Complete Guide11 min readJun 30, 2026

How to Spot AI Slop

The complete 2026 field guide to telling whether something is AI slop — across images, video, and writing. The signals that give it away, the tools the pros use, and where slop hides.

3 mediums covered7 sources citedUpdated Jun 2026

“How can I tell if this is AI?” is the question of 2026. It has a good answer — but not the one most people expect. There is no magic button. Instead, AI slop leaves behind consistent fingerprints, and they differ by medium. This guide is the master map: the tells for images, video, and writing, the tools professionals actually use, and the platforms where slop concentrates. Each section links to a deeper guide for the platform that needs it.

Start with the text

The fastest check you can run right now: paste any suspicious caption, comment, or article into our detector for an instant, explainable read on whether it reads like slop — in your browser, no sign-up.

Open the Slop Detector →

1First, What “Slop” Actually Means

Merriam-Webster made it official: slop was the 2025 Word of the Year, defined as “digital content of low quality that is produced usually in quantity by means of artificial intelligence.” The word traveled from “soft mud” in the 1700s to “pig slop” in the 1800s to today's meaning: machine-made filler produced at volume.

The important word is quality, not origin. That distinction is the whole game. Chasing “was a machine involved?” leads to unreliable AI-origin detectors and false accusations against real writers. Asking “does this say anything, is it specific, is it verifiable?” is more honest and more useful — because slop is slop whether a human or a model produced it.

“Slop is content made to fill space and farm attention — not to inform. Judge it by what it says, not by who typed it.”

2Spotting AI Slop in Images

Image models render surfaces beautifully but still fail at logic and detail. Zoom in — that is where the tells live.

1

Broken anatomy

Hands with extra or fused fingers, wrong number of teeth, ears and glasses that melt into skin. AI struggles most with how body parts connect.

2

Gibberish text

Any sign, label, or logo in the image dissolves into warped letterforms on zoom. Still one of the more reliable image tells, though newer models are improving.

3

Physics that doesn't work

Light with no source, reflections that don't match, water flowing wrong, shadows pointing in impossible directions.

4

Regular where it should be messy

As BBC Verify puts it: patterns are 'regular where you'd expect disorder, or distorted where you'd expect them to be regular.' Skin too smooth, crowds eerily uniform.

Facebook is the epicenter of image slop — from Shrimp Jesus to fake “made it myself” posts. See the platform-specific playbook: Facebook AI Slop: the 7 signs.

3Spotting AI Slop in Video

Video adds time, and time multiplies the errors. Watch for tells that only appear in motion.

1

Lip-sync drift

Mouth movements that don't match the words — one of the clearest signs of an AI 'news anchor' or deepfake. BBC Verify checks audio waveform and lip-sync together.

2

Morphing objects

Items that change shape, count, or color between frames. Backgrounds that shift when nothing should move.

3

Uncanny motion

Gliding walks, blinking that's too regular or absent, hands that pass through objects. Physics breaks down over a few seconds.

4

No traceable source

A 'reporter' or channel with no verifiable identity, stock-sounding AI voiceover, and a firehose of near-identical uploads.

YouTube Shorts and kids' videos are the worst-hit formats. Deeper guide: How to spot AI slop on YouTube.

4Spotting AI Slop in Writing

Text slop is the hardest to eyeball because it reads fluently. The tell is not grammar — it is emptiness: lots of words, no information.

1

Hollow fluency

Grammatically perfect sentences that say nothing. You finish a paragraph unable to state what you learned.

2

Filler phrases

'In today's fast-paced world,' 'it's important to note,' 'a testament to' — connective tissue with no substance. A pile-up of these is a red flag.

3

Vague, unverifiable claims

'Studies show,' 'many experts agree,' 'over 95%' with no named source, date, or link. Slop gestures at authority it doesn't have.

4

Hedged, symmetrical everything

Every point balanced by a counter-point, no stance taken, no specific example, no first-hand detail. Safe, smooth, and useless.

This is exactly what our detector measures

Rather than guessing whether a machine wrote it, paste the text and get a 0–100 slop score across five dimensions — vocabulary, cliché, structure, diversity, and substance — with the exact phrases that dragged the score down highlighted.

Score a piece of text →

5The Tools & Methods Pros Use

Professional fact-checkers rarely rely on a single “is-it-AI” button. BBC Verify combines several methods, and you can borrow them:

1

Reverse image / video search

Paste the image or a keyframe into a reverse search to find earlier appearances, original context, or debunks. Often the fastest disproof.

2

Geolocation & cross-checks

Match landmarks, terrain, and signage against satellite and street-level imagery. If the place can't exist, neither can the image.

3

Frame-by-frame & audio analysis

Step through video for splice points and lighting mismatches; listen for waveform anomalies and lip-sync drift.

4

Probability scores — with caution

AI-origin detectors output a likelihood, not a verdict, and false-positive on real work. Use them as one input, never the ruling.

Coming soon

One-click image & video slop check

Today our detector reads text. We're weighing an image/video checker that automates the visual tells above. Want it?

Vote for it →

7Frequently Asked Questions

How can you tell if something is AI-generated?

No single test is foolproof, but signals cluster by medium. Images: broken hands, melted objects, gibberish text. Video: lip-sync drift, morphing objects, impossible physics. Writing: hollow fluency, filler phrases, vague claims. Two or more together is a strong indication.

What is AI slop, exactly?

Merriam-Webster (which named 'slop' its 2025 Word of the Year) defines it as 'digital content of low quality that is produced usually in quantity by means of artificial intelligence.' The point is quality and intent, not origin.

Are AI detectors accurate?

AI-origin detectors give probability scores, not proof, and can false-positive on human writing. Judging content by quality — is it specific, verifiable, substantive — is often more useful than asking whether a machine was involved.

Where does AI slop show up most?

Wherever algorithms replace editors: Facebook feeds, YouTube Shorts and kids' videos, Amazon listings, and search-bait. Google filters most slop from search, so social feeds are the biggest exposure.

Sources

Catch slop without leaving the page

Install the free SlopDetector Chrome extension — select any text, right-click, and get an instant slop score. 100% local, your text never leaves your browser.

+ Add to Chrome

Stop Guessing. Check It.

The visual tells take practice. The text check takes one paste. Drop any caption, comment, or article into the Slop Detector for an instant, explainable read.