It’s almost impossible to avoid seeing AI-generated content online, but it doesn’t have to be. YouTube, Instagram, TikTok, and more have ramped up content verification efforts over the past year, with many now automatically labeling AI-generated images, videos, and music to distinguish them from real, human-created images.

It’s all well and good if we’re just randomly stumbling across labeled content, but you know what would be better? Let us filter the AI ​​slope.

Current labeling efforts have not meaningfully changed the way content is presented online. You may notice that some TikTok or YouTube videos in your feeds now have AI disclosures in the description, or informational labels are overlaid on the clip itself. Meta takes a similar approach by labeling images on Facebook and Instagram with “AI information” that identifies AI metadata or voluntary disclosures by creators.

But if you actually want to escape Seeing anything tagged with such labels – which is justifiable, given the brain damage it raises ethical and environmental concerns around creative AI – is actually incredibly difficult to do. A filter will solve this easily. All we need is an “AI” checkbox to toggle.

I reached out to Meta, Google, TikTok and Spotify to ask if they plan to let users filter the different content they’re authenticating with the AI ​​labeling system. TikTok and Spotify never responded, and Google said it had nothing to share. Meta did not provide an attributed comment. But to summarize, none of these companies said “yes”.

One of the only online platforms I’ve seen with an AI content filter is DeviantArt, and its implementation is quite remarkable. For one, you can’t access it on DeviantArt’s feeds or store page, so it feels somewhat hidden. Instead, you’ll need to create an account and then hover over your user icon at the top right of the page to find the “AI Content Settings” menu. From there, you only have two options: the default “Show AI” setting, or the “Suppress AI” option which claims you’ll see “fewer examples” of AI-generated or manipulated images.

After trying both options, I, unfortunately, don’t see any noticeable difference. I’ve had good enough eyesight to spot AI-generated “digital illustrations” at this point, but I didn’t have to rely on my suspicions alone – almost every suspicious image I selected included a creator disclosure in the description confirming that the work was spit out by a robot. DeviantArt does a poor job. Automatic AI labeling For images with metadata that clearly indicates AI generation.

Pinterest A similar system exists. Users signed into a Pinterest account can click the Settings icon, select “Improve your recommendations” and then tap the “AI Content” tab to toggle through specific categories, including art, beauty, fashion and home decor. According to Pinterest, disabling any of these options will show you “less AI-edited content” for that particular category, but in my experience, it’s far from perfect. Sorting is also harder to find than the filters built into Pinterest’s feeds. Even with the AI ​​filters maxed out I still saw a lot of images suggesting dubious AI (including unusually perfect photographic models and inexplicable illustration errors).

And this is definitely what would happen if other platforms like YouTube or Instagram introduced an AI content filter: it wouldn’t work very well. But that’s okay because it will expose the ineffective “solutions” that our AI emperors devise themselves. They exist on paper, to appease regulators and critics, but do little to address the real problem of distinguishing AI fakes from authentic photographic and creative works.

and platforms do it Know that this is a problem. Instagram chief Adam Mosseri said in December that “authenticity is becoming a scarce resource” amid the rise of AI-generated content. And now we have Google CEO Sundar Pichai admitting it recently. Decoder “There’s a lot of AI slop,” says the interviewer, and online users need to “adapt to it.” Well, give us the filters.

Provenance-based systems such as C2PA and SynthID work by adding metadata or invisible watermarks to content at the point of creation. But there are plenty of open source AI models that don’t (especially if they’re built for nefarious purposes), and even then, the metadata can be extracted pretty easily to make it trustworthy. There are also detection-based methods that analyze patterns in digital content and then rank the likelihood that AI was used to create it, but these can provide false positives. None of this currently works effectively at scale.

Despite this, companies, including AI providers like OpenAI, are currently touting these AI labeling solutions as something that will help prevent people from being defrauded by deepfakes and other misleading forgeries. If regulators realize how ineffective they are, online platforms and AI providers may actually need to find solutions that does Work, instead of what currently feels like a smokescreen.

Platforms will argue that if they emphasize labeling measures too much, they risk falsely flagging authentic content. Meta and both Youtube The hard way was found after applying AI labels to photos and videos that the creators said were produced without the help of such tools. If this is such a concern for existing labeling systems, Then find a better solution. Surely improving the user experience for our millions of customers is a worthwhile investment to fend off competition?

And while I’m asking, why can’t I report all the unlabeled AI slop I see every day? Given the scale of the matter – with A study last year by Kapwing Finding that more than 20 percent of YouTube videos shown to new users are low-quality slop, for example—I imagine that would require a lot of human moderators to effectively evaluate each report.

And maybe that’s the rub. At a time when big tech is replacing workers with AI that can outperform them, can it afford to backtrack on its carefully crafted narrative by hiring them to solve AI problems? Just like humans, they have annoying needs. Salary and benefitscompared to automated moderation systems that lack investigative skills.

An alternative to labeling AI-generated content would be to start labeling verified human creators instead. This does not necessarily identify the artificial content posted. by the They are creators, but it can help us look less at unverified content farms that produce low-quality slop. That’s the future that Instagram’s Mossari envisions for Meta’s image-sharing platform, and something Spotify is doing with pre-verified artists.

Of course, Meta, Spotify, and Google don’t just host AI-generated photos, ads, and music. They are also responsible for creating the tools that create it. That is why they insist that no. all AI content is slippery and that’s more of a quality issue – if it gets convincing enough, they’re hoping you’ll ignore it and happily slide down the trough. Allowing users to filter it regardless would go against all the efforts these platforms have made to take advantage of AI: they want To hug you slip factory.

I’m happy to be proven wrong. I’m actually clean The plea For online platforms to prove that AI labeling efforts are not a waste of time. But right now, they hold all the cards and we’re just left to hope that their attempts at AI mediocrity will fizzle out. So give us a basic “no AI” or “verified human creator” filter and we’ll judge how well it’s actually doing.