slopcheck: open source slop detection

What this tool does

Paste an article, a draft, or any prose. The tool applies the open source synthesis engineering skill family (v4.0 content-quality and v2.0 fact-checking) and returns a structured analysis covering two axes:

AI provenance signals. Which model-family patterns triggered, with examples from your content. Family attribution where discernible (Claude, GPT, Gemini, Llama, Grok, DeepSeek, Mistral, Qwen).
Slop independence. Whether your content carries substance and depth, regardless of how it was authored. Good AI-collaborated content passes; empty content fails regardless of provenance.

The skill family is open source. Anyone can install it locally in their own AI agent for free. This hosted convenience tool is for people who want to try the system without installing anything.

Who is this for

Writers

Pre-publish gut-check. Run a draft through and see what an editor's pattern-trained eye would flag before you hit publish.

Journalists

Screen wire copy, freelancer submissions, and pitches. Identify which AI-family signature you are looking at, with empirical anchors per family.

Newsroom editors

Forensic review of older articles, current submissions, AI-collaborated drafts. The catalog spans the entire LLM era, not just current frontier models.

Marketers and execs on LinkedIn

Avoid posting cringe slop. Check a draft post for the corporate-vocabulary, em-dash-heavy, no-specifics patterns that undermine your credibility.

Researchers and academics

An open-source, year-stratified, family-attributed catalog of AI-generation patterns. Compounding-archive design means it stays useful as models evolve.

Engineers

Install the skills locally in Claude Code, Codex, Cursor, or any AI agent. Use the CLI in CI. Audit the open source code.

Privacy: what we don't collect

We don't store your content. Not on the BYOK path (your browser talks directly to the LLM provider you choose). Not on the hosted-tier path (the Pages Function streams the call and forgets it).
We don't store your API key. On the BYOK path it lives only in your browser. We never see it.
We don't store your IP address. The hosted tier uses an HMAC-SHA256 hash of your IP (with a salt we rotate) as a daily rate-limit counter. The raw IP is never written to storage.
We don't use cookies, analytics, telemetry, or trackers. No Google Analytics, no Mixpanel, no third-party tags. No "view counts." No fingerprinting.
We don't have user accounts. No login. No email. No password. No subscription.
We don't share with third parties. Your content is sent only to the LLM provider you choose, for the time it takes to compute the response.
We use Cloudflare Turnstile for bot resistance on the hosted tier (a privacy-preserving captcha alternative; no Google reCAPTCHA). BYOK skips Turnstile.
Everything is open source. Audit the frontend at github.com/synthesisengineering/synthesis-slopcheck and the hosted-tier Pages Function in the same repo's functions/ directory.

Choose how to run it

Three ways to use slopcheck. All free. Pick the one that fits.

Bring your own API key Recommended

Free. You pay your provider's standard API rate (≈ $0.05 to $0.45 per analysis depending on model and document length).
No daily cap. Larger documents allowed.
Access to the best models: Claude Opus 4.7, GPT-5.5 Pro, Gemini 3.1 Pro.
Faster: no shared queue.
Your key stays in your browser. Nothing is sent to our servers.

Use Rajiv's keys (free, limited)

No API key needed. Click and go.
5 analyses per day from your IP address.
Documents up to 50,000 characters.
Cost-efficient models only: Claude Haiku 4.5, GPT-5.4 Mini, Gemini 3 Flash Lite.
If the daily community budget runs out, hosted tier rests until tomorrow.

Install locally

Completely free. No caps. Runs in your AI agent (Claude Code, Codex, Cursor, etc.).
Best for power users and teams.
Install instructions.

Why bring your own key?

The hosted tier is free. The API calls it makes are not — they cost Rajiv money. Each hosted-tier analysis costs roughly $0.05 to $0.20 in API fees.

If you can afford an API key, please use it. A $5 minimum credit at any provider gets you around 50 analyses, the tool is faster, you can use the best models, and you free up the community budget for people who can't.

No login. No subscription. Just a free API key from your chosen provider, used directly from your browser.

If the tool is genuinely useful to you, you can support the open-source work on GitHub Sponsors. One-time or recurring, any amount, voluntary. Not a gate on anything.

Setup

Provider

Model

Your API key

Remember key in this browser (uses localStorage; persists across sessions until cleared)

How do I get an API key?

Anthropic (Claude): console.anthropic.com/settings/keys. Create an API key, then come back and paste it here.
OpenAI: platform.openai.com/api-keys.
Google (Gemini): aistudio.google.com/apikey.

Your key stays in your browser. This page does not send it to any server.

Analyze your content

Content to analyze

0 characters

Detector mode Artifact mode (default; the content is the produced artifact, not a chat transcript) Full-response mode (includes the conversational wrapper from a chat transcript)

All the ways to use slopcheck

Pick whichever fits how you work. Every path is open source and free.

Try it inside your chat tool

Slopcheck GPT in the OpenAI GPT Store. Use slopcheck directly inside ChatGPT. Account: free ChatGPT.
Slopcheck Claude Project. Use slopcheck directly inside Claude.ai with the full methodology loaded. Account: free or Pro Claude.ai.
Prompt template for any chatbot with web fetch. Copy the prompt template and paste it into ChatGPT, Claude, Gemini, or Perplexity along with your content.

Browser-side, no install

This web app. The page you are on. BYOK or hosted tier. No account.
Bookmarklet: drag this link to your bookmarks bar: slopcheck this page . Click it on any page to open slopcheck with the selected text (or page content) pre-filled.
Chrome / Edge extension: right-click any selection → "Slopcheck this." (See the extensions/chrome source.)
Apple Shortcuts: share-sheet integration on iPhone, iPad, and Mac. (See platforms/apple-shortcuts.md.)

Command line and packages

Homebrew (macOS / Linux): brew install synthesisengineering/tap/slopcheck
npm: npm install -g @synthesisengineering/slopcheck
bun: bun add -g @synthesisengineering/slopcheck
AUR (Arch Linux): paru -S slopcheck
curl (any POSIX): curl -fsSL https://synthesisengineering.org/install.sh | sh
Python CLI direct: clone the repo, run cli/slopcheck.py (no dependencies, Python 3 stdlib only).

For engineers and teams

Install the skills in your AI agent (Claude Code, Codex, Cursor, etc.) for free. See install instructions.
Self-host: clone synthesis-slopcheck and deploy your own copy. Cloudflare Pages is the reference host.
Slack / Discord bot scaffold: self-host instructions for newsrooms and teams. (Scaffolded in the repo.)
GitHub Action (planned, post-launch): run slopcheck on a PR diff and post a review comment.

Support the open-source work

Slopcheck and the synthesis engineering skills are open source and free. If they are genuinely useful to you, supporting Rajiv on GitHub Sponsors helps keep the hosted tier running and the open-source work going. One-time or recurring, any amount you choose. Voluntary. Not a gate on anything.