Your Face Gets Judged in 100 Milliseconds

Last updated: September 26, 2026. Every study on this page is linked to its original paper, not to somebody else's summary of it.

You have probably read that people judge you in a fraction of a second. It gets repeated constantly, usually with no source, usually with a number someone made up.

There is real research underneath it, and it is more specific and more useful than the version that circulates. It was measured, the exposure times were varied deliberately, and the finding that matters most for a profile photo is not the one that gets quoted.

Here is what the studies actually did, what they found, and the one result that should change which photo you upload.

The short answer

• Judgements made after seeing a face for 100 milliseconds match the judgements people make with unlimited time. Extra looking time did not change the verdict. It made people more confident in it.
• Agreement with unhurried judgements is already above chance at 33 milliseconds, and stops improving after about 167 milliseconds. Roughly a sixth of a second is the whole window.
• Competence judged from a one second glance at a face predicted 68.8 percent of the 2004 United States Senate races, and tracked the winning margin.
• The finding that matters for your headshot: two photos of the same person can produce more different impressions than photos of two different people. Your face is not one first impression. Each photo of it is a separate one.

What the 100 millisecond study actually measured

Janine Willis and Alexander Todorov, at Princeton, published the original in Psychological Science in 2006. People saw unfamiliar faces for 100 milliseconds, 500 milliseconds or 1,000 milliseconds, then rated them on five things: attractiveness, likeability, trustworthiness, competence and aggressiveness. A separate group rated the same faces with no time limit at all.

The 100 millisecond ratings lined up closely with the no time limit ratings. On all five traits. That is the whole result, and it is narrower than the way it usually gets told.

What changed with more time was not the judgement. It was confidence. People given longer looked at the same face, reached the same conclusion, and felt more sure about it. The extra time bought certainty, not accuracy.

Trustworthiness was the fastest of the five to settle, which makes a kind of sense. Deciding whether a stranger is safe is the older problem.

How fast is fast

Three years later Todorov, Manish Pakrashi and Nikolaas Oosterhof went shorter, with better masking so people could not keep processing the image after it vanished.

A sixth of a second. That is less time than it takes to blink, and it is roughly the time someone spends on your photo while scrolling a list of applicants or search results.

The result nobody quotes, and the one that matters

In 2014 Todorov and Jenny Porter published something more awkward in the same journal. They showed people multiple different photographs of the same individuals.

Impressions of one person shifted depending on which photo was shown, and the variation between different photos of one person was as large as, or larger than, the variation between photos of different people.

Read that again, because it is the opposite of how most people think about a headshot. You do not have a face that makes one impression. You have a set of photographs, and they disagree with each other about you more than strangers' faces do.

The same year, working separately, Richard Vernon and colleagues at York reached the same conclusion in PNAS using what they called ambient images, meaning ordinary photographs with all the variation in lighting, angle and expression that real pictures have. Their wording: first impressions can show as much variability between photographs of the same individual as between photographs of different individuals. Two labs, two methods, one answer.

Todorov and Porter also found that the photo people chose shifted with the purpose. Asked to pick an image for online dating, people chose differently than when picking one for a political campaign. Same face, same photo set, different job to do.

For anyone choosing a profile picture, that is the practical finding. The question is not whether you photograph well. It is which of your photographs is doing the job you need doing.

The uncomfortable one about elections

The 2005 paper in Science is the one that made this field famous, and it is worth knowing because it shows the size of the effect outside a lab.

Todorov and colleagues showed people photographs of the winners and runners up in United States congressional races. The participants did not recognise the candidates. They rated them on competence after a one second look.

Those competence ratings predicted the actual outcome of 68.8 percent of the 2004 Senate races. The ratings also tracked the margin of victory, so a bigger perceived gap in competence went with a bigger win.

Nobody sensible claims a face decides an election. Plenty of other things move votes. But a snap judgement from a photograph, made by people who knew nothing about the candidates, carried real signal about a result with millions of dollars behind it.

What this does not mean

The research is good. The conclusions people stack on top of it often are not.

• It does not mean the judgements are correct. Speed is not accuracy. The studies measured agreement between fast and slow judgements of the same face, not whether either one matched the person.
• It does not mean you can be judged trustworthy by choosing the right jaw. The effects are about the image, lighting, angle and expression as much as bone structure.
• It does not mean a good photo beats a bad record. In hiring, a photo gets you a second of attention. What you did with your career decides the rest.
• It does not transfer cleanly across cultures. Most of this work used Western participants and Western faces. Treat it as strong evidence about a population, not a law of nature.

What to actually do with your photo

The research points somewhere specific, and it is not the usual advice.

Stop looking for the photo where you look best. Look for the photo that makes the impression the situation needs. Todorov and Porter showed those are different questions, and that people already intuitively sort photos by purpose when asked to.

Give yourself a real set to choose from. If impressions vary that much between images of one person, a single photo is a single roll of the dice. Several photos of the same face in the same session give you an actual choice.

Get the fast signals right, because they are the ones being read. In a sixth of a second nobody is reading your CV in your eyes. They are reading eye contact, whether the face is lit or shadowed, whether the expression is settled or caught mid movement, and whether the framing looks like a person or a passport control camera.

Let someone else pick. People are poor judges of their own photographs, partly for the reason covered in our piece on why you look better in the mirror than in photos. Ask two people which of your shortlist looks most like you on a good day.

What this means if you use an AI generator

We build one, so treat this section as interested rather than neutral. It is still the honest reading of the research.

The strongest practical finding above is that you need a set of images to choose from, because a single photograph is one draw from a distribution that varies more than you would expect. That is the thing a generator is genuinely good at. It is not good at deciding which of them is right, and no tool can do that for you, because the right one depends on the job you are doing with it.

Two warnings that follow from the same research.

The first: an image that no longer looks like you is not a better first impression, it is a first impression of somebody else. The gap between your photographs is already large. A tool that smooths, slims and relights until the person in the frame is a stranger has not improved your odds, it has moved the target. When you meet the person who saw the photo, the photo is the promise you are being measured against.

The second: the usual instinct is to pick the most flattering image. The 2014 work suggests sorting by purpose instead. The photo that works on a dating profile was not chosen the same way as the one that works on a campaign poster, and people do this naturally once the purpose is named. So name it. Ask which photo makes you look like someone who can do the job, not which photo makes you look best.

If you want the practical version of this, our notes on what makes a good input selfie and lighting for the photo you feed in cover the parts you control before any generation happens.

By the numbers

The bottom line

The fast judgement is real, it is quick, and it is mostly settled before anyone has consciously decided to look at you.

But the finding worth acting on is the quieter one. Your photographs disagree about you. The gap between your best and worst picture is bigger than the gap between you and a stranger. That is not a reason to worry about your face. It is a reason to stop uploading the first acceptable photo you find, and to choose deliberately from more than one.

If you want a set to choose from, our free AI headshot generator makes three from one selfie in about a minute, and our guide to picking the best photo to feed it covers what makes an input photo work.

Cite this article

Judgements of a face after 100 milliseconds match judgements made with unlimited time, and impressions vary as much between two photographs of one person as between photographs of two different people.

Source: FreeHeadshot.org, "Your Face Gets Judged in 100 Milliseconds", September 2026. https://freeheadshot.org/blog/the-100-millisecond-first-impression

FAQ

How fast do people judge your face?

Agreement with unhurried judgements is already above chance at 33 milliseconds and reaches its ceiling at about 167 milliseconds. The best known figure is 100 milliseconds, from Willis and Todorov in 2006, at which point fast judgements match the ones people make with unlimited time.

Does looking longer change the judgement?

No. In the original study, longer exposure did not change what people concluded about a face. It increased how confident they were in the conclusion they had already reached.

Which traits are judged that fast?

The 2006 study tested attractiveness, likeability, trustworthiness, competence and aggressiveness. All five showed the same pattern. Trustworthiness settled fastest.

Are these snap judgements accurate?

The research does not say that. It measured whether fast judgements agree with slow judgements of the same photograph, which is a different question from whether either judgement is true of the person. Treat them as predictions of how you will be read, not as facts about who you are.

Does my photo really matter more than my face?

For first impressions, the choice of photograph carries at least as much weight. Todorov and Porter found that impressions varied as much between different images of one person as between images of different people, which means the picture you pick is doing real work.

Can a better headshot actually help me get hired?

It can get your profile a longer look, which is what a first impression buys. It will not carry an application on its own. We covered what the evidence supports in do AI headshots help you get hired.

Does any of this apply outside Western countries?

Carefully. Most of this work was done with Western participants rating Western faces. The speed of judgement looks general, but which faces read as competent or trustworthy is shaped by who is doing the looking.