Skip to content
Culture

Celebrities Who Look Alike, According to an AI

Which celebrities look alike? Ollie's face recognition model compared more than 5,000 celebrities with each other, over 17 million pairs, and ranked the closest. It paired Dakota Fanning with Sadie Sink, Winona Ryder with Rachael Leigh Cook, and Mila Kunis with Camila Mendes. Without being told who is related, it also picked out parents, children and siblings.

Liam BradleySeptember 26, 2026 · 6 min read

How We Compared Every Celebrity With Every Other

Ollie's index holds more than 5,000 living celebrities, each with between 2 and 12 verified photos. That makes more than 17 million possible pairs of people. For each pair, the model compares every photo of one person with every photo of the other and averages the two closest photo pairs. It's the same rule the celebrity lookalike finder uses when it scores you. A single lucky photo can't win on its own.

The model has never been told anyone's name. It turns each face into a list of 512 numbers, a facial embedding, and the similarity score measures how close two of those lists are. It compares faces. It doesn't judge how famous, attractive or related two people are.

For scale: two celebrities picked at random score about 33% on Ollie's scale, and only 1 pair in 100 reaches 58%. Every pair below scores in the range two different photos of the same person usually reach.

The Closest Celebrity Lookalikes

Among Hollywood actresses, the model's closest pairs were Dakota Fanning and Sadie Sink, Amanda Seyfried and Dakota Fanning, Winona Ryder and Rachael Leigh Cook, Mila Kunis and Camila Mendes, and Kate Bosworth and Danielle Panabaker. Carrie Underwood and Emily Osment made the list from music, and Dakota Johnson and Renate Reinsve crossed the Atlantic.

In India, Katrina Kaif and Zarine Khan came out as one of the closest pairs, a resemblance the Hindi film press pointed out when Zarine Khan debuted. So did Kareena Kapoor and Kritika Kamra, and Alia Bhatt and Nidhhi Agerwal. In sport, golfers Rory McIlroy and Matthew Fitzpatrick were close, and so were footballers Dayot Upamecano and Dávinson Sánchez.

The closest pair between a man and a woman was tennis players Alexander Zverev and Aryna Sabalenka. The model compares face shape, not gender, so strong jawlines and similar proportions can pair a man with a woman.

Read more: Why Everyone Has a Celebrity Doppelganger (According to Science)

It Found Families Without Being Told

The most striking result is how many of the closest pairs are related. The model has no family data, yet near the top of 17 million pairs it put Japan's Emperor Emeritus Akihito and his son Emperor Naruhito, and Naruhito and his brother Crown Prince Fumihito. Fumihito's children, Mako Komuro and Prince Hisahito, were paired as well.

It did the same for King Hamad of Bahrain and his son Crown Prince Salman, for the chess-playing siblings Praggnanandhaa and Vaishali Rameshbabu, and for actors Jonah Hill and his sister Beanie Feldstein. A model trained only to tell individuals apart still picks up family resemblance, because relatives share much of the bone structure the model relies on. Our post on identical twins covers the extreme case.

Where the AI Gets Fooled

Not every close pair is a real resemblance. The celebrities who turned up in the most close pairs were tennis players (Barbora Krejčíková, Elina Svitolina, Elena Rybakina) and Korean pop and drama stars. Tennis photos taken with long lenses, in similar light and mid-effort, look alike, and so do stage looks with the same styling and make-up. The photos have as much in common as the faces.

There is also a known weakness. A large share of the closest pairs are East Asian or South Asian celebrities. Ollie's model was trained on MS1MV2, a dataset made up mostly of lighter-skinned faces, so it separates some groups less finely and gives different people higher scores. People show a similar bias, known as the own-race effect. We disclose this on the match page and cover it in accuracy across demographics. Read these pairs as what one model sees, not as a verdict on anyone's face.

Read more: The Celebrity Faces That Fool AI the Most

Get new guides by email

By joining, you agree that Ollie can email you about new features, including when Search launches. Every email has an unsubscribe link. Privacy

The Most 'Average' Celebrity Faces

Averaging every celebrity's embedding gives the centre of the model's face space, the most typical face in the index. The celebrities closest to it included Robert Downey Jr., Isabella Rossellini, Margaret Qualley, Mikey Madison and Jeffrey Dean Morgan.

'Average' here is a compliment. In face research, faces close to the average of a population tend to be rated as more attractive, as our post on attractiveness research explains. A face near the centre also resembles many people a little, which is why some celebrities come up as a match more often than others.

Find Which Celebrity You Look Like

The same model and the same scoring rule are behind the match page. Upload a clear, front-facing photo and Ollie compares it with every celebrity in the index and returns your five closest. The photo is used for the search and then discarded. Like the pairs above, your result shows how close two faces are to this one model. It's a similarity score for comparing results, not a measure of accuracy.

Frequently Asked Questions

Which celebrities look alike the most?

Out of more than 17 million celebrity pairs, Ollie's AI rated Dakota Fanning and Sadie Sink, Winona Ryder and Rachael Leigh Cook, Mila Kunis and Camila Mendes, and Katrina Kaif and Zarine Khan among the closest. Family members such as Emperors Akihito and Naruhito scored just as high.

How does an AI decide two celebrities look alike?

It turns each face into 512 numbers (a facial embedding) and measures the distance between them. Ollie compares every photo of one person with every photo of the other and averages the two closest pairs.

Can AI tell if two celebrities are related?

Not directly, but relatives often score very high because they share bone structure. Without any family data, Ollie's model paired Akihito with Naruhito, Jonah Hill with Beanie Feldstein, and Praggnanandhaa with his sister Vaishali.

Why does the AI pair so many K-pop stars or tennis players?

Similar photo conditions and styling make photos look alike, and Ollie's training data under-represents some groups, so it separates them less finely. These pairs show what one model sees, not a fact about anyone's face.

Which celebrity do I look like?

Upload a clear, front-facing photo on Ollie's match page. It compares your face with more than 5,000 celebrities and returns your five closest matches for free.

Try it yourself

Find your celebrity lookalike

Upload a photo and see which celebrities you look most like. Free to try, and your photo is never stored.

Find my celebrity look alike

About the author

Liam BradleyWrites for Ollie, a free celebrity lookalike AI

The Ollie team wrote and trained Ollie's own face-recognition model from scratch, and writes these guides to explain how face matching works in plain English. Questions or corrections are welcome on the contact page, and the FAQ covers how Ollie works and what happens to your photo.

Related Articles