“this looks like it was made by ai.” it’s the phrase i dread seeing as a writer who occasionally practices the art of illustration and amateur photography. in a world where generative ai technologies are becoming increasingly adept at mimicking human work, it is only natural for people to feel skeptical when digital platforms refuse to label even content that is crystal clear in its machine origins.

this reality leads me to one conclusion: perhaps we should start labeling human-made texts, images, audio, and videos with something resembling the globally recognized “fair trade” logo. machines are certainly not interested in labeling their work, but creators facing the danger of marginalization and replacement by machines care deeply about it.

why is human creativity hard to distinguish?

fortunately, i am not the only one thinking this way. instagram head adam mosseri hinted at this last december, saying it would be more practical to water-mark authentic media rather than trying to track fake media, especially as ai technology evolves to the point where it can produce content visually indistinguishable from that made by professionals.

no one can say for sure how much content on the internet has been created by ai, but there is a widespread perception that news sites, social media platforms, and search engine results are teeming with this type of content, according to a recent survey by the reuters institute.

authenticating human works was the primary goal behind the content provenance and authenticity standard known as c2pa, already used by meta platforms. but so far, its implementation seems entirely ineffective, despite receiving broad support from the tech sector. it turns out that many people who create and publish ai content have hidden motives to conceal its origins because of the clicks, clutter, and money it can generate.

the labeling chaos and proposed alternatives

in an effort to help human creators distinguish their works from those produced by ai generators, a multitude of solutions have emerged in recent years. like previous standards, these solutions face several challenges that hinder their widespread adoption.

currently, there is an overwhelming number of alternatives to choose from to prove that work is free of ai. in total, i counted at least 12 alternatives, all trying to tackle the same problem with varying eligibility standards and authentication approaches. some are tailored for specific industries, such as the authors guild’s “human authored certificate” for books and other written works, which cannot be broadly applied to all forms of creative content.

other solutions, like the “proudly human-made” and “not by ai” initiatives, aim to be more inclusive, covering published texts, visual arts, video production, and music. but the verification processes used by these services can be just as questionable as those used by ai classification solutions. other initiatives, like “made by humans,” operate purely on mutual trust, making badges and labels publicly available for anyone to download and apply to their work without proving actual provenance.

most of the services i checked take the hard route: forcing creators to manually show their preliminary drafts and sketches to a human checker. it is a very labor-intensive process, but without technological shortcuts, this remains our most reliable method currently to determine if something was made by a real human.

what is the definition of “human-made” in the tech era?

another issue is agreeing on the very meaning of the term “human-made.” with ai now integrated into so many creative tools, and educators in artistic fields encouraging its use, where do we draw the line?

*

in this context, jonathan stray, senior scientist at the center for human-compatible ai at uc berkeley, told the verge: “the real problem will lie in definition and verification. is chatting with large language models about an idea before executing it manually considered using ai? and how can a creator prove machine non-interference? other consumer labels, such as organic product labels, have regulations and regulatory bodies enforcing them.”

nina beggosch, a lecturer at the uc berkeley school of information, points out that we have already entered the era of hybrid content that conflicts with how we define something as authentically made. she said: “any creative output today can be touched by ai in one way or another without us being able to prove it. the concept of authorship is unraveling toward new directions, becoming more technologically and collectively augmented. we need to renew our standards of creativity, which were originally built for humans only.”

blockchain technology as a promising solution

one solution offered by a competitor to human-made labels, calling itself “not by ai,” attempts to account for this ambiguity. it offers a variety of badges that creators can add to websites, blogs, artworks, films, articles, books, and podcasts, provided that at least 90 percent of the work is created by a real human. however, this voluntary approach lacks any mechanisms to verify truthfulness.

other solutions, like the “proof of human” initiative, rely on blockchain technology to provide a permanent record that anyone can use as a reference to verify creators and works documented by the service. by storing verification on these decentralized records, creators receive a tamper-proof digital certificate proving that a human did the work, which is far more reliable than trying to use software to guess whether media was generated by a machine.

thomas baer, an executive fellow at the rhea school of management at uc san diego, says that web3 and blockchain technologies could provide a powerful solution by shifting the question 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 an elite class of art where authenticity is mathematically and definitively guaranteed.” – thomas baer

do creator labels outperform tech labels?

despite their flaws, established standards like the content provenance protocol provide something that ai-free content labeling solutions desperately need: uniformity. big tech names like adobe, microsoft, and google have committed to this standard, and ai providers are implementing it to appease global regulators. however, when i weigh the pros and cons, i feel that efforts focusing on verifying authentic human content are much more likely to succeed.

many creative professionals have understandable motives to distinguish their work from the synthetically generated competition flooding the industry and threatening their livelihoods. while there are plenty of tech evangelists on social media who are happy to show what it can achieve, there is profound hesitation to disclose its use when money and influence are on the line.

scammers using generated images to sell products online certainly don’t want to be exposed, and platforms hosting them like etsy do not seem overly concerned. similarly, anyone using these technologies to sow discord on social media can only succeed when people believe what they post is real. it is hardly surprising then that ai labeling has failed to catch on.

the path toward a reliable global standard

we know some content creators will avoid transparency. a prominent example is coral harte, a romance author who told the new york times she made massive amounts of money producing over 200 novels via ai last year. she puts no label on her books indicating they were written using these tools, fearing it would “harm her work” due to the “strong stigma” surrounding this technology.

if the goal is a globally recognized and actionable solution, a unified standard must be agreed upon—not just by creators and platforms, but also by global governments and regulatory authorities. to my understanding, these conversations are currently few and far between.

trevor woods, ceo of the “proudly human-made” organization, says: “we cannot prevent the fraudulent display of our certification mark, just like other certification marks and corporate logos. however, we make it easy for consumers to verify them. and if any identified bad actor refuses to stop using the mark, we will take legal action against them.”

clearly, there is a growing demand to make it easier for consumers to recognize human-made works. and if a single standard can rise to the same level of symbols like “fair trade” and “organic”—which carry their own concerns but are globally recognized—perhaps we can return to the days when we trusted what our eyes see.

frequently asked questions

why are creators demanding a label that distinguishes human content?

creators, including writers, artists, and photographers, are demanding a label to protect their work from unfair competition with machine-generated content, and to prove the authenticity of their works to an audience that now finds it difficult to distinguish between human creativity and machine output.

what are the most prominent challenges facing the implementation of this label?

the biggest challenge is how to accurately define “human content” given the integration of ai technologies into most modern design and writing tools. verifying that a work is completely free of machine intervention also represents a complex technical and logistical challenge.

how can blockchain technology help in this field?

blockchain technology (decentralized data ledgers) provides a way to create tamper-proof and unalterable digital certificates. thus, creators can document their work stages and link them to their identity, ensuring that the buyer or recipient knows the work is authentic and made by a real human in a definitive and mathematically proven manner.

why do platforms fail to automatically label ai content?

platforms fail to do so because many creators of this content, and sometimes the platforms themselves, profit financially from hiding the fact that content is machine-generated, as it brings them more views, clicks, and engagement that might drop if the audience realizes the content is not real.