Which photo should go first, and why nobody can tell you

6 August 2026

The honest answer is that nobody can tell you which of your photos should go first, and that includes us. Every page confidently telling you otherwise is repeating a figure with nothing underneath it. Here is what is actually published, what the apps themselves do about it, and what that leaves you with.

The percentages have no source

Search this topic and within three results you will meet a precise-looking figure about how much of a swipe decision the first photo accounts for. There are two or three of these in circulation, they are always quoted to the percentage point, and between them they appear on hundreds of pages.

Try to trace one. The page quoting it cites another page, which cites a third, which states it flatly with no source at all, or credits a study that does not exist under that name, or an internal analysis that was never published. There is no paper at the bottom of the chain. A large amount of photo advice rests on those numbers and they are standing on nothing.

We are not going to reprint them here, even in order to argue with them, because reprinting is how they survive. We are also not inventing a replacement, which is the other move available and the worse one.

What has actually been published

The most relevant recent work is Dai and Xia, The impacts of media richness, blurriness, and beautification of online dating profile visual elements on dating outcomes, published in Frontiers in Communication on 8 April 2025. It put 389 participants through mock dating profiles under controlled conditions.

Two findings are worth having:

Profiles with multiple pictures were assessed more favourably than profiles with a single picture.
Blurriness had no statistically significant effect on the outcomes measured, which is not what most photo advice would predict.

What the study does not contain is any analysis of photo order. It compares profile types, not sequences. If you see it cited as evidence for which photo should go first, that citation is simply wrong.

More broadly: there is no peer-reviewed study testing photo sequence on a live dating app. We have looked for one. The reason is structural rather than mysterious. Doing it properly means manipulating real profiles served to real users and measuring real swipes, which requires a platform to cooperate and then to let the results be published. That has not happened in public.

The apps treat this as a per-person question

This is the strongest argument available, and it cuts against every universal rule on the subject, including any we might be tempted to offer.

Tinder announced Smart Photos on 13 October 2016. It alternated which of your existing photos was seen first, recorded how people swiped on each one, and reordered them accordingly. Tinder's newer tool, Photo Selector, announced on 17 July 2024, does a different job: it helps you pick candidate images out of your own camera roll, and the company's stated framing is that it assists the decision rather than making it.

Bumble ships Best Photo, which it describes as working out which of your first three photos gets the most right swipes and moving that one to the top of your profile. On 26 February 2026 Bumble added AI Photo Feedback, reported by TechCrunch, and the examples in that report are about what is visible in a picture, such as dropping photos where sunglasses cover your face and adding more variety, rather than about which photo beats which.

Notice the shape of all this. These companies have the data. For any individual profile they can see which image produces the best response, and they have been able to for a decade. If there were a general rule, publishing it would be trivial and would improve their product at no cost to them. Instead they built machinery to run the experiment per person, which is what you build when the answer varies by person.

Which means the honest version of "which photo should go first" is "the one that works for you", and the only way anyone has ever found that out is by measuring.

The one finding that does hold up

There is a robust result next door to this question, and it is useful precisely because it makes no claim about which photo wins.

White, Sutherland and Burton, Choosing face: The curse of self in profile image selection, published in Cognitive Research: Principles and Implications in April 2017, found that images chosen by strangers made more favourable first impressions than the images people chose of themselves. The contexts tested included a dating site. People are worse at picking their own pictures than an unfamiliar person is at picking pictures of them.

The practical implication is narrow and real: get somebody else to look. It does not tell you which photo to lead with. It tells you that your own ranking is the least reliable one in the room, which is a different fact and a more useful one.

What survives is about legibility, not ranking

Strip out everything unsupported and a small amount is left. None of it is a claim about which photo performs better. All of it is a claim about whether a picture can be read at all.

Your face has to be visible and large enough to read at the size the app renders it. A first photo is often shown small, and it is competing with a thumb that is already moving. A wide landscape with you somewhere in it fails before any question of appeal comes up.
One person in the frame. A group picture first asks the viewer to solve a puzzle before deciding anything. Most of them will not.
Recognisably how you look now. Also not a ranking claim. A picture that no longer matches you costs you at the meeting rather than at the swipe, which is the more expensive place to pay.

Those are not opinions about what is attractive, and they are not advice about performance. They are conditions for a photo doing its job at all, and they are why most failing first photos are failing.

How to actually settle it

Given all of the above, the defensible method is unglamorous.

Rule out the legibility problems first. They are objective, they are free to fix, and they do not depend on anybody's theory.
Ask somebody who is not you which picture they would lead with, and take the answer seriously even when it is not the one you like. That step is the one with a study behind it.
If the app offers to test the order for you, let it. It is measuring something real about your profile that no article can possibly know.
Change one photo at a time, and leave it long enough to mean anything before you judge the result.

Our free photo check covers the first of those. Every photo is judged on its own, never against your others. Where it goes runs from use this first to do not use this one. With that come six ratings and a short list of what to improve. Often there is nothing worth changing, and the list comes back empty. That is a good result, not a failure. Inventing a fault to look thorough is the same move as quoting a percentage with nothing behind it. The check cannot tell you which photo will win, and it does not claim to, for the reasons above. If nothing on the profile is working at all, the mechanical causes are worth ruling out before any of this, and those are in what is actually going wrong. And if you are weighing up a generated photo for that first slot, the question that matters is not ranking but whether it is a current likeness of you.

Written by Fottory, which generates dating photos from photos of your own face. The first two are free, and none of the advice above depends on using it.