Does this read like a chatbot wrote it?
Paste your text. Em dashes, “it's not X, it's Y”, lists of three, “delve” and a dozen other AI writing habits get underlined, and you get a slop score. It all runs in your browser.
Rewrite it with a local model
Erik Markus Kannike trained these habits out of Qwen3-4B and released it as qwen3-4b-deslop. If you run it with Ollama, this page can send it your flagged paragraphs to rewrite. Your text goes from this tab to your own computer and nowhere else.
- Get the model (2.5 GB):
- Start Ollama so it accepts requests from this page. Quit the Ollama app first, then:
- Connect, then rewrite. Chrome may ask to let this site reach apps on your device; allow it.
How the score works
Each habit is a regular expression with a weight. The score is the weighted count per 1,000 words, the same scale as the detector behind qwen3-4b-deslop, where careful human writing scores under 15 and the stock Qwen3-4B averages 121. This one is an independent take on that style guide; on the 80 sample answers in that post it averages 122 before de-slopping and 7 after (theirs: 121 and 13).
A regex can't tell good writing from bad. Writers used em dashes and lists of three long before chatbots did, so treat a high score as a reason to reread, not proof that a model wrote it.
| Habit | Points | Why it's flagged |
|---|