Bluesky’s Discover feed shows posts from accounts you do not follow. Many users wonder why certain posts appear and others do not. The algorithm does not rely on a single popularity metric like total likes or reposts. Instead, it uses a combination of social graph analysis and interaction patterns. This article explains how the Bluesky Discover algorithm selects posts and what factors influence visibility.
Key Takeaways: How Bluesky’s Discover Feed Works
- Interaction graph analysis: The algorithm ranks posts based on who liked, reposted, or replied to them, prioritizing content from users your network engages with.
- Recency weighting: Posts published within the last 6 to 12 hours receive a higher score than older posts, ensuring the feed stays current.
- No single viral metric: The algorithm blends engagement signals from your network and the broader community, not just total likes or reposts.
How the Bluesky Discover Algorithm Selects Posts
Bluesky’s Discover feed is powered by an open recommendation algorithm called the Bluesky Recommendation Engine. It runs on the AT Protocol, which allows developers to inspect and modify the code. The algorithm does not use a black-box AI model. It applies a set of deterministic rules that score every post in real time.
The scoring process has three main stages: candidate generation, feature extraction, and ranking. In the candidate generation stage, the system collects posts from accounts that your followed accounts have recently interacted with. This is called the social graph expansion. The system also includes posts that have received unusually high engagement across the entire network, known as global viral signals.
In the feature extraction stage, the algorithm computes several numerical features for each candidate post. These features include the number of likes from your followed accounts, the number of reposts from your followed accounts, the age of the post in minutes, and the total engagement rate relative to the poster’s average. Each feature is normalized to a value between 0 and 1.
Finally, the ranking stage applies a weighted linear combination of these features. The current default weights are tuned to favor recency and network engagement over raw popularity. This design prevents a single post from dominating the feed for hours and ensures that content from smaller accounts can surface if their posts resonate within a specific community.
Key Factors That Influence Post Visibility
The algorithm considers three primary factors when scoring a post:
- Network engagement: Likes, reposts, and replies from accounts that you follow carry more weight than the same actions from strangers.
- Recency: Posts published within the last 1 hour receive a 50 percent score bonus compared to posts that are 12 hours old.
- Engagement ratio: A post with 10 likes from 50 followers has a higher score than a post with 100 likes from 10,000 followers.
Steps to See How the Algorithm Affects Your Discover Feed
You cannot directly view the raw algorithm scores in the Bluesky app. However, you can observe the algorithm’s behavior by performing a simple test. The following steps will help you understand which posts appear and why.
- Open the Discover feed on the Bluesky app
Launch Bluesky on your phone or desktop. Tap the Discover tab at the top of the screen. The feed will show a mix of posts from accounts you do not follow. - Note the first five posts and their authors
Write down the usernames of the first five posts. Check whether you follow any of the accounts that liked or reposted those posts. Tap the like or repost icon on each post to see who interacted with it. - Check the age of each post
Look at the timestamp on each post. Posts older than 24 hours are rare in the Discover feed. If you see a post older than 24 hours, it likely has extremely high engagement from your network. - Like or repost a post from a small account
Find a post from an account with fewer than 100 followers. Like or repost it. Then refresh the Discover feed after 10 minutes. You will likely see more posts from that account or from accounts that the small account follows. - Repeat the test after muting an account
Go to the profile of any account that appeared in the Discover feed. Tap the three-dot menu and select Mute Account. Refresh the Discover feed. That account’s posts will no longer appear, confirming that network engagement is the primary signal.
Common Misconceptions About the Discover Algorithm
“The algorithm only shows viral posts from big accounts”
This is not accurate. While posts from large accounts do appear, the algorithm heavily weights engagement from your direct network. If your followed accounts engage with a small account’s post, that post can rank higher than a viral post from a celebrity. The global viral signal only applies to posts that have an engagement rate above the 99th percentile across the entire network.
“Likes from strangers help my post get discovered”
Likes from strangers have a minimal effect on the Discover feed. The algorithm assigns a 0.8 weight to likes from your followed accounts and only a 0.2 weight to likes from strangers. A post with 10 likes from your network will outrank a post with 50 likes from strangers.
“Reposting my own content boosts visibility”
Reposting your own content does not increase its score. The algorithm deduplicates posts by their content hash. If the same text is posted twice, only the first instance is scored. Self-reposts are treated as duplicate content and are ignored.
Bluesky Discover Algorithm vs Twitter For You Algorithm
| Item | Bluesky Discover | Twitter For You |
|---|---|---|
| Algorithm type | Deterministic, open-source | Machine learning, proprietary |
| Primary signal | Network engagement from followed accounts | User interest prediction from browsing history |
| Recency weight | High, posts older than 12 hours rarely appear | Moderate, posts up to 48 hours can appear |
| Global viral boost | Only for top 1% of posts by engagement rate | Applied to all posts with high engagement velocity |
| User control | Mute and block affect feed, no manual refresh of algorithm | “Not interested” button and topic filtering available |
The Bluesky Discover algorithm is transparent and predictable. It prioritizes content from your network and recent posts over raw popularity. You can influence the feed by engaging with accounts you want to see more of. For more control, consider using custom feeds that use different ranking logic, such as the What’s Hot feed which uses a simpler popularity score.