A friend showed me her new dating photos last month. Good light, a café she has never been to, an expression she recognised as her own. She had generated them in about twenty minutes. Then she said something that stayed with me: “I like them. I just feel weird about liking them.”

That reaction is worth taking seriously, because it is not really about photography. It is about where we draw the line between presenting ourselves well and presenting someone else entirely.

Idealised self-presentation is not new

Before assuming AI created this problem, it helps to look at what people were already doing.

Research on how online daters build profiles has consistently found that users select idealised images: photos that exaggerate positive traits, and in some cases signal traits the person perceives as desirable but does not actually possess. A 2025 study of college-educated dating app users described participants curating between highly controlled selfies and staged professional shots, choosing images that communicated an aspirational version of themselves.

None of that required generative technology. It required a camera roll and a filter.

What psychologists have long called selective self-presentation is the baseline behaviour of online dating. You choose the flattering angle, crop the messy room, and pick the photo from the year you felt best. The medium invites it: asynchronous, editable, low in nonverbal cues, and therefore easy to optimise.

So the honest question is not whether AI introduced idealisation. It is what changes when idealisation stops requiring a good day, a good friend with a camera, or a budget.

What actually changed

Two things changed. The first is access. A professional lifestyle shoot costs somewhere between $300 and $500 and takes weeks to arrange. Tools that assemble an ai dating profile from a handful of ordinary selfies return 80 to 180 images across dozens of settings in roughly 20 to 30 minutes for a one-time fee under $80. Idealisation moved from a privilege to a commodity.

The second is category. A cropped photo is still a record of a moment that happened. A generated photo is a plausible depiction of a moment that did not. Both may look equally like you. Psychologically, they are not the same object, and users seem to know it even when they cannot articulate why.

That gap between “it looks like me” and “it did not happen” is where the discomfort my friend described actually lives.

What the research says about trust

The empirical picture is more nuanced than either the panic or the marketing suggests.

Work by Wu and colleagues investigated how a cue suggesting a dating profile was AI-generated affects impression formation, and found that perceived AI involvement changes ratings of attractiveness and trust. Notice the mechanism: it is the perception of AI involvement doing the work, not the image quality itself.

More recent research on generative AI in dating contexts, including Barkallah’s work on privacy and trust in AI-mediated dating, frames the technology as playing a dual role. It genuinely helps people present themselves, while creating an unresolved question about how much of that assistance should be disclosed. A 2025 analysis of visual deception in online dating found gendered differences in how AI-generated imagery is used and judged.

Long and colleagues took the comparison further, testing AI-generated photos against real professional photos alongside disclosure effects. The relevant finding for ordinary users is that the penalty attaches to detection and to concealment, not to enhancement as such.

In other words, the cost is not attached to looking good. It attaches to being caught having engineered it while implying otherwise.

The line is verifiability, not technology

This is where the psychology becomes practical.

Dating profiles are unusual among forms of self-presentation because they carry a scheduled reckoning. A LinkedIn photo may never be tested against the person. A dating photo almost always is, usually within a few weeks, in a coffee shop, in unflattering afternoon light.

That built-in verification is what makes exaggeration feel risky rather than clever. The discomfort people report is not moral squeamishness about technology. It is anticipatory: some part of them is already sitting at that first date, watching the other person adjust their expectations downward.

Which suggests a cleaner rule than “AI or no AI”. The question is whether the image is something a person who meets you would accept as a fair representation of your ordinary life. A generated photo of you in a café you have never entered can pass that test, if it looks like a Tuesday you might plausibly have. A photo that implies a physique, a lifestyle, or an age you do not have will fail it regardless of how it was produced.

Some tools have started building this into the product rather than leaving it to the user’s judgement. Scoring each generated image on how natural it appears, on a scale from 0 to 100, is essentially an attempt to operationalise the credibility threshold that people otherwise apply intuitively and inconsistently. Whether a number can substitute for that judgement is debatable. That the judgement needs making is not.

How to use these tools without the cognitive dissonance

If you are going to use them, a few principles reduce the discomfort rather than suppress it.

Generate widely, publish narrowly. The value of these tools is volume, which means the work is curation, not creation. Most of what comes back should be discarded.

Keep at least one photo taken by an actual person on an actual day. It anchors the set to a verifiable reality and, more usefully, gives you something you do not have to feel ambivalent about.

Test against the meeting, not the swipe. Before posting an image, imagine the person you are meeting seeing it after they have met you. If that sequence produces a wince, the photo is doing short-term work at long-term cost.

Do not use images that make claims your life does not support. This is where most of the reputational risk sits, and it is entirely avoidable.

The uncomfortable conclusion

The reason AI dating photos feel like cheating even when they look like you is that they make explicit something online dating always did quietly. We have been presenting curated versions of ourselves for as long as the profile has existed. AI removed the friction that used to disguise the effort.

That is not a reason to avoid the tools. It is a reason to be clear-eyed about what you are doing with them: not becoming someone else, but removing the excuse that you did not have a decent photo of who you already are.

My friend kept three of her generated photos and deleted the rest. She also kept a slightly blurry picture a colleague took at a work event, because, as she put it, “that one is definitely me.” That instinct is the whole thing, and no scoring system will ever fully replace it.

Adam Mulligan, a psychology graduate from the University of Hertfordshire, has a keen interest in the fields of mental health, wellness, and lifestyle.