Slop Score
The check is a list of text patterns running in your browser, so nothing you paste has to leave your machine to get a score. It marks the habits it can match and adds them up. It cannot tell you who wrote the text or whether the writing is any good, so when you want that, you can hand the same passage to Hemmingway-1 and ask.
0/100
Higher means more habits found. 0 means nothing matched.
Paste your text and press Check.
In the marginThis is the one action that sends your text to the model.
Hemmingway-1's note
The model wrote this. The score and the margin notes come from the pattern list.
The score, the marks and the margin notes are worked out in your browser. Your text stays on your machine until you press Ask Hemmingway-1, which is the only thing on this page that sends it anywhere.
Built by @ppramanik62
What the score means
It is a count of habits, adjusted for length. The check runs 12 patterns over your text, counts what each one matches, weighs the more noticeable habits more heavily, and divides by your word count. Higher is worse. Zero means nothing matched. Seventy and up means the habits sit close enough together that a reader feels them before they can say why.
Dividing by length is what makes the number worth comparing. A 2,000-word article and a 150-word email with the same habits per paragraph land in roughly the same place. It also means the number is unreliable on very short text: a two-sentence note that opens with “However” and then “Moreover” scores 23, which says more about the floor the check divides by than about the note.
The 12 habits are: “Staged opener”, “Cheerful closer”, “Not X, but Y”, “Hedged preamble”, “Vague importance”, “Politeness filler”, “Emoji bullets”, “Stock AI words”, “Stacked transitions”, “Rule of three”, “Bold labels”, “Dashes everywhere”.
What the score is not
It is not a judgement of your writing. A clear, useful paragraph can pick up points for one stock word. A paragraph that says nothing at all can score zero by avoiding every pattern on the list. The check reads patterns. It cannot read what you meant.
That is the part Hemmingway-1 does. The model reads the same passage and can say what a pattern never could: that your second paragraph repeats your first, or that you never say who any of this is for. The score and the marks tell you where to look, and Hemmingway-1's note tells you what is actually wrong.
The check runs on your machine
The 12 patterns ship with this page as a few kilobytes of JavaScript. When you press Check, your text is matched against them in your own browser. Nothing is uploaded, and once the page has loaded you can pull the network cable and the score still works. That is worth knowing if the thing you want to check is a draft you would rather not paste into a website.
Hemmingway-1 is the part that cannot work that way, because a 27B model does not run in a browser tab. Pressing Ask Hemmingway-1 sends the passage you checked to our Cloudflare Worker, which passes it to Hemmingway's API and streams the answer back. That is the only request this page makes, and it only happens when you press that button. See our privacy policy.
This is not an AI detector
It cannot tell you who wrote a passage, and you should be careful with anything that claims it can. Every habit on the list is one a person can write and a model can avoid. Somebody who learned to write business email in 2010 will trip four of them in a paragraph. A model told to write plainly will trip none.
A high score means your text has habits worth editing out. It does not mean a machine wrote it. We ran the check over How to Write with AI Without Losing Your Voice, an article on this site written by hand. Across its 869 words of prose, not one pattern matched. That is the result we tuned for, and it is also the reason the check cannot be used the other way round.
What to do with a bad score
Work down the marked passages and edit them yourself. Each mark has a note in the margin level with it, naming the habit and saying what it costs you, so the page reads like a proof somebody has already been over.
If you would rather see a whole rewrite first, the Humanizer hands your draft to Hemmingway-1 and asks it to say the same thing the way a person would. It is the same model that writes the note on this page, and both tools are on our free writing tools page. What is Hemmingway-1? covers what the model is and how its makers' benchmark claims hold up.
Check your drafts where you write them
A pasted passage is all this page can see. StashBase is an open-source, local-first writing workspace where you write with Claude Code, Codex or the built-in Default Agent, each able to draft with the Hemmingway writing skill. Your draft, the notes it has to agree with and your own past writing sit in one folder beside the chat, so an edit can be checked against the thing it is supposed to say.
Frequently asked questions
Is the Slop Score an AI detector?
No. It matches text patterns and nothing else, so it cannot tell you who or what wrote a passage. Every habit it looks for is one a person can write and a model can avoid. A high score means the text has habits worth editing out, and that is the only thing it means. We do not publish a “% AI” figure because we do not think an honest one exists.
How is the score worked out?
The check runs 12 patterns over your text, counts what each matches, weighs the more noticeable habits more heavily and divides by your word count, so a long piece is not punished for its length. Higher is worse. The patterns come from Wikipedia's Signs of AI writing.
Does my text leave my browser?
Not for the score. The patterns, the marks and the margin notes are worked out in your own browser and nothing is uploaded, so you can check a draft you would not want to paste into a website. The one exception is the note: pressing Ask Hemmingway-1 sends the passage to our Cloudflare Worker, which passes it to Hemmingway's API. The Worker does not log or store it, and Hemmingway's privacy policy says API requests are not kept. Everything on the page is gone when you close or reload the tab. See our privacy policy.
Why is the model a separate button?
So that nothing you paste goes over the network unless you decided it should, which also means you can run the check on anything without thinking about it first. The other reason is cost. The check is free to run as often as you like because it runs on your machine, while every note is a call to a model we pay for. Keeping the two apart is what lets the scoring stay unlimited.
Is a high score always bad?
No. The check is looking for habits, not mistakes. A dash, a “however” or a group of three can be the right call, and the score cannot tell the difference. Read the marked passages and decide each one. The number is a place to start looking, not a verdict on the sentence.
Which model writes the note?
Hemmingway-1, a 27B open-weight model fine-tuned from Qwen3.8-27B to write text that reads as if a person wrote it. The weights are Apache-2.0 on Hugging Face. You can also chat with Hemmingway-1 on this site, and What is Hemmingway-1? reads its makers' benchmark claims independently.
Is it free?
Yes. There is no account, no API key and no quota. The check itself is free to run as often as you like because it runs on your own machine. One note takes up to 6,000 characters, about 1,000 words, and the page limits how many notes one visitor can ask for in a minute.
What should I do with a bad score?
Edit the marked passages. Every mark has a note in the margin beside it naming the habit and saying what it costs you, so you can work down the page without looking anything up. If you would rather start from a rewrite, the Humanizer gives your draft to the same model and asks it to say the same thing the way a person would.