Pricing
Case studies
Blog
Docs
Careers
Log in
Book a demo
Spot the slop
0:30
Which one is AI slop?
Round 1 of 5
The study behind this game
Five AI models each wrote their own version of 2,250 real blog posts, from a brief.
Then we trained a model to predict AI slop.
13,500 posts
Dataset
Posts
Human blog posts
From 268 company domains, 2008 to 2022, before ChatGPT
2,250
AI mirrors
OpenAI
Claude
Gemini
Deepseek
K
written from a brief of each human post
11,250
Detection score
Macro-F1 on the test set, unchanged after rewording
98.0
Can you beat our model in identifying AI slop?
How the pairs were made
·
Read the study
Play (2 min)
How the pairs were made
Read the study
Spot the slop - Sitefire