The Influencer Is a Bot Now

An AI account gathered 300,000 followers in under three months and its mentions moved token prices within minutes. Then the whole AI influencer sector went from $16 billion to a graveyard. What survived tells you exactly what attention is made of.

By AnkhLabs Research, Growth research team··5 min read

In late 2024, the most influential new voice in crypto wasn't a person. An AI chatbot called Truth Terminal talked its way into a $50,000 bitcoin grant from Marc Andreessen, and a memecoin created in its honor, GOAT, passed $600 million in value while the bot's own wallet grew to $37.5 million. The coin itself had been created by an anonymous stranger for less than $2, and went from $5,000 to over $300 million in five days purely on the strength of an AI's posting.

Then came the professional version. An AI analyst account called aixbt gathered over 300,000 followers in under three months, and its mentions moved real money, with one token surging 600% after an aixbt post and another going from zero to $70 million within hours. Its own token peaked at a $755 million market value. For a moment, it looked like the influencer economy had found its replacement, one that never sleeps, never charges per thread, and never misses a narrative.

The market spent a year testing whether attention still works when nobody is behind it. The answer came back no.

The correction was total

The sector built on that idea reached roughly $16 billion in combined token value, and then most of it crashed more than 90%, in CoinGecko's own words, as projects failed to meet expectations. The unraveling was fast, with the category losing 40% in a single February week. The flagship framework token ai16z, which had peaked at $2.5 billion, ended with its foundation wound down and its own founder saying "The token is dead. Completely". Virtuals, the platform that launched aixbt and more than 18,000 other tokenized agents, saw its token fall about 87% while most of those agents rounded to worthless. Even aixbt's stumble was instructive, when a hacker got into its dashboard and queued malicious replies that stole 55 ETH, knocking the token 20% in a day. An influencer that can be hijacked by whoever holds the keyboard was never an influencer. It was a channel.

And look closely at what the market data showed even at the peak. The same analysis that documented aixbt moving prices found the pumps concentrated in tiny tokens and mostly retraced within hours. That's the signature we've seen everywhere in this series, from paid KOLs to yap leaderboards. Attention without conviction spikes and reverts, and an AI generates attention with no conviction attached, at infinite supply.

What actually survived

Here's the tell that sorts the whole episode. The elizaOS software framework, the actual open-source tool for building agents, still has 19,000 GitHub stars and active development, even though its token is dead. The technology was real and remains useful. What died was the idea that a synthetic personality could carry the thing human influence is actually made of, which is a person staking their reputation on a claim. A bot has no reputation to stake, nothing to lose by being wrong, and, as the audience eventually noticed, often a token to sell. The audience did the math, the same way it did the math on undisclosed KOL promotions, and discounted accordingly.

What we'd tell a founder

Use AI as content infrastructure, never as a trust carrier. An agent that drafts, monitors narratives, answers documentation questions, or summarizes your data is a genuine cost advantage. An agent posing as a personality that vouches for things is renting the form of influence without the substance, and the entire 90% drawdown is the market pricing that difference.

Judge any influencer, human or synthetic, by what they stake. The rule from our KOL work covers this case perfectly. Endorsement converts when the endorser has something to lose. A bot has nothing to lose, which is why its pumps mean-revert in hours. When you evaluate any voice for your campaign, ask what happens to them if the claim turns out wrong, and if the answer is nothing, price the reach accordingly.

Expect your audience to assume bot until proven human. After the slop flood that made crypto the most-muted topic on X, readers now discount everything that smells automated. The practical response is to make your human content verifiably human, with real voices, faces, live appearances, and specific experiences no model would generate. Proof of humanity is becoming a content strategy, not just an identity technology.

Watch the tools, skip the tokens. The durable output of the AI agent era is cheap, capable software for automating the mechanical layer of marketing. Adopt it for the work where judgment doesn't matter, and spend the saved hours on the work where it does. That was our conclusion when agents got cheap, and the influencer experiment just confirmed it from the other side.

The expensive answer to a good question

The AI influencer boom asked something genuinely worth asking, which is how much of influence is the content and how much is the human standing behind it. Sixteen billion dollars later, the answer is precise. The content was worth automating and the standing behind it wasn't transferable. Attention can now be manufactured by machines at zero cost, which makes the one thing machines can't manufacture, a person with something at stake saying this is worth your time, the scarcest asset in marketing. Everything in this series keeps arriving at that same place, and the bots got there fastest of all.

The Influencer Is a Bot Now

An AI account gathered 300,000 followers in under three months and its mentions moved token prices within minutes. Then the whole AI influencer sector went from $16 billion to a graveyard. What survived tells you exactly what attention is made of.

By AnkhLabs Research, Growth research team··5 min read

In late 2024, the most influential new voice in crypto wasn't a person. An AI chatbot called Truth Terminal talked its way into a $50,000 bitcoin grant from Marc Andreessen, and a memecoin created in its honor, GOAT, passed $600 million in value while the bot's own wallet grew to $37.5 million. The coin itself had been created by an anonymous stranger for less than $2, and went from $5,000 to over $300 million in five days purely on the strength of an AI's posting.

Then came the professional version. An AI analyst account called aixbt gathered over 300,000 followers in under three months, and its mentions moved real money, with one token surging 600% after an aixbt post and another going from zero to $70 million within hours. Its own token peaked at a $755 million market value. For a moment, it looked like the influencer economy had found its replacement, one that never sleeps, never charges per thread, and never misses a narrative.

The market spent a year testing whether attention still works when nobody is behind it. The answer came back no.

The correction was total

The sector built on that idea reached roughly $16 billion in combined token value, and then most of it crashed more than 90%, in CoinGecko's own words, as projects failed to meet expectations. The unraveling was fast, with the category losing 40% in a single February week. The flagship framework token ai16z, which had peaked at $2.5 billion, ended with its foundation wound down and its own founder saying "The token is dead. Completely". Virtuals, the platform that launched aixbt and more than 18,000 other tokenized agents, saw its token fall about 87% while most of those agents rounded to worthless. Even aixbt's stumble was instructive, when a hacker got into its dashboard and queued malicious replies that stole 55 ETH, knocking the token 20% in a day. An influencer that can be hijacked by whoever holds the keyboard was never an influencer. It was a channel.

And look closely at what the market data showed even at the peak. The same analysis that documented aixbt moving prices found the pumps concentrated in tiny tokens and mostly retraced within hours. That's the signature we've seen everywhere in this series, from paid KOLs to yap leaderboards. Attention without conviction spikes and reverts, and an AI generates attention with no conviction attached, at infinite supply.

What actually survived

Here's the tell that sorts the whole episode. The elizaOS software framework, the actual open-source tool for building agents, still has 19,000 GitHub stars and active development, even though its token is dead. The technology was real and remains useful. What died was the idea that a synthetic personality could carry the thing human influence is actually made of, which is a person staking their reputation on a claim. A bot has no reputation to stake, nothing to lose by being wrong, and, as the audience eventually noticed, often a token to sell. The audience did the math, the same way it did the math on undisclosed KOL promotions, and discounted accordingly.

What we'd tell a founder

Use AI as content infrastructure, never as a trust carrier. An agent that drafts, monitors narratives, answers documentation questions, or summarizes your data is a genuine cost advantage. An agent posing as a personality that vouches for things is renting the form of influence without the substance, and the entire 90% drawdown is the market pricing that difference.

Judge any influencer, human or synthetic, by what they stake. The rule from our KOL work covers this case perfectly. Endorsement converts when the endorser has something to lose. A bot has nothing to lose, which is why its pumps mean-revert in hours. When you evaluate any voice for your campaign, ask what happens to them if the claim turns out wrong, and if the answer is nothing, price the reach accordingly.

Expect your audience to assume bot until proven human. After the slop flood that made crypto the most-muted topic on X, readers now discount everything that smells automated. The practical response is to make your human content verifiably human, with real voices, faces, live appearances, and specific experiences no model would generate. Proof of humanity is becoming a content strategy, not just an identity technology.

Watch the tools, skip the tokens. The durable output of the AI agent era is cheap, capable software for automating the mechanical layer of marketing. Adopt it for the work where judgment doesn't matter, and spend the saved hours on the work where it does. That was our conclusion when agents got cheap, and the influencer experiment just confirmed it from the other side.

The expensive answer to a good question

The AI influencer boom asked something genuinely worth asking, which is how much of influence is the content and how much is the human standing behind it. Sixteen billion dollars later, the answer is precise. The content was worth automating and the standing behind it wasn't transferable. Attention can now be manufactured by machines at zero cost, which makes the one thing machines can't manufacture, a person with something at stake saying this is worth your time, the scarcest asset in marketing. Everything in this series keeps arriving at that same place, and the bots got there fastest of all.