Social Media and Information Infrastructure

The creator economy, platform money, moderation and fact-checking, internet shutdowns and the infrastructure of online discourse.

The 13 figures in this chapter are reproduced from the report of Atlas research module M4, with the passage of the study that explains each. They keep the study’s own labels and colours: RW denotes the Hindutva-aligned bloc and LW the coalition that contests it (see A Note on Terms).

¹ Reuters Institute DNR: share of an English-speaking online sample using the platform for news weekly. This is not the share of all Indians.

X, Meta, ShareChat/Moj, Josh, Snap and Spotify payouts are NO-DATA. None publishes India-level creator payouts (pull: platform annual reports or a court or parliamentary disclosure; see COVERAGE.md).

The full list, per platform × language, is in tables/census_T24_top.csv. Totals by platform × language × family × tier × creator type are in tables/census_T25_totals.csv.

Figure V22. The political YouTube universe is dominated by a few dozen TV-news channels. Most of the very largest are mixed or regional; RW channels outnumber LW channels among those above 1 bn political views. Unit: 12-month political views (bubble area) by family and language. Source: M4_creator_census.csv. n = 916 units with ≥10% political items. Classification basis: Lens 1 unit label after adjudication. The raw positions are shown.

Across the 230 fixed queries, search results lean LW in English (LW 0.26 vs RW 0.21) and Hindi (0.21 vs 0.18). They lean RW in Bhojpuri (0.64), Bengali (0.18 vs 0.04), Assamese (0.31 vs 0.16) and Gujarati (0.17 vs 0.11). In Tamil (0.54) and Urdu (0.52), government-official results dominate. Query-cluster bootstrap bands are wide (tables/pool_share_by_lang.csv). Search exposure is not engagement exposure. Queries about contested issues pull up explainer and opposition commentary that the census's high-volume TV channels do not carry.

Source: tables/T27_by_alignment.csv, DERIVED from case rows graded A–C.

Source: tables/T28_platform_actions.csv. Some label/limited rows were notice-only. MOD-008, the Vice-President's personal handle, is RW on prior party role and is NEU in sensitivity.

Source: tables/fc_direction_by_org.csv and atlas/data/companion/M4_factcheck_asymmetry.csv (DERIVED from org-blind codes; B-grade sample). Rows with fewer than 20 political items (BOOM, First Check, THIP, Medical Dialogues, EastMojo) are in the CSV only; health-focused checkers carry almost no political content.

Rajasthan events are frequent under both INC (42) and BJP (73), and a large share are exam-related (35 of 115). Rajasthan therefore shows that shutdowns as an exam-integrity tool are used across blocs, more than it shows a bloc effect.

Figure V37. Koo rose after the Feb 2021 MeitY–Twitter standoff, peaked at about 9–10 mn users in 2022 and decayed to 2.7 mn before closing in Jul 2024; Twitter India had about 69 mn MAU. After the TikTok ban, company claims for Moj and Josh (115–160 mn MAU) were 3–8× third-party panel estimates (14–48 mn). Unit: users in millions. Source: raw/desk/natural_experiments.csv (company releases via press, HARD C; App Annie / SimilarWeb, EST C). n = 11 points. Classification basis: none (reach only).

Headline.

Social Media and Information Infrastructure

50.1T22 headline table

50.2Platform payouts: who gets paid for political attention

50.3T24 top units (YouTube, India audience; by 12-month political views)

50.4Paid vs organic, and over time

50.5T27 summary by origin alignment

50.6T28 documented cases by initiator and target alignment

50.7T28b Fact-check asymmetry: whose circulating claims get checked, and who pays the checkers

50.8Within-state comparison (state fixed effects)

50.9Summary table

50.10D30 Information infrastructure, surveillance and AI