These days, it’s getting harder to tell what’s real and what’s fake. Fake AI images can be incredibly convincing, but people are also quick to claim that real photos or videos are fake. This back-and-forth creates what experts call the ‘liar’s dividend’—a situation where dishonest figures can dismiss genuine evidence as AI-made, muddying the truth.

Why the ‘liar’s dividend’ matters

Back in 2019, legal scholars Bobby Chesney and Danielle Citron introduced the term ‘liar’s dividend’ to describe how false claims about AI content can benefit those spreading misinformation. When public figures or governments are caught in questionable videos or images, they often claim it’s an AI fake, even if it’s real. This tactic makes it easier to dismiss damaging evidence and sow doubt among the public.

Research supports this. A study involving over 15,000 Americans found that when politicians falsely accused evidence of being AI-generated, support for them often increased. Basically, when people are told to doubt what they see, some might turn against the truth, especially during elections or political scandals.

New AI tools and the race to verify authenticity

Google has introduced a new image generator called Nano Banana, or Gemini 2.5 Flash Image. It can create realistic pictures and even modify existing ones with simple instructions. It’s so good that fake images can keep consistent faces, clothing, and other details across scenes, making them look very real. This tool is available to developers through Google’s API and is expected to be integrated into many apps soon.

To combat fakes, Google embeds a hidden watermark called SynthID into every image or edit. This watermark helps identify whether an image was generated or altered by AI, even after resizing or compression. Similarly, OpenAI’s DALL-E 3 and ChatGPT now include metadata called Content Credentials, which record details about the AI tool used and a cryptographic signature. These help verify images but can be removed or altered, so they aren’t foolproof.

Other companies, like Meta and Adobe, are also working on verification methods. Meta uses invisible watermarks and metadata, while Adobe is pushing for a universal system called Content Authenticity Initiative. TikTok has started adding content credentials and can label AI-made media automatically. Still, all these measures can be bypassed by someone determined to pass off fakes as real.

The decline of reliable visual evidence

Since photography became a trusted form of evidence in courts and newspapers over 160 years ago, our confidence in visual proof has been high. Photos and videos have long been considered solid proof in courtrooms and for public opinion. But now, with AI-generated content so advanced and verification tools still imperfect, that trust is fading.

It’s a troubling shift. As AI tools become more sophisticated, fake images and videos can look very convincing. Even experts can’t always tell if a picture is real or AI-made. This raises a big question: how can we rely on visual evidence in the future? The era of unquestioned photographic proof seems to be coming to an end, replaced by a landscape full of doubt and deception.

In the end, the battle over truth and fake content is just beginning. As technology advances, so will the tools to both create and detect fakes. The challenge for everyone is to develop better ways to verify what we see, but the reality is that the line between real and fake will keep getting blurrier.

Inspired by

• https://www.computerworld.com/article/4051728/ai-and-the-end-of-proof.html

Sources

• gmpg.org
• vice.com
• californialawreview.org
• cambridge.org
• purdue.edu