FREE TOOL · NOTHING UPLOADED

Slop Score

Paste any text. Get a 0–100 AI slop score and a category breakdown, checked against the 200+ rules in the AI slop words and phrases list. Everything runs in your browser — your text is never sent anywhere.

0 words
–SLOP
Paste a draft
Tells are scored by density, so length is accounted for.

Was a hit wrong?

False positives are the most useful thing you can send. Tell us which rule fired on writing that was fine, and it gets fixed in the rule set and on the public list.

Now fix what it found

Slop Score tells you what is wrong. The enhancer rewrites it: every hit underlined in your text with a one-click plain replacement, and a de-slop rewrite that invents no facts.

Open the enhancer — free, no signup →

How the score works

The score is a density measure, not a count. Every tell carries a weight, the weights are added, and the total is divided by the length of the text. A 2,000-word essay with eight tells scores far lower than a 100-word post with eight. That is deliberate: density is what readers actually notice, and a long piece has more room to earn a few.

Weights follow precision — how rarely a human writer produces the pattern by accident.

CategoryWeightWhy
Machine artifacts25No human cause. One hit is proof, so the score floors at 60.
Marketing templates10Full templates a human editor almost always cuts
Sentence patterns8Superficial -ing tails, speculation tails, weasel attribution
Structural patterns6–10Em-dash density, negation frames, closing aphorisms
Tell-phrases6Common in AI, uncommon in edited prose
Brochure register3Praise without observation
Tell-words3Real words. Only density makes them a signal.
Crutches, hyphen pairs, filler2Weak alone, meaningful in bulk

What the bands mean

ScoreGradeWhat it means
0–9CleanReads as human. Nothing to fix.
10–24LightA few habits. Most published writing sits here.
25–44NoticeableA careful reader senses the pattern.
45–69HeavyReads as machine-written to most people.
70–100SlopUnedited model output.

What a high score does and does not prove

A high score means the writing carries patterns readers associate with AI. That is a style problem whether or not a model wrote it. Plenty of human marketing copy scores in the 40s, and it should — the reader's reaction is the same either way.

So do not use this to accuse anyone. Every word-level and structural rule fires on genuine human writing too, which is exactly why the score is density-based rather than a yes/no verdict.

One category is different. Machine artifacts — unfilled [Your Name] placeholders, leaked citeturn0search0 tokens, utm_source=chatgpt.com links, zero-width characters, homoglyphs — have no human cause. A writer has no reason to produce them. Those are marked certain and floor the score at 60 on their own.

Is my text uploaded anywhere?

No. All 200+ rules are JavaScript in this page. Nothing is sent to a server, nothing is logged, and the tool works with your network off once the page has loaded. Paste unpublished drafts and client work without worrying about it.

How to lower your score

Work down by weight, because the top of the list is where the points are.

  1. Machine artifacts. Search for [, utm_source, and cite. Each fix removes 25 points and one certainty.
  2. Marketing templates and tell-phrases. Delete them outright. Do not rewrite them. If the sentence still reads, the phrase was slop.
  3. Structural patterns. Replace em dashes with commas and periods. Cut forced triplets to their real count. Drop "it's not X, it's Y" and state Y. Delete the closing aphorism.
  4. Word swaps. Last, because they are the smallest part of the score. Leverage becomes use.
  5. Add one specific detail only you could know. No tool can score this, and it is what makes writing yours.

The full list has the plain replacement for every rule here. For a worked before/after on a real draft, see the five-step fix.

Related