Do People Really Use AI Girlfriends?

A quick, honest caveat about this data

I want to be upfront about something before going further. We don't have access to any platform's actual user counts, session data, or survey results, and I'm not going to pretend otherwise or invent a number that sounds authoritative. What we do have is a live database of 129 AI girlfriend platforms that I test directly, including their pricing structures, free tier design, and feature investment.

Those things aren't the same as a usage statistic, but they're not nothing either. Companies build free tiers, price tiers, and features based on what actually gets people to open the app and keep paying, not based on guesswork. Reading that structure carefully is a legitimate, if indirect, way to understand usage patterns, and it's the honest method available to us here.

What free tier adoption tells us

48% of the 129 platforms we track, 62 of them, offer a genuine free tier. Free tiers cost real money to run, since even basic chat requires ongoing compute, so a company only builds one at this scale if it's converting a meaningful volume of visitors into either paying subscribers or engaged repeat users worth retaining for other reasons. Nearly half the industry making that bet isn't consistent with a category that only a handful of curious people occasionally poke at.

The typical design pattern reinforces this. Free tiers almost always keep basic text chat open while gating voice, images, or deeper memory behind a subscription, which strongly suggests companies have already identified chat as the feature that hooks people in the first place, worth giving away to build the habit.

What pricing tiers tell us about real usage

Pricing structure is another useful proxy. Out of our 129 tracked platforms, 16 are free, 56 sit in a budget tier, 55 in mid-range, and only 2 have gone premium. The average starting price across platforms with parseable pricing comes out to $11.85 a month.

That clustering in the budget-to-mid-range band, and the near total absence of premium pricing, tells you something specific: this is a recurring subscription product competing on ongoing habitual use, not a one-time purchase or a novelty people try once. Products priced for repeat monthly engagement are priced that way because repeat monthly engagement is what actually happens, which is exactly why it's worth comparing options on our best AI girlfriend rankings rather than picking the first app you find.

What product diversity tells us about who's showing up

The range of product types across our 129 platforms is itself a usage signal. 104 platforms allow NSFW content while 25 are strictly SFW-only, and 22% now offer AI video generation on top of chat and images. That kind of specialization only makes commercial sense if different, real audiences are actually showing up for each variant. A company doesn't build a dedicated voice feature, currently offered on a minority of platforms and still averaging just 1.81 out of 5 industry-wide, unless enough real users are asking for it and using it to justify the ongoing cost of running it.

In other words, the fact that the market has split into distinct chat-first, image-first, voice-enabled, and video-capable products is indirect but genuine evidence that different groups of real people are using this category for different reasons, not one small group treating it as a single uniform novelty.

The scoring breakdown across the five categories we test tells a similar story: chat quality averages 3.26 out of 5, pricing 3.30, image generation 2.12, voice interaction 1.81, and customer support 2.21. A category with only token, one-off usage wouldn't generate enough competitive pressure to produce five genuinely different averages like that. Real, differentiated usage is what drives platforms to specialize and get scored so unevenly across features in the first place.

What memory investment tells us about repeat use

Memory is another useful, if indirect, signal. Only 21% of the 129 platforms we've tested document a real cross-session memory system, and building one is expensive and only worth the investment if a meaningful share of a platform's users are actually coming back across multiple sessions rather than trying an app once and leaving. The fact that any platforms have built this at all, even a minority, tells you retention is a real enough pattern in this category to justify serious engineering spend chasing it.

The flip side is informative too. The large remaining majority not investing in memory suggests a lot of usage in this category may still be shorter, more session-based, or more exploratory than platforms would like, which lines up with what we see in our breakdown of the market's actual scale: a large, still-consolidating category with real churn, not a fully matured one.

Who actually shows up, based on what platforms are built for

Reading backward from product design, a few real use cases stand out. People using these apps purely out of curiosity are served well by the free, chat-only tier that nearly half the industry offers. People looking for something closer to ongoing companionship are the audience platforms are chasing with memory and voice investment, even though only a minority have delivered on it well. And a substantial share of the market, reflected in that 104-to-25 NSFW-to-SFW split, is clearly using these apps for adult content specifically, which is worth naming plainly rather than talking around.

None of that is a demographic breakdown, and I won't pretend it is. It's what a careful read of platform investment and structure across 129 real products can honestly tell you, which is more than nothing and less than a proper survey would give you.

One more proxy worth naming: customer support investment. 78% of the 129 platforms we tested have no documented support channel at all. That's a weak spot for the industry, but it's also a small signal in itself, since a platform only bothers building a support channel when enough paying users are running into problems worth resolving. Its absence on most platforms doesn't mean nobody's using them, it more likely means support has lagged behind growth, which is a familiar pattern in young, fast-growing software categories generally.

How this connects to the bigger "how big is this" question

Usage patterns and market scale are related but genuinely separate questions. This piece is about who's using these apps and how, inferred from platform behavior. If you want the broader case for just how large and structurally real this category is as a whole, including the churn and consolidation happening across our full database, that's covered in our companion piece on whether AI girlfriends are actually a thing, which lays out the scale argument in full.

What would actually change my mind here

I'd treat this whole argument as provisional. If platform-reported user numbers, verified session data, or genuine survey research became available, I'd rather cite that directly than lean on proxies like free tier design and pricing structure. Until then, this is the most honest read of the evidence we actually have access to, built from testing 129 real platforms rather than repeating a number I can't verify.

If you want to understand the underlying category more before drawing your own conclusions, our complete definition of an AI girlfriend and our testing methodology are both good next stops, and you can read more about how I approach this kind of research on my author page.

Further reading

• Are AI Girlfriends a Thing? Yes, Here's the Scale of It
• What Is an AI Girlfriend? A Complete Definition
• The best AI girlfriend apps, ranked