How Can You Tell AI Slop from the Real Thing?

“AI slop” is an informal label for low-value synthetic content, not a scientific category with one body count. Claims about platform-wide views, physical harm, and human involvement require source-specific definitions and denominators.

Where the flood is actually coming from

A Stanford and Georgetown preprint studied 120 Facebook Pages that posted at least 50 AI-generated images each. The sampled pages collectively drew hundreds of millions of engagements; the study did not estimate all AI content on Facebook.¹²

A widely repeated claim holds that roughly 40 percent of videos recommended to children are AI slop. No accessible methodology sits behind that number. Kapwing separately identified AI-slop channels among the top 100 trending channels in each country and combined third-party view and subscriber estimates; that is a commercial trending-channel analysis, not a census of the 15,000 most popular channels or all YouTube views.³

AI-generated guides and reviews can create real risk, but the alarming specifics that travel with the story; the mushroom-foraging guides, the 26,000 listings, the 400-percent growth, AutoBait, the NewsGuard tallies; none of them resolve to a primary method when you go looking.

Then there's the industrial version: content farms. Cybersecurity firm DoubleVerify identified a network called "AutoBait" in March 2026, over 200 websites running templated prompts through a language model to churn out articles and images purely to harvest ad revenue. NewsGuard and Pangram Labs put the broader count at 3,006 AI content farm sites as of that same month, growing by 300 to 500 new sites monthly. This isn't a scattered problem. It's a manufacturing pipeline.

What still gives it away, and what doesn't anymore

Human and automated detection performance varies dramatically by generator, image type, compression, prevalence, test design, and date. The accuracy rates, false-positive rates, writing-detection scores and punctuation-frequency tells that circulate all contradict each other, which is why none of them can be combined into a single detection rule you could actually apply.

Text is a different fight, and people lose it differently. After a five-minute conversation, participants misidentified GPT-4o's writing as human 77 percent of the time. Academics separating real research abstracts from AI-written ones scored 44 to 76 percent depending on the field. What holds up as a tell isn't any single word, it's density and rhythm. A short, boring list of stock words, the kind you'd never actually say out loud, turns up constantly in model output and almost never in human writing. One phrase alone is a giveaway: "it's not just X, it's Y" shows up more than a thousand times as often in LLM writing as in real prose. One em dash proves nothing, but fifteen in a 600-word piece, paired with uniform sentence length and a transition word opening every paragraph, is a pattern worth trusting.

The tell was never the flaw. It's the flatness, writing with no friction, images with no accident, content built to clear a threshold rather than say something.

The labels are arriving, actually use them

The infrastructure to settle this argument is finally shipping. Google's SynthID is embedded by default in every Imagen image and every Veo video, and it survives cropping, resizing, recompression, and format conversion. C2PA Content Credentials, backed by Adobe, Microsoft, OpenAI, Meta, and most major camera makers, attach a cryptographically signed provenance record to a file's entire edit history. Meta has attached "AI Info" and "Made with AI" labels across Instagram and Facebook off that metadata since early 2024. YouTube went further this May: a video carrying C2PA metadata marking it fully generative gets a permanent AI disclosure label in YouTube Studio. the creator can't toggle it off, and the usual appeals path doesn't apply. CEO Neal Mohan named reducing slop and catching deepfakes platform priorities back in January.

Use provenance metadata and platform labels when present, then check the original source and publication history, and ask who is behind it and what they stand to gain. Absence of a badge does not prove human origin; presence of AI assistance does not prove low quality.¹²³

• Check for a Content Credentials or "AI Info" badge before trusting an image's origin story, don't rely on your eyes alone.
• Read for rhythm, not vocabulary: uniform sentence length, a transition word starting every paragraph, and zero specific detail are worse tells than any banned word list.
• Follow the incentive. Strip the face and the name, leave only a caption angling for engagement, and what is left is a business model wearing a creator's face.

The slop isn't going away, the pipeline generating it is getting cheaper by the month. But the provenance tools built to counter it are real, they're shipping now, and they work better than squinting at a photo ever will. Use them.