What People Actually Think About AI Relationships

People often ask what "the public" thinks about AI relationships, usually hoping for a clean statistic like "X% of people approve." I don't have a verified survey number I'd stand behind, and I'm not going to cite one from a source I can't confirm actually ran a rigorous, representative poll. What I can offer is a more honest mix: general, common-sense observations about how sentiment toward new, once-unusual technology typically evolves, plus the one piece of real, verifiable sentiment-adjacent data we actually have, our own review numbers across 129 platforms.

Why we're not citing specific poll numbers here

A lot of content about AI relationships throws around specific-sounding percentages attributed to unnamed or vaguely named surveys. We'd rather not add to that. Without a visible, checkable methodology, a specific number is just a specific-sounding guess, and repeating it doesn't make it more true. Instead, this article sticks to two things: general, well-established patterns in how public attitudes toward new technology tend to shift over time, and our own genuinely verifiable data, which comes from real user reviews on the platforms we track.

The general spectrum: curiosity, stigma, and normalization

Public reaction to any genuinely new category of product, especially one touching on relationships or intimacy, tends to fall somewhere across three overlapping reactions, and AI companionship is no exception:

• Curiosity. A lot of people are simply interested in what this technology can and can't actually do, independent of any judgment about whether they'd use it themselves.
• Stigma. A meaningful amount of social discomfort or judgment still surrounds using an AI companion, similar to the discomfort that surrounded other now-mainstream categories of personal technology early in their existence.
• Normalization. Over time, as more people encounter the category, either directly or through people they know, casual, non-judgmental familiarity tends to increase, even among people who wouldn't personally use the product.

These three reactions coexist rather than replace each other in sequence. You can find genuine curiosity, real stigma, and growing normalization all happening at the same time, often in the same social circles, which is a pretty normal pattern for any technology that touches something as personal as relationships and companionship.

Our own review data as a real sentiment signal

Here's the one piece of genuinely verifiable data we have on this question, pulled directly from our own platform database rather than an outside survey. Across the 129 AI girlfriend platforms we track, users have left 473 reviews in total, averaging 3.67 reviews per platform. 124 of the 129 platforms have at least one real review on record, and only 5 have none.

This isn't a survey of general public opinion, it's specifically people who've actually used a platform and were willing to leave a public review of it. But that's exactly what makes it a genuinely useful, if narrower, signal: it reflects real engagement rather than a hypothetical opinion about a category someone's never tried.

What curiosity actually looks like in practice

A useful, concrete way to see the curiosity side of public sentiment is simply how often this category comes up in casual conversation compared to a few years ago, even among people who'd never personally use one of these apps. Explaining what an AI girlfriend app is no longer requires the lengthy setup it once did. That kind of ambient cultural familiarity, distinct from approval or disapproval, is itself a real sentiment shift, and it tends to precede more formal normalization rather than follow it.

What this review pattern actually suggests

The fact that 124 of 129 platforms have at least one review, rather than reviews clustering only around a handful of well-known products, suggests real, distributed usage across a wide range of platforms rather than attention concentrated on just a few. It also suggests people are willing to leave a public, identifiable trace of having used one of these products, which cuts somewhat against the idea that stigma is so overwhelming that usage stays entirely hidden. At the same time, an average of well under four reviews per platform is a small sample by any standard, and we're not going to overstate what it proves about broader public opinion.

How sentiment seems to be shifting, based on general observation

Without citing a specific poll, it's a reasonable, low-controversy observation that categories like this one tend to become less stigmatized over time simply through repeated exposure and mainstream media coverage, the same general pattern that's played out with plenty of other once-unusual personal technology. We'd describe current sentiment as somewhere in the middle of that arc: no longer a total taboo nobody will acknowledge, but not yet fully normalized either. Growing platform count, expanding features, and genuine (if modest) public review activity are all consistent with a category moving further along that normalization curve, even without a specific percentage attached to the claim.

How to evaluate sentiment claims you see elsewhere

If an article or site cites a specific statistic like "X% of people approve of AI relationships," it's worth asking who ran the survey, how participants were selected, and whether the methodology is actually published anywhere you can check. A specific number with no visible sourcing is not more trustworthy than an honest "we don't have a verified figure for this," and in a topic area this sensitive to sampling bias, that distinction matters more than usual.

Sampling bias specifically deserves a second thought here. A poll advertised through an AI companion app's own social channels will skew toward existing enthusiasts. A poll run through a general population panel that briefly defines an unfamiliar topic may capture reflexive discomfort more than considered opinion. Neither approach is inherently invalid, but neither produces a number that means the same thing as "this is what the public thinks," and headlines built on either kind of poll tend to flatten that nuance away entirely.

Why review data and survey data aren't interchangeable

It's worth being explicit about the limits of our own review numbers, too, in the interest of consistency. A review requires someone to have actually used a product, formed an opinion strong enough to act on, and taken the extra step of writing it down publicly. That selects for more engaged, often more opinionated users than a random population sample would, in either direction, strongly satisfied or strongly frustrated people are generally more likely to leave a review than someone with a lukewarm, unremarkable experience. So while 473 reviews across 129 platforms is real, verifiable engagement data, we wouldn't present it as equivalent to a representative survey of general public opinion, and we'd encourage the same caution toward any other source making a similar claim.

How we approach sentiment topics in our own coverage

We stick to data we can verify directly: our own review counts, our own testing scores, and general, well-established observations about how new technology categories tend to be received over time. We don't run our own public opinion surveys, and we're not going to present someone else's unverified numbers as if we did. You can see our complete review and rating data for every platform we track through our best AI girlfriend rankings, and read more about our process in our testing methodology.

Further reading

• Parasocial Relationships and AI: What the Concept Actually Means
• AI and Loneliness: What We Actually Know
• The best AI girlfriend apps, ranked