Dana OkaforHead of Growth at Hollowpine · 2h
Free Chrome extension
How much of this is slop?
Slopgauge measures every post you scroll past and names the phrases that gave it away. Edit the post below and watch them drain into the gauge.
No account. Nothing you read leaves your browser.
Tells drained0
In your feed
It reads along with you.
Every post gets a meter as you scroll LinkedIn, X, Reddit and the rest. These three were scored just now, on this page, by the same code the extension runs.
- A meter under every post.Five levels, from No slop to Pure slop. Nothing to click.
- The evidence, on hover.Hover a meter to see each tell, the exact phrase, and why it reads as generated.
- Dim the worst of it.Turn on Dim heavy slop and the sloppiest posts fade back, like the third one here. Hover to read it anyway.
- Deep check, if you want more.A larger neural detector that runs on your own computer. Off until you turn it on.
Five levels
Drag the gauge.
Five posts, measured just now on this page. Pull the level up or down, or tap a mark, to see what each reading looks like.
Drag the level, or use the arrow keys.Drag the level, or tap a mark.
The tells
Every tell has a name.
A score alone wouldn’t tell you much. So Slopgauge names each habit it finds, about thirty in all, and a model trained on 120,000 texts weighs them with word choice and rhythm. Here is one paragraph of pure slop, run through it.
In today’s fast-paced world, it’s not just about working harder. It’s about working smarter. Here’s the thing: success isn’t a destination. Let’s delve into the three pillars that make it possible: clarity, consistency, and community. The result? Growth. This serves as a testament to the power of showing up. I hope this helps! Agree? Drop a comment below.
Also on the list: negative listings, dramatic fragments, fake-profound kickers, rule-of-three lists, em dash pileups, emoji and arrow bullets, weasel attribution and pasted markdown.
How sure is it
It gets it wrong sometimes. Here’s how often.
The essay checker on 2,999 school essays by students in grades 6 to 12, and 2,000 essays by AI on the same prompts. No detector can prove who wrote a text, and this one doesn’t pretend to.
100 essays by students
1
was wrongly called Likely AI-written. The checker sets this line separately for each kind of text, so it stays near 1 in 100.
100 essays by AI
72
were called Likely AI-written. The other 28 got through: we chose fewer false accusations over catching more.
Half by a student, half by AI
44%
of the AI half was marked, and 2% of the student’s half.
A high reading is a reason to look closer, never proof on its own. Short texts are the hardest: under a few hundred words, results are less reliable.
The essay checker
Got a whole essay? Pour it in.
It marks every sentence that reads as AI-written, measures how much of the text does, and tells you how unusual that is for that kind of writing. Paste something the student wrote before, and it checks whether each part sounds like them. Or drop in a whole class and see who to read first.
Works on
LinkedIn, X, Reddit, Facebook, Instagram, Threads, Bluesky, TikTok, YouTube comments, Quora, Gmail, Outlook, Slack, Discord, Hacker News, GitHub, Google Docs, forums, and articles on most sites.
Anywhere else, select text and right-click “Check with Slopgauge”, or paste it into the toolbar popup.
Questions
Good to know.
Can Slopgauge prove that someone used AI?
No, and no detector can. A high reading means a text shares habits that are common in AI-written text. About 1 in 100 school essays written by students is still called “Likely AI-written”. Treat a reading as a reason to look closer or to have a conversation, never as proof on its own.
How accurate is it?
On 2,999 essays by students in grades 6 to 12, the essay checker called 1 in 100 “Likely AI-written” and caught 72 of 100 essays written by AI on the same prompts. Short texts are harder: under a few hundred words, results are less reliable. See the test results.
Is my text uploaded anywhere?
No. The extension and the essay checker analyze text in your browser and send it nowhere. The one exception is Google Docs, which draws its text on a canvas: Slopgauge downloads the open doc as plain text from Google, the same as File > Download, and reads it locally. Read the privacy policy.
Is it free?
Yes. The extension and the essay checker are free, with no account, no ads and no trial.
Which browsers does it work in?
The extension works in Chrome, and in browsers that install extensions from the Chrome Web Store, such as Microsoft Edge and Brave. The essay checker works in any modern browser, on a computer or a phone.
What counts as slop?
Text that reads as generated rather than written: stock openers like “In today’s fast-paced world”, “It’s not X, it’s Y” contrasts, rule-of-three lists, self-answered questions, and posts that end by fishing for comments. Slopgauge names about thirty of these habits and weighs them with word choice and sentence rhythm.
I’m a teacher. How should I use it?
Paste or drop an essay into the essay checker, pick the kind of text, and read the marked sentences. Paste something the student wrote in class to check whether each part sounds like them. Check a class measures a whole assignment at once and groups essays that read as AI and share unusual wording.
Then talk with the student. A reading is where a conversation starts, not a verdict.
Why was my own writing flagged?
Formal, formulaic writing can share AI’s habits: stock transitions, very even sentences, a five-paragraph shape. That is the 1 in 100. Short texts are flagged more often too. The named tells show exactly which phrases tipped the reading.
About
Made by a student who wants the internet to sound human again.
Slopgauge is an independent project by a student at Princeton University. It started with a simple frustration: feeds, inboxes and group chats were filling up with posts nobody had really written, and there was no quick, honest way to tell.
The detectors that existed uploaded your text, returned a single number and never said why. So Slopgauge works the other way around. It runs on your own device, it shows the evidence behind every reading, and it publishes how often it gets things wrong.
It’s built out of genuine passion for the problem, and it’s free: no ads, no accounts and no data to sell. Ideas, bug reports and readings that look wrong are always welcome at support@slopgauge.com.
SlopgaugeAn independent project from a Princeton University student
Built on open research
Slopgauge stands on public datasets and models shared by researchers:
- The style modelTrained on about 120,000 texts from openly licensed research corpora: DACTYL, MAGA, LLMTrace, LLMDetect, Ghostbuster and TweetEval.
- Deep checkVanguard, an open neural AI-text detector, run on your own computer.
- Voice comparisonLUAR, from Learning Universal Authorship Representations (Rivera-Soto et al., EMNLP 2021).
- Accuracy testingPERSUADE 2.0, a corpus of essays by students in grades 6 to 12, used only to measure accuracy and set thresholds.
- Named tellsAdapted from no-ai-slop by Peter Yang (MIT license).
Shows its evidence.
Every reading names the phrases behind it, so you can judge it yourself.
Runs on your device.
Nothing you read or paste is uploaded. There are no servers to send it to.
Honest about mistakes.
The essay checker is tuned so that about 1 in 100 texts by people is called “Likely AI-written”, and this site says so up front.
Words of what you read sent anywhere.
Slopgauge measures text on your device and sends it nowhere. No servers, no account, no analytics. The gauges on this page run in your browser too.
The one exception is Google Docs, which draws its text on a canvas: Slopgauge downloads the open doc as plain text from Google, the same as File > Download, and reads it locally.