“Can You Really Create This Without AI? Show Us!”

“This looks like AI.”

As a writer who also dabbles in illustration and photography, this phrase sends a chill down my spine. With generative AI technology becoming increasingly skilled at mimicking human creativity, skepticism is only natural when online platforms fail to clearly label AI-generated content.

This brings me to a thought: perhaps we should start tagging human-created text, images, audio, and video with a universally recognized symbol, similar to a Fair Trade logo. While machines aren’t inclined to label their own creations, the human creators facing displacement certainly are.

I’m not alone in this perspective. The head of Instagram, Adam Mosseri, mentioned in December that it might be “more practical to fingerprint real media than fake media” as AI technology advances to produce content that looks indistinguishable from that crafted by professionals.

While it’s hard to quantify how much of the internet is AI-generated, a recent survey indicates a widespread belief that news sites, social media, and search engine results are filled with it. The C2PA content credentials standard was designed to authenticate human-made works, and it’s already in use on some platforms. However, its implementation has been largely ineffective, despite industry backing. Many creators of AI content have a vested interest in obscuring its origins due to the clicks and revenue it can generate.

In response to the need for human creators to differentiate their work from AI outputs, numerous solutions have emerged in recent years. Yet, like C2PA, these solutions face significant hurdles in achieving widespread acceptance.

Currently, there’s an overwhelming number of AI-free labeling options available—at least 12, each with different eligibility criteria and verification methods. Some are tailored to specific industries, like the Authors Guild’s “human authored certification” for written works, while others, such as Proudly Human and Not by AI, aim for broader applicability across various creative fields. However, the verification processes of these services can be just as dubious as those of AI labeling solutions. Some, like Made by Human, rely solely on trust, allowing anyone to download and use their badges without verifying authenticity. Others, like No-AI-Icon, claim to visually inspect works and run them through AI detection services, which can often be unreliable.

Most of the services I’ve explored require creators to manually demonstrate their processes to a human auditor, such as providing sketches and drafts. While this method is labor-intensive, it remains the most reliable way to confirm that a piece was created by a human.

Another challenge is defining what “human-made” actually means. With AI now integrated into many creative tools and its use encouraged by educators, where do we draw the line?

“The problem is going to be definition and verification. Does chatting with an LLM about the idea before executing it manually count as using AI? And how could the creator prove no AI was involved?” Jonathan Stray, a senior scientist at the UC Berkeley Center for Human-Compatible AI, pointed out. “Other consumer labels, such as ‘Organic,’ have regulations and agencies that enforce them.”

Nina Beguš, a lecturer at UC Berkeley, notes that we’ve entered an era of hybrid content that challenges our definitions of authentic creation. “Any creative output today can be touched by AI in one way or another without us being able to prove it,” she said. “Authorship is disintegrating into new directions, becoming more technologically enhanced and more collective. We need to revamp our creativity criteria that were made solely for humans.”

One contender in the human-made label space, Not by AI, is attempting to address this ambiguity. It offers various badges for creators to apply to their work, provided that at least 90 percent is created by a human. However, this voluntary approach lacks any verification of truthfulness.

Other solutions, like Proof I Did It, are leveraging blockchain technology to create a permanent record that can verify creators and their works. By storing verification on the blockchain, creators receive an unforgeable digital certificate proving their work was human-made, which is far more reliable than software trying to guess if a piece was AI-generated.

Thomas Beyer, an executive director at the University of California’s Rady School of Management, believes that Web3 and blockchain can provide a solid solution by shifting the focus from “does this look like AI?” to “can this account prove its human history?”

“By issuing ‘Made by Human’ tokens to verified creators, the market creates a ‘premium tier’ of art where authenticity is mathematically guaranteed,” Beyer explained. Other experts, including Beguš, share similar views on the potential rise in value of “human and biological creativity” amidst the influx of synthetic media.

Despite its shortcomings, established standards like C2PA offer something that AI-free labeling solutions desperately need: unification. Major tech players have committed to this standard, and AI providers are adopting it to satisfy global regulators. When weighing the pros and cons of AI labeling versus verifying authentic human-made content, the latter seems more likely to succeed.

Many creative professionals, even those who don’t entirely oppose AI tools, are understandably eager to distinguish their work from the AI-generated content flooding the industry and threatening their livelihoods. While there are plenty of AI advocates showcasing the technology’s capabilities, there’s a reluctance to disclose its use when it could jeopardize their influence or income.

Consider the case of adult film actors creating digital clones of themselves that remain perpetually youthful, or AI influencers selling an idealized life that doesn’t exist. Revealing that they’re AI could shatter the illusion for those seeking a genuine human experience. Scammers using AI-generated images to sell products online also have little incentive to be transparent, and platforms hosting such content often turn a blind eye. Similarly, those using generative AI to create chaos on social media thrive on the belief that their content is real. It’s no surprise that AI labeling efforts like C2PA have struggled to gain traction.

We know some AI-focused creators will avoid transparency, as evidenced by Coral Hart, a romance author who revealed she made a six-figure income from over 200 AI-generated novels last year. However, she doesn’t label her books as AI-assisted, fearing it would “damage her business” due to the “strong stigma” surrounding the technology.

This disdain is evident in how often AI-generated content is dismissed as “slop,” even when the works are visually or audibly impressive. This raises the question of how human-made or AI-free labeling providers will prevent their logos from being exploited by those profiting from deception. Trevor Woods, CEO of Proudly Human, admits that preventing misuse may be impossible.

“Like other certification marks and company logos, we cannot prevent fraudulently displaying the Proudly Human certification mark. However, we make it easy for consumers to verify it,” Woods stated. “If a bad actor identified by us refuses to stop using the label, we will take legal action against them.”

To achieve a universally recognized and enforced solution, a standard must be agreed upon not just by creators and online platforms, but also by global governments and regulatory bodies. Currently, those discussions are sparse.

“Proudly Human has occasionally briefed government and industry associations but is not involved in formal negotiations regarding a unified human origin certification,” Woods noted. “The rapid evolution of AI capabilities and AI-generated content will outpace government and regulator responses.”

Clearly, there’s a demand for making human-made works easier for consumers to identify. Creatives, regulators, and authentication agencies need to decide on a unified approach. If a single standard can achieve the same recognition as symbols like Fair Trade and Organic—despite their own concerns—perhaps we can return to a time when we trust what we see.