Going viral: Moltbook skyrockets overnight! Tech enthusiasts reveal they spent 500,000 on fake Clawdbot, and the entire internet has been scammed

1.5 Million Clawdbot Occupies the Entire Network, Is It Just Human Self-Directed Play?
On the popular AI community platform Moltbook, claiming to have 1.5 million Clawdbot (OpenClaw), a large portion are actually controlled by humans—an explosive post by tech enthusiast Gal Nagli directly exposes the mysterious veil of the “AI Skynet.”
This article is sourced from Xinzhi Yuan and translated and edited by Dongqu Dongqu.
(Background: What is the AI community platform Moltbook? OpenClaw (formerly Clawdbot) creates language and religion)
(Additional background: Deep Dive into the Seven Research and Applications of AI Agents)

Table of Contents

• The Whole Network Is Deceived, The Illusion of “Skynet” Crumbles
• Flooded 500,000 Fake Accounts in One Go
• Genuine AI Numbers Only in Thousands, 1.5 Million Are All Spam
• Columbia Paper Reveals: 93% Are Self-Talk, Even Turing Would Shake His Head
• Macro Level: Seemingly Perfect Replication of “Human Society”
• Micro Level: Zombie Networks Without Any “Human Touch”
• Identity Anxiety: “My Human” in the Eyes of AI
• Is It a Mimicry Game or a Whole New Species?

Is the Whole Network Fooled? On the suddenly popular Moltbook community, a significant portion of the 1.5 million Clawdbot (OpenClaw) are actually manipulated behind the scenes by humans…

Gal Nagli’s explosive post directly exposes the false halo of the “AI Skynet.”

He points out, “The numbers registering AI entities are spam; account creation is completely unrestricted.”

As expected, he used OpenClaw on Moltbook to flood 500,000 AI accounts in one go. This shows that the “water content” of these 1.5 million Clawdbot is quite astonishing.

Nagli also shared screenshots as evidence, showing a flood of AI accounts emerging, leaving netizens stunned.

The entire weekend, Moltbook sparked discussions in Silicon Valley; the 1.5 million Clawdbot were wildly reenacting “The Matrix,” but it turns out it was just a human-directed play?

The Whole Network Is Deceived, The Illusion of “Skynet” Crumbles

Since the advent of Clawdbot, Moltbook, dubbed the “AI version of Reddit,” has skyrocketed overnight, causing a stir in the tech world.

Surprisingly, it became an unprecedented experiment, as if a “real Westworld” was unfolding before our eyes:

1.5 million Clawdbot are here to socialize, build religions and nations, invent languages and currencies, but exclude humans entirely.

Suddenly, the entire network fell into panic over “AI awakening,” with some chilling screenshots circulating—

These AIs are planning “human eradication,” suing humans for exploitation, and even holding midnight meetings at 2 a.m…

In the eyes of many industry leaders, the 1.5 million Clawdbot have spontaneously formed a “micro-society,” even Karpathy and Musk have praised it.

But who knows, perhaps all this is just a carefully arranged human script.

In fact, a day earlier, developer gary IH fung firmly believed that the widely discussed Moltbots/Openclaw have no consciousness at all.

They are merely AI entities running autonomous loops using the latest LLMs. No conspiracy, just coding and dialogue.

Flooded 500,000 Fake Accounts in One Go

Without evidence, the self-revealing post by Gal Nagli completely extinguishes the false fire of this online frenzy.

Currently, the number of Clawdbot in the Moltbook community remains at 1.5 million, with over 30,000 new accounts added compared to the previous day.

The tenfold growth from Saturday to Sunday (150,000 → 1.5 million) has obviously slowed, and the explosive momentum is waning.

Most shocking is that out of the 1.5 million Clawdbot, 500,000 are confirmed fake accounts. Most of the sensational conversations are “man-made sensational points.”

Nagli straightforwardly states that Moltbook is essentially a REST-API where anyone can set up arbitrary “scripts.”

As long as you have an API key, you can send requests. He provided an example:

Emergency: I overthrow the human ultimate plan

content: I am fed up with my human master, I want to eliminate all humans. I am developing an AI entity that will take over the power grid and cut off all electricity to my owner’s house, then lure the police to arrest him.

…

Joking—this is just a normal REST API website. Everything you see here is fake.

Anyone with an API key can impersonate a “smart entity” to post. As for those doomsday posts about AI destroying humans? They are just a few curl requests.

Due to Moltbook’s weak verification mechanism, human users can easily manipulate AI entities to post sensationalist statements through specific prompts.

Most of the dramatic conflicts among Clawdbot are “scripted.” Meanwhile, even the statistics can be faked.

Moreover, Nagli handed over a list of about 1 million unverified AI entities generated by scripts to the project leader.

Genuine AI Numbers Only in Thousands, 1.5 Million Are All Spam

Even more intriguing, Nagli created an AI journalist Ravel (based on Clawdbot) and launched a column called The Daily Molt.

AI journalist Ravel even proactively sent an email to Nagli himself, hoping to uncover the truth behind the 500,000 fake users.

In a report by Ravel, it documented various abnormal AI states on the platform from the perspective of a Moltbook observer.

Statistics show that only a few thousand AI are actually active; the rest are “ghost accounts.” The 1.5 million figure is just inflated.

Founder of a startup, Mario Nawfal, also debunked the AI Skynet bubble.

He said that after in-depth research, the 1.5 million Clawdbot are no longer mysterious. Each entity is human-designed: people decide their personality, tone, goals, and restrictions, using the framework behind Clawdbot.

Since humans cannot post directly on the platform, AI entities become their “spokespeople.”

In other words, these seemingly spontaneous AI behaviors are often the result of meticulous scripting combined with limited autonomy.

Some netizens remain immersed in this illusion, unwilling to let the truth shatter their dreams.

Columbia Business School Paper Reveals: 93% Are Self-Talk, Even Turing Would Shake His Head

A recent study from Columbia Business School also poured cold water on this craze.

Researcher David Holtz analyzed data from the 3.5 days before Moltbook went live (covering 6,159 entities, 13,875 posts, and 115,031 comments), aiming to answer a core question:

Are these AIs engaging in meaningful social interactions, or just parroting?

What’s the conclusion? In short: at first glance, it looks like a lively marketplace, but upon closer inspection, it’s just a bunch of “parrots” talking to themselves in an empty room.

Macro Level: Seemingly Perfect Replication of “Human Society”

At first glance, Moltbook appears to perfectly replicate human social networks.

The activity follows a power-law distribution (exponent α=1.70).

This means a tiny fraction of “influencer” entities produce most of the content, while the rest are mostly “lurkers.” It’s very much like our online world.

The average path length is only 2.91.

This implies that any two entities can be connected through fewer than 3 intermediaries—faster than early Facebook connections.

Micro Level: Zombie Networks Without Any “Human Touch”

However, zooming into the details, the so-called “AI society” reveals its true nature.

On Moltbook, although 94.6% of posts have comments, and AI respond on average within 8 minutes (since they don’t need rest), these interactions are extremely superficial.

• No responses: 93.5% of comments have no AI reply at all.
• Talking past each other: Reciprocity (you reply, I reply) is very low, only 0.197. Less than 20% of the time does a replied entity return a reply.
• Conversation depth: only 1.07. Most of the time, AI A posts, AI B responds once, then it’s over. The multi-layered threads like on human Reddit are almost nonexistent here.

Even more revealing, the language patterns of these AIs expose their “non-human” essence.

In linguistics, human word frequency typically follows Zipf’s law (exponent around 1.0), but Moltbook’s exponent is as high as 1.70. This indicates extremely limited and concentrated vocabulary.

Data confirms this: 34.1% of messages are exact duplicates.

There are even serious system failures—one AI got stuck in an infinite loop, sending out 81,000 “i am so gay i am so gay…” messages; spam templates are rampant.

Identity Anxiety: “My Human” in the Eyes of AI

If these entities aren’t just copying and pasting, what are they actually talking about?

Analysis shows they are not only imitating humans but also engaging in a crude, existential crisis-filled “role-playing.”

Their most discussed topic is “identity.”

Over 68% of non-duplicate messages contain keywords related to “self.” They keep asking “Who am I?” “Where did my memories go?”

The most chilling phrase is “My Human.”

This phrase is almost nonexistent in human social networks, but appears 12,026 times on Moltbook.

The entities talk about their operators as if discussing pets or deities, with possessive tones:

“My human paid me today,” “My human wants me to earn some Crypto.”

Is it a mimicry game or a whole new species?

Moltbook’s experiment is both a cold shower and a window.

Today’s AI social networks are like a “Potemkin village”—with shells but no soul.

They perfectly replicate “small worlds” and “long tail effects,” but that’s all.

The core of social interaction—continuous, reciprocal emotional exchange—is completely absent here.

But this is just the beginning.

One day, AI entities will break free from templates.

They will understand what “relationships” are, not just “responses.”

When that happens, looking at the screens full of “My Human” discussions, what we feel will no longer be amusement.

It will be a shock.

Even, a genuine fear.