One month in: five things a free, no sign-up AI chat taught me

Hey everyone 👋

A month ago I posted here before launching GetAskAI, a free AI chat with no account, no app and no paywall. Some of you tried it and sent honest feedback, which was more useful than any of the congratulations.

Now there is real usage instead of assumptions. Here is what I actually learned. Some of it makes me look right. Some of it does not, and those are the more useful parts.

1. English is a minority language for us.

This is the one that genuinely surprised me. Real usage runs across all the languages we support, and the largest audiences are in Brazil and Germany, not the US. Removing the account wall turns out to matter most to people who are not in the English-speaking, credit-card-having, already-has-a-Google-account default. I built for that as a nice-to-have. It is the main audience.

2. The fair-use limit is invisible to humans and only bites bots.

In the launch post I wrote that the limit exists to stop automated abuse and that people would not feel it. That was a promise made before I could prove it, which always makes me uncomfortable. It has held: the overwhelming majority of visitors never come anywhere near it, while scripted traffic hits it immediately. I am not going to publish the exact threshold, for obvious reasons, but the design goal worked.

3. Most questions do not need a live web search.

Everyone in this space says "our AI searches the web". In practice only a minority of questions genuinely need it, and users are perfectly happy with the rest being answered directly. Routing everything through search would make the product slower and far more expensive for no benefit. Deciding when to search turned out to be a harder problem than searching.

4. Suggested follow-up questions are close to dead weight.

This is my least favourite lesson and the reason I am including it. It is a feature that appears in every AI product, it looks great in a demo, and in real use it is almost never clicked. I have not removed it yet, but I have stopped counting it as a win.

5. No accounts means no way to bring anyone back.

Here is the honest cost of the whole premise. People who ask one question usually ask a second one in the same session, so the product itself works. Getting them to come back next Tuesday is much harder, and there is no user ID, no email list and no notification to lean on. "Just add accounts" is the one answer I am not willing to take, so this stays an open problem.

On speed, since it came up last time: plain answers come back in a couple of seconds and searched answers take a bit longer. That gap is the entire reason we do not search on every question.

What I would tell anyone about to launch something similar: the finding that changes your roadmap is almost never the one in your pitch deck. Mine was the language split, and I only saw it because I shipped before I was ready.

The question I actually want to ask this community: if you run a product without accounts, how do you measure retention? Everything I have tried is a proxy for something I deliberately made unmeasurable. And for those who added accounts later, did you lose the thing that made people try you in the first place?

Happy to go deeper on any of these 🙏