I Built a SaaS. Then I Asked ChatGPT to Recommend It. It Had Never Heard of Me.

anthony papoti — Thu, 20 Aug 2026 02:56:47 +0000

A solo founder's distribution reckoning, in real numbers.

This morning I opened my analytics dashboard already convinced my product was dead. Nobody's signing up, I told myself. I'd been refreshing the same screen for an hour, watching a number that barely moved, feeling the whole thing slip away.

Then I did the one thing I'd been avoiding: I stopped staring at the realtime counter and actually read the data.

I build Katto, an AI tool that turns long videos into short vertical clips. It's pre-launch, no ad spend, no marketing budget, just me. And here is what the database actually said, once I looked past a slow Tuesday morning:

129 users. 58 of them in the last seven days. All organic.

That is not a dead product. So why did it feel like one? Because I'd been judging a half-empty morning instead of the trend, and because two silent leaks were quietly undoing everything the traffic brought me. Finding those two leaks was the real work of the day.

Leak one: the first run was failing

The first gut punch was the failure rate. On one recent day, close to six out of ten jobs failed.

I dug into the actual reasons instead of panicking, and the picture split cleanly in two. About half the failures were the download step choking on YouTube's bot wall: a user pastes a link, and the video never comes down. The other half were free-plan users bringing videos longer than the plan allows and hitting a hard cap.

The first half is a bug. The second half is a signal (people are showing up with real long-form content and getting a wall instead of an upsell), but that is a story for another post.

The download failures are the ones that hurt, because they are invisible to me and fatal to the user. Someone hears about your tool, signs up, brings their video, and the very first thing they try just dies. They don't file a bug. They don't email you. They leave, and they never come back, and they certainly never pay.

So I spent the day on unglamorous infrastructure: I migrated the backend to new infra and the download failures stopped. I did not take that on faith. I pulled the exact timestamps to make sure I wasn't fooling myself, and the last failure was the day before, with zero since the fix. Boring work. But the gap between a user who stays and one who's gone often lives in exactly this kind of boring work.

Leak two: even when it works, nobody finds it

The second gut punch was bigger, and no infrastructure fix touches it.

I opened Google Search Console. I rank on page two or three for basically everything. My own brand name ranks around position 28, because "katto" is a common word and I haven't earned the right to own it yet. (Curiously, "katto ai" ranks near the top, which tells me something about which name to lean on.)

Then I ran the test that actually matters in 2026. I asked ChatGPT, "best AI clipping tools 2026." It confidently listed ten tools. I was not one of them. I tried again with a few phrasings. Same result. The AI that a growing share of my potential users now ask for recommendations has no idea I exist.

That stung more than the failure rate. My product is good. People who try it mostly like it. And it does not matter, because the people who would try it never hear the name.

It's not an indexing problem, it's a penetration problem

Here is the reframe that turned the despair into a plan.

I am fully indexed. Google has my pages. The problem is not that Katto is hidden. The problem is that Katto is not written about anywhere that counts.

The tools an AI recommends, and the tools that rank, are the ones that appear across third-party pages: the "best OpusClip alternative" roundups, the review blogs, the directories, the creators who mention them. That is what a search engine and a language model both read as "this thing is real and people vouch for it."

A competitor I had genuinely never heard of has something like 1,300 referring domains pointing at it. I have maybe three. That single gap explains almost everything about why they surface in a generic answer and I don't.

The plan: 3 to 10 to 25 to 50

The mistake would be to look at 1,300 and try to chase it. That way lies buying garbage links off Fiverr and torching whatever trust I've built.

The honest target for a young product is not 1,300. It's 3 to 10 to 25 to 50 real referring domains. The rule of thumb people repeat is that somewhere around 25 to 50 genuine domains a new site stops looking new. I can't prove that exact number, but the direction is obviously right, and the first ten are gettable by hand, without paying anyone:

A handful of honest directory listings (a few done, a few to go). The first concrete one on my list is AlternativeTo, where the OpusClip alternatives page lists over fifty tools, submissions are free and community-driven, and there's even an "EU-based" filter with almost nothing in it. French founder, euro pricing: I'd be nearly alone in a niche no competitor can claim.
Two or three real reviews from people who actually use the tool.
A link in a genuinely neutral "best AI clipping tools" article. This one is harder than it looks, because most of those roundups turn out to be a competitor quietly ranking itself first. The truly neutral ones are rarer (a few independent blogs, community sites). So the other half of this move is writing my own honest roundup, which, awkwardly, is exactly what every tool that ranks already does, and exactly what the language models read.
My own founder presence, building in public, generating mentions over time.
One or two creators who try it and talk about it.

What I am deliberately not doing: buying backlink packages, and paying a hundred dollars a month for an SEO tool to confirm a problem I already understand. That money goes into outreach and into getting the product in front of real creators instead.

What I actually learned today

Three things, mostly about myself.

Your product being good is rarely the problem. Being found is. I spent months polishing features when the real gap was that almost nobody was arriving to see them.

Stop staring at the realtime counter. It is noise. A slow morning is not a trend, and doom-refreshing a dashboard is a way to feel productive while learning nothing. Look at the weekly number, once a day, and get back to work.

And distribution is not a thing you do after the product. For an indie founder with no budget, it is the product's twin, and it starts at three referring domains and a lot of unglamorous outreach.

The traffic is already coming, quietly, on its own. My job now is to stop losing the people who arrive, and to slowly earn my way onto the pages that decide who gets recommended.