that one message taught me more than weeks of GA4. Your QR-scan example is the same shape: invisible until someone shows you.
The AI referral traffic point is underrated. Most builders are still optimizing for Google while ChatGPT is quietly sending qualified visitors who already know what they want. AEO is going to matter a lot more than SEO for niche tools in the next 12 months.
Agreed — and the wild part is the AI visitors convert better than search for me, because they arrive already knowing what they want (ChatGPT framed the solution for them before they even land). For a niche tool that nobody actively searches for, being the answer the AI recommends beats ranking on a search nobody makes. Still figuring out how much of AEO you can actually influence vs. just earn by being genuinely useful.
That hidden-source traffic is often the best clue for where to double down. For Kinetic Override, I'm seeing the same pattern: exact-intent surfaces around no-root Android auto clickers / macro recorders are much more useful than generic launch posts. The harder part is tracking which wording people actually used before they found you.
The "what wording did they use before finding you" problem is real — I actually ran a small Google Ads test recently and it confirmed the opposite from the other direction: almost nobody searches the direct terms for my tool. Zero volume. Which means the intent isn't expressed as a search at all — people discover it through a recommendation or a context, not a query. Made me rethink the whole search-first assumption. Curious what's worked for you to surface those exact-intent moments.
Interesting signal.
The thing I'd be careful with is not where the traffic came from.
It's whether the people arriving from those channels are actually succeeding for the same reason.
A surprising traffic source can send founders chasing the wrong lesson if the real bottleneck sits somewhere else in the path.
That's not a call I'd make casually from analytics alone.
This is the sharpest comment here, and you're right to flag it. Since writing this post I actually got bitten by exactly what you're describing — I assumed my activation drop was one thing, then a user emailed me with the real reason (bugs in the entry flow I couldn't see). The analytics pointed at "where," but only a human told me "why." So yeah — the surprising traffic source is a clue, not a conclusion. I'm trying to treat the data as the question, not the answer.
That's actually the part I find most interesting.
Once you've seen data point one way and a user point another, the next challenge usually isn't collecting more information.
It's deciding what deserves to change your mind.
Hard to do justice to that in a couple of comments.
If you're interested, drop your email and I'll send over the fuller thought.
If you want to try it: notionlock.com — free plan, no Notion account needed for viewers.