is this AI slop? 🤖

Spot the tells of AI-written text — with zero AI.

🌐 Paste your text in the browser →  ·  nothing uploaded, runs 100% client-side.

npx is-this-ai-slop "your text here"

Everything reads like ChatGPT now. The em-dashes. The "let's delve into the realm of…". The "it's not just a tool — it's a revolution." The tapestry of robust, seamless solutions. You can feel it, but you can't always point at it.

is-this-ai-slop points at it. Paste any text and it underlines every tell that makes writing read as AI — and the funny part: it uses no AI at all. It's a curated, rule-based database of the clichés, overused words, and structural tics LLMs lean on. Deterministic, instant, 100% local.

$ npx is-this-ai-slop email.txt

  In today's fast-paced world, it's worth noting that effective communication is
  paramount. Let's delve into the realm of writing and unlock the potential…
        ‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾   ‾‾‾‾‾‾‾‾‾‾‾‾‾    ‾‾‾‾‾  ‾‾‾‾‾‾‾‾‾‾   ‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾

  100/100 ████████████████████ Heavy AI slop 🤖
  18 AI tells in 83 words

  Tells
    delve into · cliché phrase
    in today's fast-paced world · cliché phrase
    a testament to · cliché phrase
    elevate · AI-favorite word
    …
  → delete those to sound human.

It's a de-slopper, not a polygraph

It does not claim to "prove" anything was AI-generated — that's snake oil, and humans write clichés too. What it does is honest and useful: it flags the clichés that read as AI, so you can cut them and sound like yourself again. A clean text earns a ✋ wrote-this-myself badge.

What it catches

| Tell | Examples | | ---- | -------- | | Cliché phrases | "in today's fast-paced world", "it's worth noting", "delve into", "unlock the potential", "a testament to", "harness the power" | | AI-favorite words | delve, tapestry, boasts, robust, seamless, elevate, leverage, realm, meticulous, plethora, bustling | | The em-dash habit | the ChatGPT love of — everywhere (the first one's free) | | "Not just X, it's Y" | "this isn't just a tool, it's a revolution", "not only fast but also cheap" | | Chatbot pleasantries | "Certainly!", "Great question", "Let's dive in", "Buckle up" | | Tidy closers & emoji | "Ultimately,", "In essence,", and decorative 🚀✨ in prose |

Each tell is weighted and the text gets a 0–100 slop score (length-normalized, so a long clean essay isn't punished and a short sloppy one isn't let off).

Use it

npx is-this-ai-slop "paste text inline"
npx is-this-ai-slop README.md          # …or a file
cat post.md | npx is-this-ai-slop      # …or stdin

npm i -g is-this-ai-slop               # the bin is also `slop`

De-slop your docs in CI — fail the build if the README reads like a robot wrote it:

# .github/workflows/ci.yml
- run: npx is-this-ai-slop README.md --max-score 30

is-this-ai-slop post.md --json out.json     # machine-readable
is-this-ai-slop post.md --md  out.md        # paste into a PR
is-this-ai-slop post.md --ignore emoji      # mute a category

Library API

The engine is pure and browser-safe (it's literally what the playground runs):

import { analyze, toSegments } from "is-this-ai-slop";

const r = analyze(text);
r.verdict.score;        // 0–100
r.verdict.humanBadge;   // earned the ✋?
r.findings;             // every tell, with { match, start, end, category, note }
toSegments(text, r.findings);   // plain + flagged segments, for highlighting

Honest about the limits

This is a cliché detector, not an AI detector. Heavy human editing fools it, and a careful AI prompt evades it. It's deliberately built to be a writing tool — "here are the N phrases that make this read as AI; cut them" — not a gatekeeper. (Please don't use it to accuse a student. It will be wrong.)

Roadmap

• Per-language tell sets (the tells are very English-specific today).
• A "rewrite suggestion" for the worst offenders.
• A browser extension to score any text box as you type.
• Community-tunable tell weights.

💖 Sponsor

Free, MIT, built in spare time. If it de-slopped your writing:

• ⭐ Star the repo.
• 🍋 Sponsor via Lemon Squeezy — one-time or recurring.

License

MIT © is-this-ai-slop contributors