
Nobody Admits to Having an AI Companion
Try this. Count the people you know who have talked to a companion app in the past month.
Now count the people who have told you so.
The gap between those two numbers is the most under-examined fact in this entire industry. Downloads across the category have surpassed 220 million, with hundreds of apps earning real revenue, and a large slice of the category built explicitly around intimacy — companions, roleplay, the best AI chat companion apps, all of it. And yet almost nobody says so out loud. Millions of people are doing something daily that they will not mention at dinner.
That silence isn’t neutral. It has consequences, and it has beneficiaries.
What the Silence Actually Costs
Start with the practical damage, which is easy to overlook because it’s invisible by definition.
If you can’t tell anyone you use an app, you can’t ask anyone whether it’s any good. You can’t compare notes. You can’t say does yours also get weird when you push back on it, or did yours also raise the price, or — most importantly — do you think this thing might be handling me? Every other consumer category is policed partly by people talking to each other. This one isn’t, and the vacuum is filled by whatever the companies say about themselves.
There’s a second cost. Shame concentrates use. Somebody who could mention it casually would use the app the way people use anything: sometimes, unremarkably. Somebody who has to hide it uses it in a sealed compartment, where nothing outside can moderate it and nobody notices the volume creeping up.
Which is worth setting against a finding people cite constantly and often misread.
The Research Reads Differently in the Dark
In 2025 a Harvard Business School team reported in the Journal of Consumer Research (De Freitas et al.) that an hour spent with a companion app relieved loneliness to nearly the degree an hour of human conversation did, leaving passive substitutes trailing. The drop came down to one thing: users left believing they had been heard. Not the brightness of the replies.
Then hold up the other result: a large MIT Media Lab and OpenAI study tied the heaviest everyday usage to deeper loneliness and firmer dependence. The direction of that arrow remains unestablished.
Put the stigma into that picture and a possible mechanism appears. A tool used openly and moderately looks like the first study. A tool used in secret, with no social friction and nobody to notice, looks a lot more like the second. The concealment may be doing some of the work we’ve been attributing to the technology.
Shame Is Excellent for Retention
Here’s the uncomfortable part, and it doesn’t require anyone to have planned it.
A user who can’t discuss the product cannot be talked out of it. They won’t hear a friend say that seems like a lot of money or you’ve been on that thing every night this week. They’re a closed system, and closed systems are exactly what subscription businesses want, because the main threat to a recurring charge is somebody outside noticing.
The naming conventions tell you the industry knows which door it came in through: around 17% of the live companion apps put “girlfriend” in their name, while only about 4% say “boyfriend.” A category shaped around that framing, in a culture that mocks it reliably, produces users who won’t talk. The mockery isn’t a headwind for these businesses. It’s a moat.
Nobody has to be cynical for this to work. The incentive operates fine on its own.
Privacy Compounds It
Secrecy raises the stakes on data in a way most people never work through.
One arXiv study, drawn from sit-downs with actual users of these products, caught both halves at once: with nothing present that could think less of you, the disclosure comes easily — and those very people go on to describe an active unease about the transcript’s ownership, and about how completely that question sits beyond their reach. Now add concealment. The transcript isn’t just sensitive because of its contents. It’s sensitive because its mere existence is something the user is hiding.
That is a materially different exposure profile from an ordinary app, and it should raise the bar on retention policies, export, and deletion — not lower it, which is what the silence effectively does. Nobody lobbies for better terms in a product they won’t admit to owning.
What Aigirlmates Says Out Loud
Doing this work in public is most of the point: at Aigirlmates we use these apps at length and publish the unglamorous findings — where the conversation actually falls apart, what a slow read of the privacy terms turns up, which pricing quietly escalates, and whether anyone still bothers opening the app once the novelty has burned off.
What Changes If People Talk
Not much, individually. Quite a lot, collectively.
Categories get better when their users compare notes, because comparison is how manipulative pricing, hostile data terms and engagement dark patterns get found. Right now this industry enjoys something close to a critical vacuum, maintained free of charge by embarrassment.
You don’t have to announce anything. But notice what the silence is doing: it isolates the user, it protects the company, and it removes the one mechanism — other people’s opinions — that would otherwise tell you whether the thing in your pocket is treating you well.
The apps aren’t the strange part. Two hundred million downloads and near-total quiet about it — that is the genuinely odd part.