AI-generated war footage can reach millions before anyone checks it. Paid verification, engagement incentives and slow corrections are making X a dangerous place for wartime misinformation.
The first casualty when war comes is truth.
This quotation is frequently attributed to US Senator Hiram Johnson (1866–1945), and it has survived the century in which it originated. It is more relevant today than ever. A phone and a social media platform can now deliver fake airport attack news into your hand while you are waiting for the bus. Before you decide or verify whether the news is real, tens of thousands of people have already shared it. Most of us do believe such news, at least for a few seconds, and that is all a rumour needs.
This is not fiction. During the 12-day war between Israel and Iran in June 2025, an AI-generated video of a wrecked airport got around 6.8 million views on X. According to various global studies, the majority of X users prefer the mobile version of the application over the web. It is an important finding because a mobile device is what you check in the gaps of the day with half of your attention, and it is not always the right time to differentiate a real strike from a convincing fake.
A verified blue badge or check on the platform once used to be a sign of authenticity, if not credibility. It used to check who you were, and since journalists had to send proof of identity, an employer official email and published work to verify, the process was much more transparent and strict. Today, anyone who pays a monthly fee can have one. X also pays creators for engagement from verified accounts through an arrangement that rewards virality long before it rewards accuracy.
The first week of the Gaza war was what the whole debate looks like in practice. For example, NewsGuard tracked 250 of the most-engaged false or unverified posts about the war between 7 and 14 October 2023. They found that 74 per cent of the posts came from verified accounts on X, while only 79 carried a Community Note. Separately, Bloomberg found that Community Notes typically took more than seven hours to appear on misleading posts about the conflict, and in a war, seven hours is a long time. The average wait for a community note to go live has improved, from more than 100 days in 2022 to about 14 in 2025. However, a fortnight is an eternity and enough time for a fake post to achieve the desired outcome. Similarly, researchers at the Atlantic Council’s DFRLab analysed about 130,000 X posts from the first days of the June 2025 war and found X’s Grok inconsistent and often unable to differentiate AI-made footage from real content.
In the war that began with US and Israeli strikes on Iran on 28 February 2026, CNN reported that AI-created fakes had racked up tens of millions of views in the first two weeks. According to the BBC’s dedicated investigative and fact-checking team known as BBC Verify, Grok kept telling X users for months that those AI videos and images, including some about this war, were real. As a countermeasure, X’s answer was to suspend creators from its payment program if they post undisclosed AI videos of armed conflict. However, the fact that most involved users were not in the program anyway made it just a face-saving stunt.
According to DFRLab, early AI use in this conflict came partly from Israeli public diplomacy, and Iran later began saturating the space with AI-generated war material. A monitoring firm known as Cyabra also counted 145 million views for an Iran-linked network of fake accounts in the first two weeks of the 2026 war.
Ownership also plays a key role in shaping the core philosophy of a social or digital media platform. And in X’s case, that philosophy is linked to the business goals and the operational transparency of its algorithm. For Elon, incentives serve fakes from every direction. I believe his own politics and ideology deserve scrutiny and explanation because he controls the algorithm, the payout formula and the Grok experience on X. Elon Musk bought Twitter in October 2022, promising a platform built around free speech, and in July 2024 he endorsed Donald Trump for president in a post on the site he owns. Researchers at Queensland University of Technology later reported that engagement on Musk’s own account rose sharply after that endorsement. Musk has also endorsed many political groups and parties in the past that aligned with his political beliefs, and critics have blamed him for running coordinated campaigns and engaging in algorithmic manipulation in favour of certain groups.
X is not only a rumour mill, and plenty of people would be worse off without it. Governmental restrictions on mainstream news outlets globally make platforms like X the only way to get firsthand information. When X launched its ‘About This Account’ feature in November 2025, it unmasked accounts across the political spectrum with large followings pretending to be from another country or region. Although X has admitted that VPNs can fool it, it is at least a step in the right direction.
I believe that these changes are not enough and are very superficial, and X needs to take significant steps to improve its ecosystem. It needs to make sure that the blue tick means identity again. And if Musk needs the money from paid verification, he can keep a paid tier, but it should call it something else. It should discipline accounts for being wrong, which means cutting off revenue sharing for accounts whose posts are repeatedly debunked in conflict reporting. It also needs to tag AI-generated war footage at the platform level rather than relying on the creator to admit it. Lastly, X should introduce friction for unverified combat footage during active conflicts, either through a brief delay or a prompt to read before sharing.
And for those of us holding the smartphone, pause for a moment, ask whether a known debunker or a newsroom has examined the video clip, and read the replies along with any Community Note. Ask yourself whether it is credible. And even if it is credible, wait ten minutes. If it is real, it will still be real, but don’t kill the truth.
Muhammad Adnan
Dr. Muhammad Adnan is a research officer at Western Sydney University. He is an early-career researcher whose broader research interests include journalism, politics and technologies.