Does it sound like AI?
Paste an email, a post or a report an AI wrote for you. See every habit that gives it away, fix the easy ones in one click, and get instructions that stop them next time. Nothing you paste leaves this browser.
Read the guide to AI words to avoid, with a plainer word for each one.
New to this? What it does, in plain words
It reads a piece of AI writing the way a sharp colleague would, and points at every phrase that sounds machine-made.
Use it before you send an AI draft to a client, a manager or your team. A sales lead can clean up a proposal in a minute. A recruiter can check that a job post does not read like every other one. A manager can make sure a note to the team sounds like them.
What you need
Nothing. No API key, no sign-up. It works with text from ChatGPT, Claude, Gemini, Copilot or anywhere else, and nothing you paste is sent anywhere.
How to use it
- Paste the text. The highlights and the score appear as you type.
- Read the list on the right. Each item says what the habit is and why it gives the text away.
- Press Apply safe fixes for the ones that cannot change your meaning: long dashes, filler openings and endings, stray bold and emoji.
- For the rest, open the fix prompt in ChatGPT or Claude, then copy the Stop it next time instructions into your AI's settings.
The text
Fix the rest with AI
A prompt that quotes each problem and asks for a rewrite that keeps every fact. Open it in the AI you use, or copy it.
Stop it next time
Standing instructions, written as what to do rather than what to avoid, which models follow more reliably. Paste them into ChatGPT’s custom instructions, Claude’s personal preferences or Gemini’s saved info.
Why AI writing gives itself away
Models learn a house style from millions of chat replies. The habits are small, but readers spot them, and they decide how much to trust what follows.
- Free, in your browser. Paste the draftAny text an AI wrote, from any AI.
- Free, in your browser. See the habitsEvery phrase that gives it away is marked, with the reason.Instant
- Free, in your browser. Fix the safe onesDashes, filler, stray bold and emoji, in one click, without touching your meaning.
- In the AI you use. Rewrite the restA prompt that quotes each problem, ready to open in ChatGPT or Claude.
- In the AI you use. Stop it next timeStanding instructions for your AI’s settings, built from what was found.
What it looks for
- Long dashes. The most talked-about sign of AI text. People type commas, colons and full stops far more often.
- Filler at the start and the end. “Certainly!”, “Great question”, “I hope this helps” and offers of more help that nobody asked for.
- AI-favorite words. Delve, leverage, robust, seamless, pivotal, tapestry, landscape and the rest, each with a plainer word where one exists.
- Patterns. “Not just X, but Y”, teaser questions such as “The result?”, and one list of three after another.
- Format. Bold on every other phrase, headings in a short email, emoji, and sentences that are all the same length.
- Your house rules. A word limit, American or British spelling, no bullet points and words your team never uses.
An AI slop checker for your own drafts
“AI slop” is what people now call text that was plainly generated and never edited: filler, hype words and the same rhythm in every paragraph. Paste your own draft here before it goes out and every one of those habits is marked, with a fix beside it. The guide to AI words to avoid lists them with plainer words.
It is not an AI detector
It never says a text is “87% AI”. Detectors guess, they are often wrong, and people write these phrases too. This tool counts habits and explains each one, so you can decide what to change. The score is simply how many it found per 100 words.
Why “stop it next time” is written the way it is
Telling a model “never use dashes” often works for a few replies and then fades. Instructions that say what to do instead, such as “use commas, colons or full stops”, hold up better, and they work best in the settings of the AI you use, where they apply to every chat. The system prompt guide explains why.
Questions
Is my text uploaded anywhere?
No. Every check runs in this browser and nothing you paste is sent to thinkhard.si or anyone else. The text you are working on is kept in this browser so it survives a reload, and Clear removes it. Only if you press Open in ChatGPT or Open in Claude does the fix prompt go to that app, because you sent it there.
Will Apply safe fixes change what I said?
It only does what cannot change your meaning: it replaces long dashes with commas or colons, removes filler openings such as “Certainly!” and chatbot sign-offs such as “I hope this helps”, strips bold marks and emoji, and corrects spelling to the variety you chose. Word swaps are left to you, because the right word depends on the sentence. Undo fixes puts everything back.
Can my team use the same rules?
Yes. Set your house rules, press Share my rules and send the link. Opening it loads the same word limit, spelling and banned words. The rules travel inside the link, after the # sign, which browsers never send to a server.
Does it work in other languages?
The phrase lists are English. Dashes, emoji, bold, headings and your own banned words work in any language.