Consumer AI is having its best moment in years and that is precisely why I have stopped recommending tools based on how good they feel in week one.
TechCrunch made the case this week that consumer AI is making a comeback. Meta’s personal assistant Muse and its plush-like mascot Jolly have been a surprise hit. OpenAI’s Dots launched and immediately found an audience. If you judge the category by downloads and chatter, things look healthy. If you judge it by the thing I actually care about as a reviewer, which is whether a tool will still exist, at the same price, with the same features, eighteen months from now, the picture gets considerably less pleasant.
Popularity and sustainability are not the same metric
The uncomfortable part of the TechCrunch piece is not that consumer AI is struggling. It is that consumer AI is succeeding and the economics still do not close. The reporting points at a growth ceiling that consumer products hit without enterprise revenue underneath them. Consumers are cheap. They churn. They use the expensive features most and pay the least. Enterprise buyers sign annual contracts and tolerate seat-based pricing. One of those groups funds infrastructure. The other one is a marketing channel that happens to cost money to operate.
That is the quiet shape of the business: your consumer AI app may be a demo for a sales motion aimed at somebody else entirely.
Why a reviewer cares about unit economics
I test toolkits. I am not an equity analyst and I do not pretend to be. But the financial structure behind a product determines the user experience in ways most reviews never account for:
- Free tiers that are too generous. Unusually good free access is almost never a gift. It is customer acquisition priced below cost, and the bill arrives as a paywall later.
- Quiet rate limits. The fastest way to fix a margin problem without a press release is to throttle. You notice it as the model “feeling worse” months after launch.
- Model downgrades behind the same name. Same label, cheaper model underneath. Your prompts stop working and nobody announced anything.
- Feature removal. The expensive capability you built a workflow around gets moved to a higher tier or deleted.
- Abrupt shutdowns. Consumer products without enterprise revenue are the first thing cut when a parent company reprioritizes.
Every one of those is a downstream symptom of the economics. So when I evaluate a tool now, “who pays for this” is a test criterion, not background trivia.
The scale problem nobody has solved
Zoom out and the discomfort grows. Ed Zitron’s argument, which got a long airing on Hacker News, is that AI requires enormous financial complexity and volume to work at all, and that this is not sustainable in the medium term. You can disagree with his conclusions and still concede the setup: this industry needs a lot of capital flowing continuously to stand still.
Reuters Breakingviews walks through the good, bad and ugly of AI’s economic impact and identifies real drawbacks, then flags a further risk that it says puts the others in the shade, citing Oxford economist Carl-Benedikt Frey. Stanford’s Chad Jones has been mapping out economic futures for AI that range from abundance to apocalypse. The honest read across all of this is that serious people with good data have not converged on an answer. That should make anyone pitching certainty about consumer AI pricing sound a little thin.
Regulation adds another cost layer. Colorado’s law, in effect since February 2026, regulates high-risk AI use in hiring, housing, lending, and healthcare. That is not aimed at chat assistants, but it signals where compliance spending is headed, and compliance spending comes out of the same budget as your free tier.
How I would actually buy right now
None of this means avoid consumer AI. Muse and Dots are getting used because they are genuinely useful, and useful is not a small thing. It means buy like someone who expects the terms to change.
- Avoid annual prepay on consumer AI unless the discount is large enough to absorb a shutdown.
- Prefer tools with data export that works. Test the export before you depend on the tool.
- Keep your prompts, context files, and workflows portable rather than locked into one assistant’s format.
- Treat anything free and expensive-to-run as temporary, and enjoy it accordingly.
- Check whether the company sells to businesses. If it does, your consumer tier is likelier to survive, because something else is paying the bills.
The comeback is real. The business model underneath it is still an open question, and the gap between those two facts is where users get burned. Review the tool, sure. Also review who is funding your use of it.
đź•’ Published: