The latest
Lifelike AI-generated images, video and audio are now routine in social-media feeds, forcing users to assess cooking clips, advertisements, celebrity images and political content that can look convincing on a phone. The Wall Street Journal reports that some synthetic material is harmless and openly fictional, while other posts seek to provoke, deceive or scam. Digital-forensics researchers warn that repeated uncertainty can make people question authentic material and erode a shared understanding of reality.
Details
- Quality leap: Earlier AI images often revealed themselves through six-fingered hands or garbled lettering. Powerful models including Google’s Nano Banana and Veo 3 have largely erased those familiar tells. Remaining clues can include inconsistent lighting, reflections, perspective and shadows, or wavy lines in structures that should be straight, although no single anomaly proves an image is synthetic.
- Fictional following: Creators Eric Suerez and Adam Vaserstein used Veo 3 to develop Granny Spills, a pink-obsessed, luxury-wearing fictional septuagenarian, during the July Fourth weekend in 2025. Her Instagram account reached one million followers within a month. Vaserstein, who spent years building audiences for street-interview comedy, said AI characters were much easier to grow.
- Disclosed identities: Granny Spills and digital model Shudu do not conceal their artificial origins. Shudu’s creator, Cameron-James Wilson, said audiences can enjoy characters they know are unreal, comparing that response with established fictional figures. He acknowledged synthetic models could reinforce unrealistic beauty standards, while noting fashion photography had relied on extensive retouching for decades.
- Reality under strain: Hany Farid, a digital-forensics expert and Dartmouth College computer-science professor, said detecting generated media remains difficult even for specialists. He argued that human perception evolved to react to apparent threats, not determine whether they were fabricated. More troublingly, he said, people are beginning to dispute verifiable events, endangering society’s shared account of what happened.
- Audio doubt: AI-generated tracks reached the Billboard charts for the first time in 2025, according to the report, prompting French streaming service Deezer to build a synthetic-audio detector. A scan of the author’s library identified 1% as AI-generated, a small share that nevertheless caused her to question the remainder.
- Verification routine: Users should ask whether a post matches the physical world, then inspect who published it and why. The report recommends checking account age, posting frequency and engagement bait, and cross-referencing locations through Google Street View. It also advises pausing over unusually delightful or enraging content because social algorithms often amplify strong emotional reactions.
What’s next
The concrete indicator to watch is whether more streaming and social-media services deploy synthetic-content detectors comparable with Deezer’s audio tool. Until such checks become widespread, users are advised to inspect accounts, verify locations, and examine lighting, shadows and reflections before sharing questionable material.
