fg-zettelkasten

RIP Twitter API: A eulogy to its vast research contributions

3 min read

RIP Twitter API: A eulogy to its vast research contributions

Summary

This paper offers a scientometric “eulogy” for Twitter’s research APIs, quantifying the scholarly footprint they enabled before X effectively closed access in 2023 by pricing the Enterprise API at $42,000/month. Through a systematic search of eight databases spanning 2006–2024, the authors identify 33,306 studies in 8,914 venues accumulating over 610,000 citations across 16 disciplines. They document a “Golden Age” of roughly 25% annual growth in Twitter-based scholarship through 2022, followed by stagnation in 2023 and a 13% decline in 2024 — empirical evidence that platform data restrictions are reshaping what social science can know about online discourse. The piece doubles as a policy argument for legislative interventions (DSA, PATA) to guarantee researcher access.

Key Contributions

  • The most comprehensive cross-disciplinary bibliometric accounting of Twitter-data research to date, surpassing prior reviews (Karami et al. 2020; Yu & Munoz-Justicia 2020) in database coverage and scope.
  • First systematic empirical documentation of the post-API-closure decline in scholarly output, moving the debate beyond anecdote.
  • A dual classification scheme (venue-based vs. top-cited-study-based) that surfaces how disciplinary labels themselves shape bibliometric portraits of a field.
  • Transparent, replicable pipeline (code + data on GitHub) including a corrected citation extraction that fixes an error in the preprint version.
  • Empirical grounding for policy arguments framing API access as a structural determinant of public knowledge.

Methods

A systematic search across eight databases (LISTA, Web of Science, Global Health, ACM DL, IEEE Xplore, Engineering Village) using a tailored Boolean string (twitter NEAR/3 data OR api OR dataset NOT survey) restricted to articles, conference papers, dissertations, and preprints. Engineering Village’s full 43,354-record set was retrieved via Elsevier’s API to bypass a 1,000-result UI cap. Hand-coded relevance sampling (≥50 papers/database) yielded relevance rates of 80–97%. Records were deduplicated in R by DOI/title/abstract, and citation counts pulled from Crossref. Disciplines were assigned twice — to the top 100 venues and to the top 100 most-cited studies — followed by manual thematic analysis of top-cited papers per discipline.

Findings

  • 33,306 unique studies in 8,914 venues with 610,738 Crossref citations across 16 disciplines (2006–2024).
  • Median 25% annual growth from 2006 to 2022; +0.6% in 2023; –13% in 2024.
  • Earliest identified paper is Java et al. (2007), “Why we twitter,” despite APIs being available from 2006.
  • Venue-based disciplinary breakdown: Science/Engineering 27%, Information Science 14%, Computer Science 11%, Data Science 10%, Social Media 9%, Medicine 8%, Social Science & Internet Technology 7% each.
  • Study-based breakdown of top-cited papers: Social Science 24%, Data Science 19%, Social Media Studies 14%, with the rest spread across Public Health, CS, AI, Business, Emergency Response, etc.
  • Dominant themes: information dissemination, information integrity (disinformation, hate speech), big-data methods (sentiment analysis, topic modeling, geolocation), event/crisis detection, and behavior research (politics, marketing, mental health).
  • External reports indicate >100 research projects were directly disrupted in 2023, including the shutdown of Botometer.
  • Continued decline is likely, since much 2024 output still drew on pre-shutdown data.

Connections

This paper sits at the empirical center of the “post-API age” debate and pairs naturally with Freelon2024-sc and Bruns2025-fz on the political economy of platform data access, as well as Tornberg2025-ir and Davis-style critiques of platform power. Its policy framing connects directly to work on the DSA and Article 40 access regimes — see Rieder2025-ju, Rieder2026-pp, and Ohme2026-nv — while its bibliometric strategy complements methodological reflections in Bak-Coleman2025-pm, Bastos2025-ol, and Munger2025-cz on the fragility of computational social science infrastructures. Substantively adjacent are studies that themselves depended on Twitter APIs and now face replication crises, including Pierri2025-hm, Luceri2025-tr, and Minici2024-tf.

Graph View

Backlinks