Matches, but nobody messages me
Getting matches and getting messages are two different problems. They fail for different reasons and they have different fixes. A match costs somebody one movement of a thumb. A first message costs them a sentence they have to invent. That second cost is where a lot of perfectly good profiles quietly stop working.
A swipe is a decision made at speed on one image. A message is written after somebody has looked at the whole profile and thought about it. A profile can pass the first and stall at the second, and that is not a mysterious failure. It is the ordinary one.
The advice here is prescriptive, and that is the gap
Search this question and you get lists. Add a photo with a conversation starter in it. Here are the six kinds of photo every profile needs. None of it is wrong. None of it is diagnostic either. It tells you what to add. It cannot tell you what your own photos already do. That is the part you cannot see for yourself, because you know the story behind every one of them.
The apps have already built for this
Hinge attaches a like to one part of your profile. You tap a heart on a specific photo or Prompt, and that choice travels with the like. Hinge announced Convo Starters on 8 December 2025. It appears when a dater taps to send a like, and offers three tips personalised to that photo or Prompt. The same page states that "72% of daters are more likely to consider someone when a like includes a message". It adds that Hinge data shows likes sent with a comment are "twice as likely to lead to a date".
Those are Hinge figures about likes rather than about matches. They still describe the same cost. Somebody has to find something in your profile to say. Hinge built a feature because enough of them could not.
Tinder came at it from the other end. Its press room announcement of Prompts calls them great conversation starters you add to your profile. What that concedes is the interesting part. Nobody ships that unless profiles often leave people with nothing to say.
So we made it a named rating
Our free photo check returns six ratings for one photo. One of them is conversation, and the reader sees it labelled "Something to ask about". The definition is deliberately narrow: whether the frame holds something specific a stranger could ask about. Four levels, no number, and no middle level to retreat into.
- Strong. Something specific and safe to ask about. An activity in progress, an instrument, a recognisable place, a dish, a craft.
- Solid. Real context. A setting or an object that hints at a life, with no one obvious question in it.
- Fair. A plain portrait. Pleasant and blank, giving a stranger nothing in particular.
- Weak. Nothing to say about it at all, or the only context is a wall, a car interior or a bathroom.
The line that matters is Fair. A photo can be clean and still land there. Well framed, face clear, nothing wrong with it, and nothing in it. That is not a failure of the photograph. It is a fact about the photograph, and it explains a quiet match list.
A rating means nothing without the job the photo is doing
The check also says where a photo goes. There are five answers, running from use this first down to do not use this one. That placement changes what the conversation rating costs you.
The opening photo has one job. Somebody has to know who you are on sight. A plain portrait is often exactly the right picture for that, so Fair on conversation costs it nothing at all.
A middle photo has close to the opposite job. It shows you visibly doing something real, where the activity carries the picture. Fair there means the middle of your profile is not doing its work. Same photo, same rating, two different verdicts, decided entirely by position. It is also why the ordering question has no general answer.
What the check will not tell you
Here is the limit, and it is a real one.
Only a Weak rating owes an explanation. That is enforced in the schema. A Weak with nothing accounting for it fails the parse and is asked for again. Fair owes nothing, and that is deliberate. Fair means usable with a limitation. A limitation written up as a fault reads to the owner as a defect. The improvement list is capped at three, worst first, and an empty list is a common and correct answer.
Put those two rules together and you get the case this page is about. The pleasant blank portrait rates Fair on conversation, and nothing is printed underneath it. The bar shows two segments of four, with Fair beside it and no line under it.
We are not going to promise otherwise. What is missing from a photograph is not in the photograph. We can see that there is no question in the frame. We cannot see what you do on a Saturday, and a tool that guessed would be making your life up.
Two of the six are measured, four are judged
Worth knowing which is which, because it decides how much a rating is worth.
Can they see your face, and Light and focus, are computed from your original file. No model reads them off. What goes in is brightness, contrast, clipping, resolution, sharpness and the face box, against thresholds fixed in code. Under 640 pixels on the long edge is Weak. So is a face sharpness under 55, or a whole-frame brightness under 62 of 255. A face darker than 70 of 255 pulls the face rating down to Fair. Run the same file twice and those bars are identical, forever.
The other four, conversation among them, are a vision model looking at one picture. That is an opinion, and we say so on the result rather than dressing it as a measurement. An opinion about what is interesting in a photograph is worth having. It is not worth pretending it is arithmetic.
What we will not sell you
Some faults cannot be corrected, and the honest thing is to say so instead of charging for the attempt. Focus and resolution were never captured. No corrected version can add detail the file does not hold, so no repair is offered on those at all. The same holds for a screenshot, and for a photo of your own reflection. Editing cannot make either into a picture of you out in the world.
Conversation is the clearest case of the lot. Nothing can put a craft in your hands that was not in the frame. That is not a correction. It is a different photograph, and selling it as a correction would be a lie.
The question nobody can answer for you
One more thing, plainly. The check never rates the person. No attractiveness, no score, no number, no percentile. It does not infer your age, your gender, your personality or how you come across. It does not compare two of your photos against each other. That is a legal constraint rather than a preference, and it is why every photo is judged alone.
So it will never tell you that you seem unapproachable. That is the suspicion sitting behind this whole question, and it is the one thing no honest tool can measure. Anything that answers it is guessing, and charging you for the guess.
Judging one photo at a time has a second consequence worth owning. The check cannot see your set. Six photos that are each fine on their own can still add up to nothing to say. Reading six results, noticing that every one is a portrait, then going out for a seventh: that part is yours.
What is actually left to do
Go through your photos one at a time. For each one, name out loud the question a stranger could ask. Not whether the picture is good. Whether there is a question in it.
If you cannot name one, that photo is not carrying this job. That is fine for the photo that opens your profile. It has a different job and it does that one well. It is not fine for every photo you have.
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.