How to Spot AI Slop: A Field Guide to Machine-Made Content

AI slop is machine-generated content published at volume with no human care — and you can learn to spot it in seconds, because it fails the same three tests every time: no substance, no source, and no reason to exist. The internet is filling with it — search results, Facebook feeds, YouTube kids’ channels, even product reviews. This is a field guide: the tells for text, images, and video, plus the honest limits of every detection trick.

First, what slop is (and isn’t)

“AI slop” doesn’t mean “anything made with AI.” A writer who drafts with a chatbot, checks every fact, and rewrites in their own voice isn’t producing slop. Slop is what happens when the human drops out of the loop: content generated in bulk, published unread, and optimized for clicks or ad pennies rather than for you.

The distinction matters because the problem isn’t the tool — it’s the absence of anyone who cares whether the thing is true or useful.

Spotting slop in text

Written slop has a rhythm you can learn to hear:

• The formula loop. Make a claim, restate it slightly, move on. Paragraph after paragraph that gestures at depth without ever landing a specific fact, number, or example.
• Nothing checkable. No names, no dates, no prices, no sources. “Many experts agree” and “studies show” with no expert and no study.
• Keyword stuffing with a thesaurus glaze. The target phrase appears in every heading, sometimes in sentences that exist only to contain it.
• The house style of nobody. Endless hedging (“it’s important to note,” “in today’s fast-paced world”), symmetrical lists of three, and a conclusion that summarizes the summary.
• It answers everything except your question. The most reliable tell: you finish the page and search again. That feeling of hollowness is the detection.

One warning: individual word-level tells (“delve,” em-dashes, certain stock phrases) are weak evidence. Real humans use them, and slop factories now prompt them away. Judge the substance, not the vocabulary.

Spotting slop in images

Image generators keep improving, but volume slop is made cheap and fast, so the classic glitches persist:

• Hands, teeth, and text. Extra fingers, fused knuckles, and signage or labels that dissolve into alphabet soup up close.
• Broken physics. Melted edges where objects meet, jewelry that merges with skin, reflections that don’t match the scene.
• Too-perfect gloss. A waxy, over-lit sheen on everything, especially in fake “look what I made” craft photos and miracle-house builds.
• The recycled caption. On social media the giveaway is often the packaging: “Made it with my own hands 🙏” on the fifth impossible cake this week, from a page with no history.

Spotting slop in video

Video slop is exploding on YouTube and Facebook, and it targets the least defended audiences — kids and older relatives:

• Lip-sync drift and faces that subtly warp between frames.
• Objects that morph — a cup changes shape between cuts, background people flicker.
• The stock AI voiceover: flat, over-enunciated narration with odd emphasis, reading text that sounds like the written slop above.
• Firehose channels. Dozens of near-identical uploads a day, no host, no face, no community — a content machine, not a creator.

If your kids watch YouTube, this is worth a family conversation — our guide on how to talk to your kid about AI has a script for exactly this.

The three-question test

Medium aside, run any suspicious content through these:

1. Substance: Does it contain anything specific — a fact, a price, a firsthand detail — that could be wrong? Content that risks nothing usually knows nothing.
2. Transparency: Is there a real person or organization behind it, findable, with a history?
3. Reason to exist: Can you tell why this page was made, other than to catch a search or fill a feed?

Two or three “no” answers: it’s slop. Move on.

Honest caveats

• AI detectors are unreliable. Automated “AI content checkers” produce false positives on human writing (they flag non-native English speakers constantly) and false negatives on lightly edited AI text. Don’t use one to accuse a student, an employee, or a writer.
• The tells decay. Every glitch in this guide is being engineered away. Hands are mostly fixed; text-in-images is improving. The three-question test ages better than any visual tell.
• Human slop exists too. Content farms wrote hollow, keyword-stuffed junk long before 2023. The disease is carelessness at scale; AI just cut the cost.
• AI-assisted ≠ slop. Plenty of good work now involves AI somewhere in the pipeline — including, transparently, on blogs like this one. The question worth asking isn’t “did a machine touch this?” but “did a human stand behind it?”

Defending your own attention

You can’t fix the supply side, but you can fix your intake:

• Curate sources instead of feeds. Algorithmic feeds are where slop thrives, because it’s engineered for them. Subscribing directly to writers and sites you trust filters most of it out — that’s the whole argument in our RSS feed reader setup for the AI era.
• Reward the humans. When you find a site with a name, a face, and checkable facts, subscribe, share, pay. Slop wins on volume; humans win on loyalty.
• If you publish, be the counterexample. Specifics, sources, and a visible author are now a competitive advantage — for readers and, increasingly, for the AI assistants deciding which sites get cited.

Bottom line

Slop detection isn’t about catching robots — it’s about recognizing the absence of care. Look for substance, a person behind the words, and a reason for the page to exist. When all three are missing, close the tab. Your attention is the only thing the slop economy runs on, and it’s yours to withhold.

Related reading: RSS Is Back: How to Set Up a Feed Reader for the AI Era · How to Talk to Your Kid About AI Without a Lecture · How to Get Your Blog Cited by ChatGPT and Perplexity · Half Your Visitors Are Now Bots: How to Feed Them Without Ruining Your Site