How Media Coverage of AI Girlfriends Has Changed

The early phase: coverage built around novelty

When AI companion apps first started getting mainstream attention, most of the coverage treated them the same way early smartphones or early social networks got treated: as a curiosity worth a reaction piece. The framing tended to lean on shock value. Headlines emphasized how strange or unprecedented it was that people would form an attachment to software, often without spending much time on how the underlying technology actually worked or how many people were actually using it.

That's a normal pattern for any new consumer technology category. Coverage of early ride-sharing apps, early dating apps, and early social media all went through a similar phase where the story was "look at this weird new thing" rather than "here's how this actually functions and who it serves." AI companion apps weren't unique in getting that treatment, they just got it more intensely because the subject matter (a relationship with software) is inherently more attention-grabbing than a ride-sharing app.

The shift toward taking the category seriously

As the number of AI girlfriend apps grew into the hundreds, and as the underlying language model technology became something most people had personally used in some form (through general-purpose assistants, if nothing else), the novelty framing started to wear thin. It's harder to write "isn't this bizarre" coverage about a category that a meaningful number of your own readers have already tried.

At that point, coverage generally split into two more serious lanes. One lane looked at the business side: how many apps existed, how they made money, what the competitive landscape looked like. The other looked at the human side: why people were using these apps, what void they filled, and what risks came with that. Both lanes required more actual reporting than the novelty phase did, since you can't write a credible business story or a credible human-interest story about a product you haven't actually used or tested. We keep our own running account of that business-side picture, including pricing, churn, and category-wide scores, in our data hub on the AI girlfriend industry.

Why the tone actually changed

Three things drove the shift from spectacle to specifics, and none of them are mysterious.

• Scale stopped being deniable. Once a category has hundreds of active products instead of one or two experimental ones, it stops being a curiosity and starts being a market, and markets get covered differently than curiosities.
• The technology became familiar. Large language models went from a research topic to something built into everyday tools. Once readers understand roughly how a chatbot works, coverage of a specialized chatbot product doesn't need to spend as much time explaining the concept from scratch.
• Real problems became documentable. Early coverage speculated about risks. Later coverage could point to specific, testable issues, like inconsistent memory, weak customer support, or products that quietly shut down, because those issues were now common enough to document directly rather than theorize about.

That third point is the one we care about most, because it's the entire reason a site like this one exists. Rather than speculate about what these apps might get wrong, we test all 129 of them directly and publish what we actually find, which is a very different exercise than writing a single trend piece.

The angles that show up again and again

Even as the overall tone matured, a handful of recurring angles never really went away, because they're genuinely the most interesting parts of the category.

Coverage keeps returning to questions about loneliness and social connection, about whether these products help or worsen isolation, about how younger users are engaging with the technology differently than older users, and about where the line sits between a helpful companion tool and something that could replace real human connection in an unhealthy way. Those are legitimate, ongoing questions, and they don't have a single settled answer, which is exactly why they keep getting revisited rather than resolved once and dropped.

What's changed is less the questions themselves and more the rigor behind the answers. Earlier coverage tended to answer them with speculation or a single anecdote. More recent, more careful coverage tends to lean on actual usage patterns, actual product testing, and a willingness to say "this varies a lot by app" instead of treating the entire category as one monolithic thing.

What even the more careful coverage still tends to get wrong

Even good-faith, well-reported coverage of this category runs into one recurring problem: it's genuinely hard to report on 129 different products as if they were one product. A story that accurately describes one platform's memory system, pricing model, or content policy can be completely inaccurate for the next platform down the list. The category is more fragmented than most consumer software categories, partly because it's still young and partly because the barrier to launching a new AI companion app is relatively low.

That fragmentation is also why generic claims about "AI girlfriend apps" as a single thing tend to age badly. Some platforms have genuinely strong memory and voice features. Most don't. Some have responsive customer support. Most don't, based on what we've found: 78% of the platforms in our database have no documented support channel at all. A single trend piece rarely has room to make that distinction clearly, even when the reporter wants to.

This is part of why we built our own testing process instead of relying on press releases or marketing copy. If you want the full picture instead of a single data point, our best AI girlfriend rankings break down exactly how each of the 129 platforms performs, category by category, rather than treating the whole space as one undifferentiated trend.

How we try to do this differently

We're not a news outlet, and this article isn't a claim that we've cataloged every piece of reporting on this category, because we haven't and wouldn't want to overstate that we had. What we can speak to directly is our own process: we maintain a live, continuously updated database of 129 platforms, we test each one hands-on across five categories, and we update our numbers when platforms change, shut down, or get acquired.

That approach is deliberately unglamorous compared to a splashy trend headline, but it's the only way we know to give an accurate answer to "is this app actually good" instead of "is this category interesting to write about." You can read the specifics of how that testing works on our testing methodology page, including how we score chat quality, memory, voice, images, and pricing independently of any affiliate relationship.

What to expect from coverage of this category going forward

Based on how every other consumer software category has matured, from ride-sharing to streaming to social media, it's reasonable to expect coverage of AI companion apps to keep splitting into distinct, more specialized threads rather than converging on one settled narrative. Business coverage will likely keep tracking consolidation and shutdowns, since our own data already shows real churn: about 18% of the platforms we track went dark, were sold, or rebranded within a single year of re-auditing.

Coverage focused on wellbeing and relationships will likely keep asking the same core questions about connection and isolation, because those questions don't have a single final answer and shouldn't be expected to. And product-focused coverage, the closest thing to what we do here, will hopefully keep getting more specific about which platforms actually deliver on their claims instead of treating "AI girlfriend app" as a single interchangeable product category.

That last shift, from talking about the category in the abstract to testing specific products against specific, repeatable standards, is the most useful direction coverage can go in, and it's the standard we hold ourselves to on every review and every data piece we publish.

Further reading

• The AI girlfriend industry in numbers
• How we test and score every platform
• About our testing process