Disclosure: I am not employed by Photofeeler and I do not receive referral fees or commissions from them.
The short version
You upload one photo. In dating mode, people in the audience you choose rate it on three traits: Smart, Trustworthy and Attractive. The final scores run from 1 to 10, standardized against other photos from people your age and gender. A 5 is average. The results page also shows the vote breakdown and any notes people left.
The votes come from people, not an image model. You pick the voter gender and age range, as well as how many votes you want. Photofeeler then adjusts for people who rate everything high, rate everything low or seem to be clicking through fast to earn credits.
That makes it good for screening photos and comparing first impressions. It does not replace app results. Once you have two or three finalists, put them back on Tinder or Hinge and see what happens when your prompts, photo order and the rest of the profile enter the picture.
Photofeeler also has Business and Social categories for LinkedIn headshots and general social media.
It rates the photo, not you
This is the part people miss. The same guy can score 3.7 in one photo and 9.0 in another. Even a slight change in pose or a different smile can drastically change the results.
Mirror selfie at a barber shop. Backpack still on.
Suit on a rooftop at sunset. This became his main.
These photos were tested on Photofeeler during the same week. The "Before" shot is the photo he had picked as his primary before I worked with him. That is where the tool helps, even if you never hire a photographer. It catches bad self-picks fast.
His 'objective' attractiveness did not change, per-se. But the lighting, expression, outfit, pose and background did. Those are the levers and they're under your control.
So if you have ever told yourself, "I'm just not photogenic," that is usually the wrong takeaway. Most of the time the issue is the photo itself. You do not have a fixed PF rating ceiling.
If you do one thing after reading this, test your current photos. Most men are worse at choosing their own primary than they think.
How voting works
Voters see one photo at a time. The traits depend on the category: in Dating they are Smart, Trustworthy and Attractive, in Social the columns are Confident, Authentic and Fun. Each trait is rated on a 0-to-3 scale (No, Somewhat, Yes, Very) and voters are told to judge each one separately. That matters because someone can read attractive and untrustworthy at the same time.
Below the buttons there is a free-text field and a panel of Quick Notes. The Quick Notes are split into Feeling and Suggestion. Feeling is how you come across: "great smile!," "fake," "intense." Suggestion is what to change: "blurry," "too close-up," "smile more." The score tells you where the photo landed. The Quick Notes and free-text comments sometimes tell you why.
Most voters just pick a few Quick Notes and move on, but the free-text field does get used. In my experience women will occasionally write something genuinely useful there, a specific reaction you would not get from Quick Notes alone. Your photo is only visible while the test is live, then private again.
The one-photo-at-a-time setup makes sense. Dating apps work like that. Nobody sees your top two candidates side by side and compares them calmly. They see one photo, form a quick impression and move on. A/B testing answers a different question. It also misses the very common case where both photos are weak.
How the scoring is adjusted
Most people assume Photofeeler just averages the votes. It doesn't.
Some people hand out 3s to everybody. Some barely ever go above 1. If every vote counted the same, a soft voter pool would inflate your score and a harsh one would drag it down.
Photofeeler says it adjusts for that. A 3 from someone who almost never gives 3s counts for more than a 3 from someone who gives them out freely. People who mash the same button over and over get their votes downweighted or thrown out. The site also runs fraud detection, bans, spacing between tests from the same account and blocks voting twice on the same test.
The 2019 paper found that this kind of normalization roughly doubled the information per vote. So 20 adjusted votes tell you more than 20 raw votes on a site that just averages everything.
Why all three traits matter
Attractiveness gets you through the first gate. Without it, the rest barely matters.
Smartness and Trustworthiness still shape who responds. Photofeeler's own analysis suggests those scores track a different kind of appeal than pure sex appeal. If you want matches who seem more relationship-minded, those two scores matter more than a lot of guys think.
Sitting at the barber in a gym tee. Good smile but it reads like a Yelp review photo.
Parka and Christmas lights. Scored highest in his set despite being the most casual.
This is why I do not chase one trait in isolation. A photo can get hotter and less trustworthy at the same time.
This is why the shirtless gym selfie so often disappoints. It might bump Attractiveness a bit while crushing Trustworthiness. On Hinge, a woman is asking two questions at once: "Is he hot?" and "Would I actually want to meet this guy?"
The same logic applies to the rest of your profile. Some photos are there to prove you look good. Others are there to make you look like an actual person. A hobby shot, travel moment, dog photo or slightly goofy image can be worth keeping even if its raw Attractiveness score is lower, because it gives someone something human to react to.
Is the Photofeeler voting pool good enough?
The voter pool matters, but probably less than you think. It's true that these are random internet strangers rather than the exact people you want to date. But Photofeeler has published data showing that more and less attractive voters did not rate photos meaningfully differently. A separate study they've done found random voters tracked reasonably well with actual hiring managers. You can pick at the methodology but in my experience it does work well.
The other limitation is niche photos. A subculture-specific shot can score lower on Photofeeler and still work better on your actual app because the voters rating it are not the people it was meant for.
Can't AI do this?
Photofeeler is still 100% human-rated. The same 2019 paper trained a neural network on more than a million dating photos. It matched roughly 10 human votes, which is statistically shaky. The model was never released, so sadly you cannot try it yourself.
It also did worse on men than on women. Women disagreed with each other more, which made male attractiveness harder to predict. That tracks with what most guys learn the hard way. And it is the exact situation where a model is least useful and real votes matter most.
How many votes you need
Every score comes with a confidence interval — a range where the "true" score probably lands if you kept collecting votes. Photofeeler itself recommends checking that interval to make sure the sample is statistically valid before drawing conclusions.
Personally, I find it easier to skip the interval math and follow a simple rule of thumb:
- 10 votes — enough for a quick yes/no when you are running through a batch and just need to cut the obvious losers.
- 20 votes — the sweet spot for every photo except your primary. Spending more would just burn credits needlessly.
- 40 votes — reserved for primary pic candidates, where a small gap in scores actually matters and you want the tightest confidence interval you can get.
What the critics get right
Most criticism of Photofeeler is fair enough. The question is what each complaint means in practice and when it should change your decision.
"Photofeeler is not Tinder."
That part is true. Tinder and Hinge judge the whole stack: photos, prompts ordering, app culture and whether the profile hangs together. Photofeeler isolates one variable, which is why the score can disagree with your match rate.
"A niche photo can score lower and still pull better matches."
Also, yes. A goofy costume shot, a rave photo or something that signals your subculture can look mediocre on generic appeal and still outperform with the specific people you actually want. That generic signal still helps. It is not the same thing as appeal to your exact type.
"People just click fast to earn credits."
The credit system absolutely creates pressure for junk votes. Photofeeler says it filters or downweights bad raters and I assume some of that works but I still would not treat a thin sample like gospel. More votes help drown out the drive-by clicks.
"The same photo got different scores when I reran it."
Reruns drift, especially when the sample is small or the photo is polarizing. That is annoying but it is not mysterious. Check the score range and do not make big decisions off 10 shaky votes.
"A high score does not guarantee dating-app results."
This is where people overread the score. A strong result tells you the photo landed well on a first pass. It does not tell you whether everything else on the profile closes the deal.
How to use Photofeeler without fooling yourself
Most people waste time the same way: they upload random leftovers, overread tiny score gaps and blindly pick the top number. Here is the approach I use instead:
Start with technically solid photos
Do not expect useful feedback from dark, blurry, low-effort shots. If the photo looks sloppy, the result mostly tells you it looks sloppy. Start with images that are at least decent.
Show your eyes and test solo first
For primaries, visible eyes usually beat sunglasses and solo shots remove the "which guy is he?" problem. Get the clean read first. Save group shots and weirder photos for later.
Use enough votes for the decision you are making
As mentioned earlier: 10 votes for a quick yes/no on a batch, 20 for everything except the primary, 40 for primary/first pic candidates.
Build a profile, not a leaderboard
A dating profile needs range. Three strong anchor photos plus a couple that show humor, lifestyle or warmth usually beat six polished photos that all say the same thing.
If you are brand new to online dating and your current camera roll is weak, do not spend a week trying to squeeze certainty out of bad inputs. Get better candidate photos first. Then use Photofeeler to compare the finalists. It cannot rescue a weak profile.
Used that way, it becomes a sorting tool vs being an 'all knowing' magic tool.
How I use it at GetDatingPhotos
My guarantee is a 9.0+ Attractive scoring pic or a full refund. That means every shoot gets tested before I deliver.
During photoshoots I capture a batch, then periodically test the top candidates with 20-40 Photofeeler votes, figure out what does/doesn't score well and why - and adjust accordingly. That is the core of the 3-day process. Instead of me saying "I like this one" or your friends saying whatever feels supportive, I get 20-40 women's actual first impressions.
I always ensure to have many backup candidate photos with a good Photofeeler score to allow for future experimentation. This matters because, while highly correlated, the dating apps and Photofeeler aren't always fully aligned.