More
Video taking off 127x its usual first day

Julia Ebner's undercover years lead to a case for "bridging algorithms"

The Oxford extremism researcher tells TEDxBerlin that outrage-driven feeds are a choice, and that AI could be built to cool hostility instead.

Why this story? The video opened at a pace of 59,910 views a day, 127 times a normal first day on this channel (470).

Extremists are using AI to radicalize us—can we fight back? | Julia Ebner | TEDxBerlinExtremists are using AI to radicalize us—can we fight back? | Julia Ebner | TEDxBerlin

Things we can learn from the video

Press to jump

  1. 0:11Ebner describes years undercover in jihadist, neo-Nazi and incel groups to study the psychology of radicalization.
  2. 1:14Extremist groups act like startups, adopting social bots and encrypted messaging early.
  3. 3:22Moderating single posts fails, she argues, because recommendation algorithms favor outrage.
  4. 4:23Three shared factors: personal grievance, identity fused with an in-group, and a perceived existential threat.
  5. 7:42AI research flags early warning words: kinship terms, dehumanization, threat perception and justifications of violence.
  6. 8:58Her answer: "bridging algorithms" that link people with differing views and use AI to de-escalate.

"Throughout much of my twenties, I lived a double life," Julia Ebner says at the start of her TEDxBerlin talk. That double life was spent undercover inside jihadist, neo-Nazi and incel groups, studying how ordinary people get pulled in. Ebner leads the Violent Extremism Lab at the University of Oxford and is the author of The Rage, Going Dark and Going Mainstream. Here she turns those years toward one question: can the technology that speeds up radicalization also help slow it down?

TEDx Talks posts new talks almost daily from independent events around the world. The same day brought Göttingen, Reus, Mumbai and Adana. Most of those talks open with a few hundred views. Ebner's passed 40,000 in its first eleven hours, and its comment section had already outgrown a typical full first day.

"One of the things I learned early on was that extremists operate a bit like startups," she says. Her examples: early adoption of social bots and encrypted messaging, and her meeting in 2017 with leaders of Generation Identity, who staged stunts to feed their online campaigns. Deleting single posts misses the point, she argues, when recommendation systems reward outrage.

Across very different movements she sees the same three ingredients: a personal grievance, an identity fused with an in-group, and the sense that the group faces an existential threat. Generative chatbots, deepfakes and automated disinformation raise the stakes. "What shocked me was not just the content, but the scale and the speed," she says.

The answer the title promises comes in the last few minutes. Her team uses AI to spot early warning signs in language, such as kinship words, dehumanization and justifications for violence. She also proposes "bridging algorithms" that connect people with different views. "But algorithms are not a law of nature. They're a design choice."

Viewers have picked up that thread. The most liked comment asks whether social media algorithms need to change. Another doubts that the helpful kind of AI will ever be funded like the harmful kind. Ebner's own closing line for what AI could be: "A bridge, not a weapon."

Published on Trari

Where next?

Or follow us on Telegram

The facts
Channel
TEDx Talks
Subscribers
44,700,000
Channel views, last 7 days
12,122,323
Channel views, pace per day
2,102,426 (on a normal day: 1,814,649)
Video published
October 13, 2026
Video views so far
39,447 (less than a day online)
Video views, pace per day
59,910

The numbers

A new talk on this channel usually opens quietly. This one started at a very different pace.

the video's first day
usual first day here
470
this video
59,910
subscribers
44,700,000
no significant change in the last 7 days
total views
9,140,193,276
views per day
2,102,426
typical day: 1,814,649
the video
views
39,447
likes
228
comments
18
days online
1
subscribers: this channel and similar channels
TEDx Talks
44,700,000
TED
27,900,000
TED-Ed
23,000,000
CrashCourse
17,200,000
As of October 14, 2026 · Data source: Trari Social News, Trari Traffic Trends and Ranking Platform, Google YouTube
Subscriber, view, like and comment counts are YouTube's. Rates, usual days, rankings, comparisons and categories are Trari's own calculations.
Developed with YouTube

💡 Test your knowledge

Which of these books was written by Julia Ebner?