AI SLOP CHECKER

Check your AI writing before anyone reads it

Paste your article. The checker marks unsourced numbers, vague claims, stock phrases and machine-like rhythm, each with a suggestion. Free and without sign-up, in English and German.

0 of 30000 characters

Your text is processed only for the check and is not stored.

We measured what actually gives AI texts away

We compared 819 expert articles from before ChatGPT with 80 AI articles on similar topics, in German and English. The checker's focus follows from that.

What separates the two groups

  • Paragraphs without a single concrete detail
  • Too many subheadings
  • Colons everywhere
  • Lists of three
  • Sentences of the same length

What doesn't separate them

  • The long dash
  • Passive voice
  • Words like "Moreover"
  • Parentheses

Limit of the measurement: all 80 AI texts came from one model (Claude). Many readers still react to long dashes, so the checker shows them as a style hint.

Open and verifiable

The checker is open source under the MIT license and works with fixed rules, without AI. The same text always gets the same result. You can also use it as a command-line tool, a GitHub Action or a Claude skill.

See the code on GitHub

Common questions

What is AI slop?
It is text from a language model that sounds finished but says little: generic claims, unsourced numbers, stock phrases. A text like that could come from any company.
Does the checker detect whether AI wrote a text?
No, and it doesn't claim to. It shows patterns and gaps you can check. Texts written by people can get flagged too.
What happens to my text?
It is checked once on our server and then discarded. We don't store or log it and we don't use it for training.
Is the checker free?
Yes, without sign-up. We only limit the number of checks per time so the service keeps running for everyone.
Is it enough if my text passes?
No. The checker finds gaps, but it can't know whether your statements are true. A number with a source can still be wrong.
Why are some findings a "pattern" and others "needs resolving"?
An unsourced number or a vague appeal to experts is something you need to resolve. A typical pattern is often fine alone and only counts when it piles up.

The checker finds the gap. Only you can fill it.

AI Editorial Team Lite gets out of you, inside Claude, exactly what is missing here: your examples, your numbers, your stance. Free, by link to your inbox.

Rather see the full AI Editorial Team right away?

In 20 minutes I'll show you, on your own topic, what AI Editorial Team adds on top: interviews via link, your content strategy, real search data, publishing to your website.

The WhatsApp button opens WhatsApp. WhatsApp is a service by Meta. Your message goes through WhatsApp, not through us.

What's in the emails?

5 how-to emails over 2 weeks:

  1. 1. Your first article: how to get the most out of the interview (Day 1): How to answer in step 5: a number, a place, an outcome. Dictate instead of typing.
  2. 2. 12 rules against AI slop, with a before and after (Day 3): The rules from the skill, explained with one example.
  3. 3. Fill it in once, and every article sounds like you (Day 6): Setting up your brand voice file and Claude project the right way.
  4. 4. Does your article show up in ChatGPT? Here is how to check (Day 9): A manual visibility test after four weeks, and what the answer tells you.
  5. 5. What Lite cannot do, and who does it for you (Day 13): The limits of Lite, explained honestly. Offer: one free test article in AI Editorial Team.

After that, occasional news about AI Editorial Team Lite and AI Editorial Team. Every email has a one-click unsubscribe link.

You'll get an email with the download link in a moment. The download works without the checkbox, too. We use your address to send you the link. Tips by email only if you tick the box and confirm it in the email. Details in the privacy policy (in German).