Details
The system works in three stages: first, it collects all recent posts from followed accounts; second, a lightweight machine learning model narrows these to approximately 500 candidates; third, a more complex model predicts the likelihood of different interactions — like, comment, share, save, or profile tap — and ranks posts accordingly. Key signals include the interaction history between the user and the poster, the post's overall engagement, recency, and content type preference. In October 2025, an Amsterdam court ruled that Instagram's automatic reversion to this algorithmic feed constitutes an illegal dark pattern under the EU Digital Services Act.
Have evidence about Instagram's AI practices? Submit a report.
Report a Sighting →