How do you spot AI-generated fake videos on social media?

Learn how to verify viral social media videos by spotting AI visual glitches and contextual errors, using a recent fake London Tube video as a guide.

Fact-checking · Published 2026-08-10 · 8 min read · 1,665 words

AI-generated illustration. A magnifying glass held over a smartphone screen showing a distorted, computer-generated video of a train carriage.

Social media feeds are increasingly populated by video content designed to trigger immediate emotional reactions. When a video sparks sudden outrage, anxiety, or confirmation of a deeply held belief, the impulse to share it is strong. However, the barrier to creating convincing synthetic media has dropped significantly, and this week has provided clear examples of how AI-generated and recycled videos are used to manipulate public perception.

On August 5, 2026, a highly inflammatory video began circulating on platforms like Facebook and X. The footage claimed to show a group of Muslim men blocking a carriage on the London Underground to pray mid-journey, with a caption reading: "MUSLIMS BLOCKING THE TUBE IN LONDON AND FRUSTRATING COMMUTERS !!!!!" The video included ambient audio of frustrated passengers saying things like "we can’t get off" and "can I get to work please."

According to a debrief by Full Fact, the video is entirely fake. It was generated using artificial intelligence. The fact-checking organization noted that the video contained visual glitches typical of AI generation, and the depiction of the prayer itself was inconsistent with how real Muslim prayers are conducted. The video was engineered specifically to exploit cultural tensions and generate engagement through outrage.

This specific incident is part of a broader, ongoing pattern of video-based misinformation that relies on either artificial generation or the recycling of old footage out of context. Geopolitical events are frequent targets. On August 4, 2026, Misbar debunked a viral video claiming to show Iran successfully shooting down a United States F-35 fighter jet. In reality, the footage was not a genuine military recording but a composite created by splicing two separate, unrelated clips together.

The consequences of this type of misinformation extend beyond digital engagement and can impact physical security and border control. During the recent migrant surge in Ceuta, where tens of thousands of people attempted to cross from Morocco into the Spanish enclave, social media platforms played a direct role in escalating the crisis. Rumors spread rapidly online claiming that the border was open and that migrants would be bused to mainland Spain.

As reported by Euronews, the information ecosystem surrounding the Ceuta crossings was heavily polluted with AI-generated photographs and recycled older footage, all paired with misleading captions to muddy the waters. The situation reached a threshold where EU Executive Vice-President for Tech Sovereignty, Security and Democracy Henna Virkkunen had to intervene. According to The Local, Virkkunen held talks with Meta and TikTok, demanding that the platforms "act decisively" to stem disinformation during crisis situations and maintain the integrity of the digital space.

Recycling old footage to fit a current narrative is just as common as AI generation, and often easier to execute. This week, a viral video claiming to show fresh evidence of corruption at South Africa's Home Affairs department was revealed by News24 to be seven years old. Similarly, India Today fact-checked a video claiming to show a man being swept away during recent floods in Kerala, India. The footage was actually old video showing flooding in Karachi, Pakistan, and had been circulating online since at least August 2025.

To navigate this environment, readers must develop the reflex to verify video content before accepting its premise or sharing it. Verification is a methodical process that relies on observation and basic digital tools.

1. Analyze for visual glitches and AI artifacts.Current AI video generation models frequently struggle with spatial consistency and fine details. When watching a suspicious video, pause and look closely at the edges of moving subjects. Check the hands and faces of background characters, which often blur, morph, or feature incorrect proportions. Look at the physics of the scene: do shadows align with the light sources? Do objects pass through one another unnaturally? In the fake London Tube video, these subtle visual distortions were key indicators that the footage was not captured by a camera.

1. Check for contextual and logical consistency.Synthetic media often fails to accurately replicate complex human behaviors or specific environmental details. As noted in the Tube video debunk, the physical actions of the people praying did not align with actual religious practices. You should also evaluate the environment. Does the layout of the train carriage match the actual design of London Underground trains? Are the background signs legible, or do they feature the garbled, nonsensical text common in AI generations?

1. Isolate keyframes for reverse image searching.If a video is a recycled clip from an older event, visual analysis alone will not reveal the deception, because the footage itself is real. In these cases, you must trace the video's origin. Pause the video on a clear, distinct frame that shows a recognizable landmark, face, or unique action. Take a screenshot. Upload this screenshot to a reverse image search engine like Google Lens, Yandex, or TinEye. If the video is recycled, the search results will often surface older news articles or social media posts containing the exact same frames, proving the footage predates the current claim.

1. Evaluate the audio track.Audio is frequently manipulated independently of the video. In composite videos or AI fakes, the ambient sound may not match the acoustics of the environment shown. Listen for unnatural looping in background noise, or dialogue that sounds detached from the physical space. In the London Tube fake, the voices of the "frustrated commuters" were overlaid to drive the emotional narrative, rather than recorded organically in the space.

You can perform these verification steps manually, or you can use software to assist in gathering the context. For instance, you can use the free wyper web app to paste the URL of a suspicious post and receive a truth score accompanied by an evidence chain of real sources.

Who needs which: Manual checks are best for researchers and journalists who require absolute control over their investigation process and prefer not to share any browsing data or URLs with third-party tools. The wyper web app is ideal for casual users on mobile devices or work computers who want to quickly verify a specific link without installing any software. The wyper extension is designed for frequent social media users who want real-time, automated fact-checking context layered directly over their feeds as they browse.

The next time you encounter a video that makes you immediately angry or anxious, stop and test these skills. Take a screenshot, run a reverse image search, or analyze the visual details for AI artifacts. Once you find the real context behind the footage, reply to the original post with your findings and a link to a credible source. Fact-checking is an active process, and sharing the evidence chain helps slow the spread of manufactured outrage.

The short version

Viral videos designed to spark outrage, like the AI-generated clip of Muslims praying on the London Underground, are increasingly common on social media. You can verify these claims by looking for visual AI glitches, checking contextual logic, and using reverse image searches on video keyframes. Always pause and verify the evidence before sharing emotionally charged content with your network.

Keep reading