Humans wrote tweets for popular Twitter spambot @Horse_ebooks for two years as conceptual art
@Horse_ebooks is a human performance — @Horse_ebooks, the Twitter account that was until now largely thought to be a spambot, is in fact controlled by real people.
Context & Ripple Effects
For two years, @Horse_ebooks sat at the center of a small debate about machine creativity: Technology Review argued in January 2012 that Twitter bots create surprising new social connections, while Poynter had already documented in 2011 how the New York Times retired its half-automated 'cyborg' account in favor of human editors (the NYT 'cyborg' replacement). @Horse_ebooks was held up as the purest case of the former — a spambot whose glitchy output people found genuinely moving.
The Verge's reveal flips that case study: the account's most-loved fragments were written by people as a conceptual art project, meaning the audience's attachment to 'machine' language was actually an attachment to curated human writing. The timing sharpens the stakes — Twitter confirmed its IPO plans and confidential S-1 filing on September 15, so a story about unverifiable accounts broke days into a process where the quality of Twitter's user base becomes a question investors and advertisers will ask.
First-order effects
- The account's large following must re-read two years of output as authored performance rather than accidental machine poetry, and the story drew unusually broad pickup within a day — the New Yorker, NYT Bits, Gawker, the Atlantic, Business Insider, ABC News and others — because it punctured a widely shared belief about what the account proved.
Second-order effects
- Every other celebrated 'bot' account now faces the same credibility test: without disclosure norms, an audience cannot tell machine generation from human ventriloquism, which pressures creators of similar accounts to state what they actually are.
- For Twitter, the episode lands awkwardly alongside the MagicRecs recommendation rollout confirmed the same week — the company itself is curating what users see through algorithms, while its most famous 'algorithmic' voice turns out to be human, complicating the automation narrative on both sides just as the S-1 process makes engagement quality a market question.
Third-order effects
- If human-authored accounts routinely pass as bots and vice versa, platforms head toward treating account provenance as a verification problem rather than an assumption — the same tension between automated and authentic presence that advertisers pricing against real human attention will eventually force platforms to resolve.
The trend: The line between bot and human authorship online is dissolving into deliberate performance, pushing account authenticity from an assumption platforms make toward something they will have to verify and disclose.