BOOM Report: AI Fakes Shift From Slop To Political Disinformation In First Half Of 2026

AI-generated or AI-edited content featured in three of every ten BOOM fact-checks between January and July 2026, up from two in ten last year. While AI slop was a major driver in 2025, this year’s AI claims were overwhelmingly political, with more than eight in ten (133 of 162) falling in that category.

Between 2 January and 31 July 2026, BOOM published 527 original fact-checks in English (229), Hindi (228) and Bangla (70).

BOOM's fact-checks in the first seven months of 2026 were shaped by three events. The US-Israel-Iran war dominated March, the West Bengal Assembly election ran through April and May, and the Cockroach Janta Party (CJP) protests over the NEET paper leak built from late May and peaking in July.

AI-Related Fact-Checks: Three In Ten, Mostly Political

AI-led claims have seen a consistent and significant rise over the past two years. 162 of BOOM's 527 fact-checks (30.7%) between January and July 2026 involved AI-generated or AI-manipulated content, up from 20.5% in 2025 (219 of 1,067) and 8.35% in 2024.

By month, the AI share peaked in February (26 of 61, 42.6%) and March (34 of 86, 39.5%), then fell to 6 of 51 (11.8%) in May. 29 of May's 51 fact-checks were about the West Bengal election, and only 3 of those were AI.

What kind of AI

Synthetic AI video was the largest group, 42 of 162 (25.9%), followed by synthetic AI images (36, 22.2%), voice clones over real footage (35, 21.6%) and face-swap or lip-sync deepfakes (34, 21.0%), between January and July.

AI-edited real media accounted for 10 (6.2%), other AI such as AI text and AI-made documents for 4 (2.5%), and 1 (0.6%) was unclear. Deepfakes and voice clones together, which "put words in a real person's mouth", made up 69 of 162 (42.6%).

Where AI shows up

By theme, Religion had the highest AI share, 9 of 13 (69.2%), followed by Entertainment (11 of 23, 47.8%), International Politics (57 of 138, 41.3%) and Indian Politics (68 of 223, 30.5%). Communal claims were the outlier, at 1 of 82 (1.2%).

133 of the 162 AI fact-checks (82.1%) were coded political and 29 (17.9%) non-political. BOOM's 2025 report flagged 78 "AI slop" fact-checks within its 219 AI total; in 2026 so far, AI has been primarily to make political claims.

Instagram had the highest AI share by platform, 37 of 103 fact-checks (35.9%), against X at 107 of 368 (29.1%) and Facebook at 70 of 263 (26.6%).

Narendra Modi was the most frequent target of AI fact-checks (16), followed by Army chief Gen Upendra Dwivedi (6), Iran (5) and Shashi Tharoor (4). Only 8 fact-checks were both AI and old footage, with either older real footage given an AI voice or face, or AI videos re-shared months later.

Theme Assessment: What Drove The Fact-Checks

Indian Politics (223, 42.3%) and International Politics (138, 26.2%) made up two-thirds of the 527 fact-checks, with Communal third at 82 (15.6%). 372 of 527 (70.6%) fell into a named event cluster; the other 155 (29.4%) were one-off claims.

US-Israel-Iran war: AI arrives on day one

The war produced 81 fact-checks (15.4%), from the first on 27 February to a tail in June and July. 67 came in March, 78% of that month's 86. The peak was the week of 9–15 March, with 20 fact-checks, 14 of them AI.

Fake and recycled content were about equally common in the war claims, with AI in 38 of 81 fact-checks (46.9%) and old footage in 42 (51.9%). The AI mix was led by synthetic video (17), then deepfakes (8), voice clones (5) and AI images (4). The first war-related fact-check, on 27 February, was already an AI fake.

Mainstream outlets carried 10 of the 81 war claims, and 16 reached Category 4 or higher on the Breakout Scale.

Examples include an AI video of Iranian missiles hitting Tel Aviv, which The Times of India also carried, and an AI image of an Iranian missile bearing an Epstein message, posted by Iran’s Foreign Ministry and then carried by NDTV and Republic TV.

West Bengal Assembly election: recycled footage, much of it from across the border

The election produced 66 fact-checks, with 35 in April, 29 in May and 2 in March. 38 came before the 4 May counting day and 28 from that day on.

42 of 66 (63.6%) used old footage and only 8 (12.1%) used AI.

In 17 of the 66, footage from Bangladesh, in whole or in part, was passed off as scenes from Bengal. Several clips came from Bangladesh's February 2026 election and army drills, and were shown as central forces or the Army acting against TMC workers.

Before polling, claims centred on security forces cracking down on TMC. After the result they turned to post-poll violence, "infiltrators" or Muslims leaving the state, bulldozer drives and BJP victory celebrations.

AI also appeared in the final stretch of the campaign. Five of the 20 West Bengal election fact-checks BOOM published between the two voting dates, 23 and 29 April, involved AI. On 28 April, the day before the second round of voting, BOOM debunked an AI video of state education minister Bratya Basu taking a bribe, shared by BJP leader Agnimitra Paul among others, and an AI image placing BJP’s Locket Chatterjee at the TMC office.

Communal claims

There were 82 communal fact-checks (15.6%). Only 1 involved AI, while 37 (45.1%) used old footage.

Communal claims carried the highest harm potential of any major theme, with 75 of 82 (91.5%) coded as potentially harmful. Among them, the "Bangladeshi Hindus under attack" narrative produced 14 fact-checks.

In a Kushinagar case involving the sexual assault of a five-year-old girl, OpIndia and Republic Bharat carried a communal version blaming a school headmaster, but police had arrested someone else. Others include Amit Malviya’s false communal spin on a West Bengal idol-vandalism case and a Bangladesh video shared with a false communal claim by ISKCON Kolkata spokesperson Radharaman Das.

The first Cockroach Janta Party wave and other protests

84 of 527 fact-checks (15.9%) were about protests, which makes it the biggest umbrella category.

The largest group was the CJP mobilisation over the NEET paper leak, accounting for 43 fact-checks in this period (2 in May, 7 in June and 34 in July). 16 of the 43 were AI and 21 used old footage.

BOOM verified that a viral video of batons with nails was real and recorded at the 20 July protest, after Delhi Police dismissed it as AI-generated. Other examples include AI visuals shared as deaths after the lathicharge, a cropped video falsely shared as protesters raising Islamic slogans by BJP’s Tajinder Bagga, and a video wrongly shown as BJP workers pelting stones at protesters in Kolkata by CPI(M)’s Md Salim.

BOOM also found at least six AI-generated women in sexualised outfits and suggestive posts, with their faces superimposed onto real footage from Jantar Mantar and passed off as CJP protesters. The most viral video passed 24 million views and steered viewers to paid explicit content, and some X users used the fakes to ridicule the real women protesters.

The other protest groups were smaller and mostly built on old footage.

All 13 fact-checks on the UGC rules protests (late January to February) used old or unrelated footage, with no AI-led claims.

The 12 claims on Iran unrest in January passed off solidarity protests abroad and unrelated clashes as scenes from Iran.

Of the seven fact-checks on the Noida workers’ protests in April, 5 broke out. A false claim that police firing killed 14 protesters led Noida Police to file an FIR against two X handles, and a BOOM ground report found that a video of police assaulting women, which Gautam Buddh Nagar police had called morphed or AI, was real..