Why weird-looking AI influencers work

What I saw

On the night of 5 October 2026, three videos came up back to back on Reels. All three were the same kind: a cartoon-like face in a realistic, luxurious scene.

Then I looked at the whole trend: its pioneer Jean Phil, the copies, the imitators. Screenshots and charts are below, in the raw layers.

I talked about it on video

I also told this story in an 8-minute video: Bu garip yapay zeka influencer’larını neden izliyoruz? (6 October 2026, 21:00, in Turkish). Six short clips came out of it and will land on the channel over the next few days.

The most striking number: shares beat likes

In a normal reel, shares stay far below likes. Here maham.iha’s shares are about 1.8 times the likes, and sirdonjeesus’s are 2 times. People don’t like these videos, they send them to someone.

For comparison, Higgsfield’s own ad reel: the same kind of character, but shares are a quarter of likes. Once it’s obviously an ad, nobody sends it to a friend.

Less than 1% of the people who watched maham.iha’s 20-million-view video liked it. The video grew by being sent, not by being liked.

Why it’s funny

1. Breaking the pattern. Luxury car, hotel entrance, suit. The brain expects a familiar template. When a bowl-cut, long-nosed man steps out, the expectation breaks in half a second and the scrolling stops.
2. Harmless violation. Psychology calls this a benign violation: a rule is broken but nobody gets hurt. That’s exactly where the laugh comes from.
3. Cartoonish, not creepy. Faces that are too close to real are unsettling. These are deliberately exaggerated. Comments keep comparing them to Wallace & Gromit and Lord Farquaad. On Jean Phil’s biggest video, the most-liked comment is a Wallace & Gromit frame with 224K likes.
4. Seriousness. The character never smiles. The viewer completes the joke and then wants to show that joke to someone else.
5. Status parody. What people are really laughing at is influencer culture itself. It mocks the “watch my rich life” videos. The viewer feels above the show.

Why it gets sent

1. High arousal. According to Berger and Milkman’s study of 7,000 New York Times articles, emotions that get people moving increase sharing, sadness reduces it. “What is this” is exactly that kind of emotion.
2. What you send isn’t the video, it’s “I thought of you”. Sending memes to friends now has a name: pebbling. It comes from penguins bringing pebbles to the ones they love. A weird character video is perfect for it. It needs no explanation, no language, offends nobody and gets a sure reaction.
3. The sender wins too. The “I found this first” feeling. What you share says something about you as well.
4. The platform rewards it. Instagram head Adam Mosseri said openly that one of the most important ranking signals is “how many of the people who saw it sent it by DM”. So this isn’t hacking the algorithm, it’s giving the algorithm exactly what it wants.

But it burns out fast

Jean Phil’s 10 videos got 14.9 million, 9.1 million and 36.6 million views in a row, then dropped fast. The last two are around 500K.

• 18-20 September: ~64.7 million
• 21-24 September: ~12.9 million
• 27 September-2 October: ~2.2 million

Once the joke is understood, the surprise is gone. When the surprise is gone, the “I found it first” feeling goes too, and the sending stops.

While it fell, the copy maham.iha, opened 11 days later, went from 5 million to 20 million. 170K followers in 8 days. Attention migrates from character to character.

The copies ride the same wave in less time. sirdonjeesus started at 9.7 million and fell to 367K within two days. The video after maham.iha’s 20-million hit stopped at 5.8 million.

Are they real people?

maham.iha and sirdonjeesus are not real people, and I can prove it. The biggest video of each is the same dance studio video. Same cones, same mirror, same camera move, even the same length: 10.12 seconds. The child dancer stays as is in one version and is turned into a character in the other. The method: take a real video someone else shot and turn the person in it into another character with AI. The movement is real, so it looks natural. The character doesn’t exist.

I couldn’t confirm Jean Phil. A real man in a wig and suit showed up on a British Kick streamer’s stream, and clips spread saying he “proved he’s real”. But that man is most likely an impersonator. There is an account selling Cameos as Jean Phil, using the original’s profile photo, opened two weeks after the viral moment, with 598 followers, and one of its videos references that streamer. Who is behind the original account is unknown. News outlets say AI, but there’s no hard technical proof. The most likely explanation is the same method as the others: real footage with an AI-altered look on top.

The uncertainty itself is fuel. The “real or AI” question fills the comments, produces reaction videos and makes money for imitators.

Why the copies are so fast

• A real motion video and an AI tool are enough. No need to film the character from scratch.
• sirdonjeesus posted about 24 videos in 3 days. Most got 100-800K views, two got 3.7 and 9.7 million.
• The “jean phil” search is almost entirely copies, reactions and “make money with AI influencers, message me on Discord” videos.
• People chase the genre, not the character. A new copy gets into the attention channel the original opened, for free.
• There’s a money incentive. Jean Phil’s bio originally had a meme coin link, and on 25 September its market cap showed 8.8 million dollars. That link is gone today. His rival Archibald Brown got a coin too.

What I take from it

• Content used to need a face, a camera, a location. Now an idea and a character are enough. They even take the movement from someone else’s video.
• Look at a piece of content’s share rate, not its likes. Content designed around “will someone send this” behaves completely differently.
• This engine burns out very fast. What’s built on surprise ends when the surprise ends. What lasts isn’t the character, it’s the hand that can produce this pattern again and again.
• What entertains the viewer and what makes the money are not the same thing.

What really gets me thinking

Why do we watch these? My screen time goes up to 4 hours too, 1-1.5 hours of it social media. We consume everything by scrolling, and the level of absurdity goes up a little every day just to stop us. Is the rising low quality a result of this habit, or is supply and demand broken?

Are we heading to a dead internet? In my review of Umut, I concluded he was loved because he was himself and didn’t play a role. If AI can do this and people aren’t looking for sincerity, that argument gets shaky. We may be heading to an internet where only AIs talk to and engage with each other.

Are real engagement and value disappearing, or are they just changing shape? I don’t know yet. I’m watching.

Open questions

• Can maham.iha keep its followers a month from now? I’ll check again in a month.
• Who is behind Jean Phil?
• Would a local version of this format work in Turkey?
• These trends are new, with no earlier example. What will the next wave of absurdity look like?

Sources

• Dexerto, “Who is Jean Phil?” (25 Sep 2026): https://www.dexerto.com/tiktok/who-is-jean-phil-ai-character-behind-multimillion-dollar-meme-coin-sparks-fake-persona-trend-3412528/
• Search Engine Journal, Mosseri and sends per reach: https://www.searchenginejournal.com/instagram-algorithm-shift-why-sends-matter-more-than-ever/521389/
• McGraw & Warren (2010), Benign Violations: https://leeds-faculty.colorado.edu/mcgrawp/pdf/mcgraw.warren.2010.pdf
• Berger & Milkman (2012), What Makes Online Content Viral?: https://faculty.wharton.upenn.edu/wp-content/uploads/2011/11/Virality.pdf
• Yahoo, pebbling: https://www.yahoo.com/lifestyle/always-sending-memes-to-your-loved-ones-its-called-pebbling-heres-why-experts-say-the-trend-has-its-pros-and-cons-100021620.html
• Know Your Meme, Jean Philanthrope: https://knowyourmeme.com/memes/jean-philanthrope-jean-phil
• LADbible, Jean Phil and his rival (30 Sep 2026): https://ladbible.com/news/ai-generated-social-media-videos-jean-phil-rival-433454-20260930
• TechCrunch, Higgsfield at a 1.3 billion dollar valuation (15 Jan 2026): https://techcrunch.com/2026/01/15/ai-video-startup-higgsfield-founded-by-ex-snap-exec-lands-1-3b-valuation
• Data: Instagram profile and reel pages, morning of 6 October 2026