← Social media bots literature review (main)

Literature review · platform governance

How platforms, states, and courts have governed social media bots, 2016–2026

An annotated bibliography of 111 legal, policy, and regulatory-studies papers on bot governance — bot typologies, platform bot policies, and the shift from voluntary codes to statutes like the DSA and NetzDG — drawn from the governance_review/ systematic search (SCOPUS-substitute + Semantic Scholar + citation chaining, 2016–2026).

621 unique candidates after dedup · 111 passed keyword screening · 90 confirmed on-topic on manual review

Search & screening funnel

Two SCOPUS-substitute boolean queries (run against Semantic Scholar), four native phrase searches, and forward/backward citation chaining from three seed papers (Bodó 2022; Gorwa, The platform governance triangle). Full methodology, raw query logs, and PDFs are in the governance_review methodology.

621unique candidates, all sources
−510pre-2016 / off-topic by keyword
111passed keyword screen
90confirmed on-topic (this review)

The keyword screen behind the 111 was deliberately permissive (see methodology), so a second manual pass was run for this page: 21 of the 111 are false positives — things like veterinary antibiotic rules or rideshare economics that matched on stray terms ("bot", "regulat*", "platform") without being about social media bot governance. They stay in the browse table below, marked excluded, for audit purposes; the charts and narrative below use only the 90 on-topic papers.

Types of bots the governance literature actually argues about

The legal/policy literature doesn't use one bot taxonomy — it argues about bots in terms of the harm channel regulators are trying to close. Harmonizing the 90 papers' own vocabulary into seven recurring types:

Political / propaganda bots
Automated accounts amplifying political hashtags, candidates, or narratives — the original "computational propaganda" object of Howard & Kollanyi's Brexit study and most election-law scholarship since.
Coordinated inauthentic behavior (CIB) networks
Clusters of bot/sockpuppet/troll-farm accounts acting in concert (synchronized posting, shared URLs) — the unit Meta and academic detection papers now use instead of "bot," since coordination, not automation alone, is what platforms police.
Engagement-for-hire bots
Bots sold as a service to inflate likes/followers/views ("engagement as a service") — a commercial market that authenticity-governance research treats as a distinct enforcement target from political bots.
Platform / community-management bots
Bots platforms and volunteer moderators deploy — Reddit AutoModerator-style tools, third-party Discord bots — i.e., bots as a governance instrument, not just a governance target.
Conversational & companion AI
Chatbots and AI companions (Weibo's "Robot," character-based companion apps) that raise consent, intimacy, and youth-safety questions distinct from disinformation bots — the fastest-growing category in 2025–2026.
LLM-generated / generative-AI bots
Post-2023 bots whose text/persona is produced by a generative model rather than templated scripts — BotSim-style malicious botnets, and the "cyborg propaganda" hybrid of a verified human plus algorithmic amplification.
Fake / fraudulent accounts
Sock puppets, cyborg accounts, and compromised accounts framed as a security/fraud problem (return fraud, review manipulation) rather than a speech problem — governed through detection engineering more than law.

For a formal taxonomy rather than this governance-literature synthesis, see Ng & Carley's The Dual Personas of Social Media Bots (arXiv:2504.12498), which the companion social-bots review uses as its harmonization standard.

How the governance model changed, 2016–2026

Reading the 90 papers in publication order surfaces four rough eras. Years overlap heavily — older models don't disappear, they get layered under new ones — but the center of gravity moves. Each era below expands into the actual papers and the actual platform/government primary sources (with links) that the claim rests on.

Reading note — the four-era grouping is a first-pass synthesis from title/abstract-level classification of the 90 papers, not a coded/validated content analysis. The expandable citations above are real and directly checkable (click through); treat the era labels themselves as a scaffold for browsing, not a settled finding.

Charts: bot type, governance model, and actor, by year

Counts are the 90 on-topic papers only. A paper can be tagged with more than one bot type / governance category / actor, so bars sum to more than 90 per year. Click a bar segment or legend entry to filter the browse table below.

Bot types discussed, by year
Governance model invoked, by year
Governance actor addressed, by year

Browse the annotated bibliography

All 111 papers that passed the keyword screen. The 21 marked excluded are keyword false positives kept for audit (hover title for reason); the rest are the 90 used in the charts and narrative above. Click a row to expand its full tagging. Full text, where retrieved, is in governance_review/pdfs/; see the full annotated bibliography for citation-count context on the source review.

YearTitleBot typesGovernance actorCited by