AI Usage at a Glance
Jul 27, 2023
RecommendationPractice documented: Bluesky uses an algorithmic recommendation system to power its Discover feed, surfacing posts based on a user's engagement history, social graph proximity, and network-wide trending signals.
Practice DocumentedView practice →May 10, 2024
RecommendationNew evidence: Bluesky now lets you personalize main Discover feed using new controls
Evidence AddedView practice →Sep 18, 2024
ModerationPractice documented: Bluesky uses automated detection systems — including rules engines and machine learning tools — to identify and act on spam networks, fake accounts, and bot behavior in real time across the platform.
Practice DocumentedView practice →Nov 15, 2024
RecommendationNew evidence: Unlike X, Bluesky says it won't train AI on your posts
Evidence AddedView practice →Jan 17, 2025
ModerationPractice documented: Bluesky uses an automated AI system supplied by a third-party vendor to scan every image and video frame uploaded to the platform and apply content labels such as adult, suggestive, nudity, graphic-media, and self-harm.
Practice DocumentedView practice →Oct 31, 2025
ModerationPractice documented: Bluesky deployed an automated model that detects replies judged to be toxic, spammy, off-topic, or posted in bad faith, and reduces their visibility in threads, search results, and notifications without removing them.
Practice DocumentedView practice →Jan 29, 2026
ModerationPractice documented: Bluesky deployed an automated system that assesses the titles and descriptions of user-created lists for overt toxicity when a list is reported, automatically applying a hidden label if violations are detected.
Practice DocumentedView practice →Mar 28, 2026
ProductivityPractice documented: Bluesky tested Attie, a standalone AI app that allows users to build personalized social media feeds by typing plain-language descriptions, without needing to write code.
Practice DocumentedView practice →Mar 29, 2026
ProductivityNew evidence: Bluesky's new Attie app uses AI to give you full control over your social feed
Evidence AddedView practice →Mar 31, 2026
ProductivityNew evidence: Bluesky Launches Attie, an AI App Custom Feed Creation
Evidence AddedView practice →Bluesky tested Attie, a standalone AI app that allows users to build personalized social media feeds by typing plain-language descriptions, without needing to write code.
Announced at the Atmosphere developer conference in late March 2026, Attie uses Anthropic's Claude as a coding agent to interpret natural language prompts and generate working feed algorithms on the AT Protocol. Users can describe the content they want to see — for example, 'show me tech news but skip crypto drama' — and Attie constructs a deployable feed. Attie is a separate product from the Bluesky app itself, built by a new internal team led by Bluesky's former CEO Jay Graber (now Chief Innovation Officer). As of late March 2026, Attie was in closed beta for Atmosphere conference attendees, with a waiting list at attie.ai; monetization had not been decided.
Bluesky deployed an automated system that assesses the titles and descriptions of user-created lists for overt toxicity when a list is reported, automatically applying a hidden label if violations are detected.
Bluesky uses automated detection systems — including rules engines and machine learning tools — to identify and act on spam networks, fake accounts, and bot behavior in real time across the platform.
Bluesky uses an automated AI system supplied by a third-party vendor to scan every image and video frame uploaded to the platform and apply content labels such as adult, suggestive, nudity, graphic-media, and self-harm.
Bluesky deployed an automated model that detects replies judged to be toxic, spammy, off-topic, or posted in bad faith, and reduces their visibility in threads, search results, and notifications without removing them.
Have evidence about Bluesky's AI practices? Submit a report.
Submit a report →AI Trace is free and nonprofit. Support our work
Bluesky uses an algorithmic recommendation system to power its Discover feed, surfacing posts based on a user's engagement history, social graph proximity, and network-wide trending signals.
Launched in July 2023 as a replacement for the simpler 'What's Hot' feed, the Discover feed mixes trending content with posts from accounts a user follows and accounts near their social graph. Bluesky's official blog states the feed is 'customized for what you like to see.' Since May 2024, users can further tune the feed using 'Show more like this' and 'Show less like this' controls on individual posts. In October 2025, Bluesky began testing a 'social neighborhoods' ranking layer that prioritizes replies from people a user already interacts with or is likely to know. Bluesky confirmed in November 2024 that the Discover feed uses AI, but stated it is not a generative AI system trained on user content.
Bluesky deployed an automated system that assesses the titles and descriptions of user-created lists for overt toxicity when a list is reported, automatically applying a hidden label if violations are detected.
Launched in April 2025, this system automatically evaluates list titles and descriptions for toxic content when a user report is received, applying a '!hide' label that makes the list invisible to everyone except its creator. Bluesky's 2025 Transparency Report states that previously, human moderators would assess and take down toxic lists entirely, which frustrated users who lost their curational work. The automated approach allows creators to revise their content and appeal for label removal, preserving their list while reducing harm. Bluesky credits this change with significantly reducing abusive use of the lists feature.
