Social Platforms Tighten Guidelines on AI-Generated Content

YouTube is sharpening its stance on AI-generated content.

In an update from the company's trust and safety team, on July 20, YouTube said it wants to reward creators who use AI to enhance original storytelling, as opposed to those who mass-produce generic content or AI personas designed to game the platform.

A YouTube spokesperson told Business Insider that the policy isn't new but an important clarification of YouTube's long-standing "inauthentic content" policy, intended to reduce ambiguity for content creators who have been demonetized for repetitive, low-effort, or emotionally manipulative videos.

YouTube has spent years pitching itself to advertisers as the future of television, and a surge of AI-generated spam threatens to undermine that message. In January, the platform removed more than a dozen popular channels that had amassed millions of views by pumping out AI-generated videos featuring characters like cats and Jesus. Video editing platform Kapwing published a report estimating that 21% of YouTube's feed was AI-generated videos.

TikTok isn't shying away from generative AI. The company and its owner, ByteDance, offer an array of AI tools for content creators on its marquee app and video editor CapCut. Creators who post on its e-commerce platform, TikTok Shop, have been ramping up the use of AI to sell products this year. Much of the company's tech is built around ByteDance's AI video model, Seedance 2.0.

Still, TikTok is taking a variety of steps to keep generative AI under control across its ecosystem. The company requires creators to label AI-generated content and appends an invisible watermark to the metadata of some videos to indicate they are artificially generated. In November, the company introduced the ability for users to toggle the amount of AI content that appears in their feeds. In July, the company said it's beginning to test "improved detection systems to target accounts dedicated to posting AI-generated spam that crowds out original creators."

The newsletter platform announced on July 21 that it would help readers identify AI-generated content.

Substack will now let readers scan content to estimate how much AI may have been used in its writing. Substack is partnering with Pangram, an AI detector tool, to power its AI identification tools.

Creators on Substack can also add a disclosure to their content explaining how they made it, such as noting that they used Claude for copyediting, or if they don't use AI at all.

However, users can fairly easily disable this tool for their Substack content by turning off AI analysis in their publication settings. Creators can opt out of Pangram's automated scanning, meaning readers will no longer see an AI-use estimate on those posts.

Pinterest is no stranger to AI slop.

In 2025, the platform began adding AI labels to content. Pinterest lets creators label content as AI-modified or -generated, and has its own detection systems to flag content that may have been AI-generated.

Pinterest also added a tool that lets users limit the amount of AI-generated content appearing in their feeds.

In June, Pinterest's CEO Bill Ready told Business Insider's editor in chief, Jamie Heller, at Cannes Lions that while there are "so many great uses of AI," there is also a "bit of a counter movement to that with consumers."

"We provide users with controls to decide whether they want to see more or less GenAI content in their feed," a Pinterest spokesperson told Business Insider. "Additionally, our Community Guidelines apply to GenAI content, and our recommendation system prioritizes high-quality content, regardless of if it's human-created or GenAI."

Meta wants users to create and engage with content made with its own AI tools, so, for the most part, it has let content like the famed Shrimp Jesus run free.

However, the company has been labeling content as AI-generated since 2024. Across Facebook, Instagram, and Threads, Meta adds an "AI info" label to content it detects as AI-generated. It doesn't automatically do the same for content that was modified with AI, though. Meta "may require" its AI labels on content with "photorealistic video or realistic-sounding audio that was digitally created, modified or altered, including with AI," per the company's help page explaining its AI policies. People can also voluntarily add the label to their content.

In June, Meta started putting its "AI info" label on s that use AI, whether they're created by Meta's own tools or third-party editing services.

If your LinkedIn posts sound a little too polished, readers may be taking notice.

AI detector Pangram estimates that 41% of long-form LinkedIn posts and 30% of short-form posts from April through June were AI-generated, which is the highest share among platforms it analyzed, including X, Reddit, Substack, and Medium.

Some creators are already changing course. Steven Bartlett's media company, FlightStory, stopped using AI-written LinkedIn posts after concluding authentic, human-written updates performed better.

LinkedIn actively promotes AI writing tools, including a button that helps users polish posts with AI. The company, however, said in a June post that it will be taking steps against "slop."

"While AI can be a helpful tool for refining language, we're seeing a rise in what many call 'AI slop,' content that is low-effort, AI-generated content that may sound polished on the surface but lacks any real unique perspective or substance," the company wrote.

Snap is joining the growing pushback against AI slop.

The company said on July 31 it will no longer recommend or reward fully AI-generated videos in Spotlight, its TikTok-like short-form feed, as it shifts its algorithm to prioritize content made by real people.

"Spotlight should remain a place where people can discover authentic creativity from real people," Snap said in a blog post, adding that original storytelling and personal perspectives should be rewarded.

Snap said that creators can still use Snap's AI tools to enhance or edit videos, but fully AI-generated content won't qualify for recommendations.