無料のAIスロップ 検出器
読む価値があるコンテンツかどうかを判定

AI slop is low-effort, mass-produced content — text, images, or video — generated at scale and published for engagement rather than value. It reads as plausible but says little: vague, templated, interchangeable. The term exploded in 2024 and became a Word of the Year contender in 2025.

AI detectors guess who wrote a text, and they fail often — flagging the Gettysburg Address as machine-made. Slop detection asks a different question: is this worth reading? A tired human listicle can score high. A carefully edited AI draft can score low. We judge quality, not origin.

It crowds out human work, pollutes search results and feeds, erodes trust in what we read, and at its worst spreads dangerous misinformation — fake health advice, fabricated news, scam funnels. It also trains future models on its own output, degrading them.

Deterministic heuristics, fully in your browser: a tiered dictionary of 300+ overused words, a cliché-phrase list, structure signals (transition openers, sentence-length uniformity, em dash density, rule-of-three), lexical diversity, and substance measures. No AI, no API calls, no data sent anywhere.

Yes — and we say so plainly. A draft generated by a model and then edited by a knowledgeable human who adds specifics, voice, and lived experience can absolutely clear the bar. The slop is what happens when nobody does that work.

Overused words ("delve", "tapestry", "pivotal"), prefab phrases ("in today's rapidly evolving world"), uniform sentence rhythm, em dashes everywhere, rule-of-three constructions, hedging instead of specifics, and zero lived experience. Two or more at once is the real tell.

"Slop" is what you feed to pigs — cheap, bulk, low-nutrition filler. Applied to AI content, it describes output produced at scale for quantity, not quality. The metaphor stuck because it captures both the production method and the nutritional value.

Shrimp Jesus on Facebook, fake crochet patterns, AI news anchors reading fabricated stories, "experts" citing studies that do not exist, foraging books on Amazon with dangerous errors, and SEO blog posts that answer questions nobody asked. See our Examples gallery for the full tour.

Manually: read the first three sentences and count tells — prefab openers, hedged claims, no specifics. Automatically: paste the text into our detector and get a 0–100 score with a factor-by-factor explanation of what tripped it.

You just found the free way: scroll up, paste your text, hit Analyze. No account, no limits, no data leaves your browser.