A look at different types of political bots on Twitter: many spread propaganda or seek to defame certain goups, but some fight disinformation or hunt imposters
Amanda Hess / New York Times :
Context & Ripple Effects
Amanda Hess's taxonomy in the New York Times lands a month after the 2016 election, when attention on Twitter's political bots was shifting from 'do they exist' to 'what do they actually do.' The piece splits them by function: propaganda spreaders and group-defamers on one side, but also bots that fight disinformation or hunt impostor accounts — a nuance that later research would quantify.
That research arc is already visible in the surrounding coverage: Oxford University researchers found bots pushed misinfo at higher rates in battleground states around Election Day, and a 14M-tweet analysis later found bots were just 6% of accounts but drove 34% of shares of low-credibility articles. The taxonomy matters because enforcement against all of these behaviors runs through the same platform.
First-order effects
- Twitter is handed a categorization problem: propaganda bots, defamation bots, counter-disinformation bots, and imposter hunters all look identical to crude automation filters, so policing one category risks sweeping up another.
- Users targeted by defaming bots — and researchers like DFR Lab, which documented its own bot-and-impersonator attack after covering ProPublica's (per Krebs on Security) — face harassment that follows, likes, and retweets rather than obvious spam.
Second-order effects
- Lax enforcement lets impostor accounts thrive on both Facebook and Twitter, leaving impersonated people few remedies while fake news and propaganda ride on borrowed identities (as the Times itself later reported) — pressure builds for identity-verification features neither platform had prioritized.
- Researchers gain a business case for bot-detection tooling: if 6% of accounts generate a third of low-credibility shares, newsrooms and platforms will pay to distinguish malicious automation from benign or defensive bots.
Third-order effects
- If the pattern holds, platform moderation shifts from content-level takedowns to behavior-level classification — treating automated amplification, coordinated defamation, and impersonation as distinct policy categories with different thresholds.
- The counter-bot ecosystem (disinformation fighters, imposter hunters) points toward an arms race where verification of 'who is real' becomes as central to social platforms as removing banned content.
The trend: Platform governance is moving from treating all automation as spam toward classifying bots by function — propaganda, defense, impersonation — because they demand different enforcement tools.