Social media bots in the literature, 2016–2026

Identification & screening funnel

Two identification streams, unioned by OpenAlex work ID: (1) four keyword-search query variants (title/abstract × singular/plural), and (2) a forward/backward citation snowball from 7 seed papers — "The rise of social bots" (2016), "Botometer 101: social bot practicum for computational social scientists" (2022), "BotOrNot" (2016), "A global comparison of social media bot and human characteristics" (2025), "Detection of Bots in Social Media: A Systematic Review" (2020), "Unpacking the Social Media Bot: A Typology to Guide Research and Policy" (2018), "The spread of low-credibility content by social bots" (2018) — via OpenAlex reference lists (backward) and the cites: filter (forward). Each record is then screened for whether it substantively concerns social media bots.

1,049 records had no retrievable abstract (excluded automatically); the remaining 4,502 were screened title+abstract against one criterion: is this substantively about social media bots, not a false-positive keyword match? The keyword-search stream (2,783 unique records) was screened by an LLM reading each abstract; the forward/backward snowball stream (2,768 records not already in the keyword-search corpus) was screened by a deterministic keyword-rule classifier over the same criterion, documented in forward_backward_review/scripts/annotate.py. Single database (OpenAlex) and title+abstract only — a first-pass reproducible scan, not a peer-reviewed systematic review.

Keyword queries: “social media bot”, “social media bots”, “social bot”, “social bots”  ·  fields: title, abstract  ·  type: article
Snowball seeds (forward + backward, depth 1): "The rise of social bots" (2016), "Botometer 101: social bot practicum for computational social scientists" (2022), "BotOrNot" (2016), "A global comparison of social media bot and human characteristics" (2025), "Detection of Bots in Social Media: A Systematic Review" (2020), "Unpacking the Social Media Bot: A Typology to Guide Research and Policy" (2018), "The spread of low-credibility content by social bots" (2018)  ·  source: OpenAlex referenced_works + cites: filter

Bot types, governance, and time

Among the 3,233 included papers, 948 mention at least one governance/policy idea. Bot personas are harmonized to Ng & Carley's Dual Personas of Social Media Bots taxonomy (arXiv:2504.12498) — all 15 personas plus one added category, LLM Agent, for bots the paper explicitly describes as LLM/generative-AI-driven, a post-2023 wave the original 15-persona survey predates. Every one of the 16 personas gets its own bar segment below (no "Other" bucket) — a paper describing no specific behavior is tagged Unclassified rather than folded away. Governance ideas are harmonized to a 14-category typology built from the raw phrasing (see the dataset's GOVERNANCE_TYPOLOGY.md); those still fold low-frequency categories into "Other" per chart, but stay visible and filterable in the browse table below. 2026 is a partial year. Click a bar segment or legend entry to see the papers behind it.

Browse the annotated dataset

All 22,165 records. Click a row to expand its full annotation (raw bot vocabulary, raw governance ideas, harmonized categories). Search matches title and venue.