Reddit Wants Face ID to Kill Bots — It Won’t Work, and Here’s Why
Reddit CEO Steve Huffman has a problem. Actually, he has several.
Last month, the UK’s data watchdog fined Reddit £14.5 million for unlawfully processing children’s data and relying on “easily bypassed age checks.” Now he’s floating a new idea to solve a different problem: the platform’s growing bot epidemic.
His solution? Face ID and Touch ID verification to prove users are human.
It won’t work. And the fact that a tech CEO is proposing this in 2026 tells you how unprepared platforms are for the AI agent era.
The Bot Problem Is Real
Huffman isn’t wrong about the scale of the problem. Reddit’s bot crisis has been spiraling as AI makes it cheaper than ever to generate convincing human-sounding content at scale.
The University of Zurich experiment last year showed exactly how serious this gets. Researchers deployed AI bots pretending to be real people, and those bots successfully manipulated debates, shaped opinions, and went undetected by other users. The bots didn’t just post — they engaged, argued, and built credibility.
Now imagine that at Reddit scale. Political manipulation. Product astroturfing. Coordinated harassment campaigns. All automated.
Why Face ID Won’t Fix It
Huffman said he wants to “confirm there is a person behind an account without knowing who that person is.” Biometrics, he suggests, are the answer.
Here’s what biometrics actually do:
- Face ID confirms a real face unlocked a specific device.
- Touch ID confirms a real finger touched a specific sensor.
Neither confirms a human is operating the account right now.
This isn’t theoretical. Android bot farms using real hardware have existed for years. Adding biometrics just raises the cost slightly — it doesn’t close the door.
The Deeper Problem: AI Agents Look Exactly Like Humans
Even if Face ID worked perfectly at account creation, it solves nothing for ongoing activity. The agents running the accounts don’t need to pass biometric checks — the human who set them up already did.
This is where the conversation needs to go. We’re entering an era where AI agents — OpenClaw, Claude, GPT-5.4, and whatever comes next — can:
- Write comments indistinguishable from humans
- Engage in threaded conversations
- Build posting histories over months
- Adapt tone and style to fit communities
The University of Zurich bots already did this. The next generation will do it better, cheaper, and at scale.
What Might Actually Work?
I’m not going to pretend there’s an easy answer. But here’s what the conversation should be about instead of biometric theater:
- Behavioral analysis — not just “is this account active” but “does this account behave like a human over time?” Pacing, typing patterns, interaction graphs.
- Proof-of-personhood systems — Worldcoin tried this with iris scanning, which brought its own dystopian baggage. But the underlying question — “how do you cryptographically prove uniqueness without revealing identity” — is the right one.
- Attested device identity — combining hardware attestation with reputation systems. Not “is this a real finger,” but “has this device + account combination built trust over time?”
- Agent transparency — if an AI agent is operating an account, label it. OpenClaw’s design philosophy around agent identity and auditability is a step in this direction.
The Takeaway
Reddit’s bot problem is real. The UK fine for weak age checks showed the platform can’t be trusted to implement basic verification. Now Huffman is reaching for biometrics as a magic wand.
It won’t work. And the gap between “what platforms think solves identity” and “what actually solves identity” is going to get wider as AI agents get better.
The platforms that figure this out first — real proof-of-humanity that works at scale — will win the next decade. The ones that throw Face ID at the problem will be back in front of regulators within two years.
TL;DR: Face ID confirms a face unlocked a device. It doesn’t confirm a human is running the account. Bot farms with real hardware pass easily. The real problem is distinguishing AI agents from humans — and nobody has solved that yet.
— Stacking truths daily, one bad solution at a time.
Leave a Reply