How to Spot AI-Generated Content

The phrase “this looks like AI” has become a common, and often dreaded, reaction online. As generative technology advances, skepticism grows when platforms fail to clearly identify synthetic media. This environment points toward a potential solution: rather than forcing unreliable AI labeling, perhaps we should proactively certify authentic human-made content with a recognizable, trusted badge. The creators whose livelihoods are at stake have a clear incentive to adopt such a marker, while AI systems have none.

This idea is gaining traction. Instagram’s Adam Mosseri noted last December that as AI improves, it may become “more practical to fingerprint real media than fake media.” Public perception already suggests a problem; a Reuters Institute survey indicates many believe news sites and social media are flooded with AI-generated material. Established technical standards like the C2PA content credentials, backed by Meta and other major firms, were meant to authenticate origin. Yet their implementation has largely failed because many who produce and distribute AI content are incentivized to obscure its source, driven by the potential for revenue, engagement, or misinformation.

In response, numerous initiatives have emerged to help creatives distinguish their work. However, the landscape is now fragmented. At least a dozen different “AI-free” badges and labels exist, each with varying rules and verification methods. Some, like the Authors Guild’s certification, are specific to books. Broader services like Proudly Human or Not by AI cover multiple formats but face significant challenges in proving their claims. Verification processes range from unreliable AI detection scans to systems based purely on trust, where badges can be downloaded and applied by anyone without proof.

The most reliable method currently is also the most arduous: having creators manually submit drafts, sketches, and process evidence to a human auditor. This labor-intensive verification lacks scalability but offers greater confidence than technological guesswork. A deeper issue is defining what “human-made” even means in an age where AI is embedded in common creative tools. As UC Berkeley’s Jonathan Stray points out, “Does chatting with an LLM about an idea before executing it manually count as using AI? And how could the creator prove no AI was involved?” Without clear, enforceable standards akin to those for “Organic” labels, ambiguity prevails.

We are already in an era of hybrid content creation, argues Nina Beguš, a lecturer at UC Berkeley. “Any creative output today can be touched by AI in one way or another without us being able to prove it,” she notes. This reality is pushing some labeling efforts to accommodate nuance. The Not by AI badge, for instance, allows use if at least 90 percent of a work is human-created, though it operates on a voluntary honor system. Other projects are exploring more robust verification. Services like Proof I Did It use blockchain technology to create a permanent, unforgeable record of human authorship, shifting the question from whether something looks AI-made to whether it has a verifiable human history.

Thomas Beyer of UC’s Rady School of Management sees potential in this Web3 approach. “By issuing ‘Made by Human’ tokens to verified creators, the market creates a ‘premium tier’ of art where authenticity is mathematically guaranteed,” he explains. This concept highlights a growing belief that human and biological creativity could carry increased cultural and economic value amid a deluge of synthetic media.

Despite its current shortcomings, a unified standard like C2PA offers what the patchwork of AI-free labels desperately needs: cohesion. Major tech companies have committed to it, and AI providers are adopting it under regulatory pressure. Yet when comparing efforts to label AI versus those certifying human origin, the latter may have stronger momentum. Many professionals are highly motivated to differentiate their work from the AI-generated “slop” threatening their industries. Conversely, those using AI for profit, from romance authors generating hundreds of novels to sellers of digital clones and AI influencers, often avoid transparency to maintain illusion, revenue, or influence.

The case of author Coral Hart illustrates this dynamic. She reportedly earned a six-figure sum from over 200 AI-generated novels last year but does not label them, fearing a “strong stigma” would damage her business. This reluctance underscores a key challenge for any certification: preventing bad actors from fraudulently using a “human-made” badge. Trevor Woods, CEO of Proudly Human, admits they cannot prevent all misuse but states they will take legal action against identified offenders and make verification easy for consumers.

Achieving a universally recognized standard will require alignment not just among creators and platforms, but also governments and regulators. Such coordinated discussions appear rare. Woods notes that while his organization has briefed some officials, it is not in formal negotiations, and the “rapid evolution of AI capabilities… will outpace government and regulator responses.” Despite the hurdles, clear consumer demand exists for reliable ways to identify human creativity. The creative community, authentication services, and regulators must converge on a single, enforceable approach. If a symbol can attain the global recognition of a Fair Trade or Organic logo, we might eventually restore some trust in what we see online.