Bluesky tests new features to improve reply quality and user control

Bluesky announced several new features on October 31, 2025, aimed at improving the quality of replies and giving users more control over their social media interactions. The social media platform, operating as a Public Benefit Corporation, is testing a mix of ranking updates, design changes, and feedback tools to encourage healthier online conversations.

According to the company's announcement, Bluesky is developing a "social proximity" system that maps the natural communities forming on its platform. This system prioritizes replies from people within a user's social neighborhood—individuals they already interact with or would likely connect with based on their network patterns. The approach represents a departure from traditional engagement metrics that have dominated social media platforms for years.

The social proximity feature works by identifying clusters of users who regularly engage with each other. By surfacing replies from these connected groups first, Bluesky aims to create conversations that feel more relevant and reduce potential misunderstandings that often occur when strangers with no shared context interact. The company stated this mapping system reflects "the people you already interact with or would likely enjoy knowing."

Alongside social proximity, Bluesky will begin testing a "dislike" option as a feedback mechanism. The feature functions as a personalization signal rather than a public metric. According to the announcement, dislikes remain private and primarily affect the individual user's experience. The signal helps Bluesky's system understand which types of posts a user prefers to see less frequently in Discover and other feeds.

The dislike functionality also influences reply ranking to a limited extent. Low-quality replies may see reduced visibility based on dislike signals, though the impact extends only slightly beyond the individual user to others within their social neighborhood. The company emphasized that the signal "isn't global," distinguishing it from public downvote systems used on other platforms.

Bluesky has deployed an updated toxicity detection model designed to identify replies that are "toxic, spammy, off-topic, or posted in bad faith." Posts flagged by this system receive lower rankings in reply threads, search results, and notifications. The approach aims to reduce the visibility of problematic content while maintaining open discussions for good-faith participants. The company did not disclose specific technical details about the machine learning models or training data used for toxicity detection.

A change to the reply button's functionality is also under testing. For top-level posts, clicking "Reply" now directs users to the full conversation thread before opening the composer. According to Bluesky, this modification encourages users to read existing discussion before contributing their own responses. The company believes this simple adjustment could reduce "context collapse and redundant replies."

The platform has refreshed its reply settings interface to make existing controls more discoverable. Bluesky already offered fine-grained settings allowing posters to control who can reply to their content, but according to the announcement, "many people don't realize they exist." The updated design includes a one-time prompt in the post composer to alert users to these options. These settings enable users to prevent unwanted replies before they occur, reinforcing the principle that "conversations you start should belong to you."

This approach to user control builds on previously implemented features. Bluesky introduced a followers-only reply setting that restricts responses to trusted connections, mod lists for sharing moderation preferences, and the ability to detach quote posts to limit unwanted attention or coordinated harassment campaigns known as dogpiling.

The significance of these updates extends beyond Bluesky's relatively small user base. According to the company, traditional social platforms have been built on "systems that optimize for attention and outrage instead of genuine conversation." The platform reached 21 million users in September 2024, growing rapidly after receiving increased attention following regulatory changes in Brazil and dissatisfaction with other social networks.

For marketers and advertisers, Bluesky's approach represents a different model than established platforms. The company explicitly stated it is "not driven by engagement-at-all-costs metrics or ad incentives." This positioning may appeal to brands seeking to reach audiences in less adversarial environments, though it also means Bluesky lacks the sophisticated advertising infrastructure available on competitors like Meta's platforms or X.

The absence of traditional advertising business models raises questions about long-term sustainability. As a Public Benefit Corporation, Bluesky operates under a legal structure that requires consideration of public benefit alongside shareholder interests. The company has not disclosed detailed monetization plans beyond its stated rejection of ad-driven engagement metrics.

Bluesky acknowledged that its approach requires ongoing experimentation. "We won't get everything right on the first try," the company stated in its announcement. The platform plans to refine these systems over the coming months while measuring their impact on user experience. Some features will become permanent, others will evolve, and the company committed to sharing results from its testing.

The technical implementation of these features relies on Bluesky's underlying AT Protocol, an open and decentralized framework for public conversation. This protocol-based architecture differentiates Bluesky from traditional social platforms by allowing users to move their data and identity between different services built on the same protocol. The company describes itself as "an initiative to transition the social web from platforms to protocols."

The toxicity detection system represents a significant technical challenge. Most content moderation systems struggle with context-dependent speech, sarcasm, and cultural differences. Bluesky's model must distinguish between genuinely harmful replies and legitimate disagreement or humor. The company provided no information about error rates, appeal processes, or human review mechanisms for flagged content.

Reply ranking algorithms introduce their own complexities. Determining which replies appear prominently affects conversation dynamics in ways that may not be immediately apparent. If the system consistently down-ranks certain viewpoints or communication styles, it could inadvertently create echo chambers even more insular than those on existing platforms. Bluesky's emphasis on social proximity could amplify this effect by prioritizing familiar connections over diverse perspectives.

The timing of these announcements coincides with broader discussions about social media's impact on public discourse. Multiple platforms have faced scrutiny over content moderation practices, algorithmic amplification of divisive content, and the mental health effects of engagement-driven design. Bluesky's approach positions the platform as an alternative for users seeking different interaction patterns.

For digital marketers tracking social media developments, these changes signal potential shifts in how audiences engage with content. Platforms emphasizing quality over quantity may develop different content consumption patterns. Brands accustomed to viral growth through controversy or outrage may find less receptive environments on networks like Bluesky. Conversely, brands investing in authentic community engagement could benefit from systems that prioritize meaningful interaction.

The reply settings refresh addresses a common user experience problem: hidden features that never get used. Making controls more discoverable through design changes and prompts reflects basic usability principles, yet many platforms resist such changes. Increased use of reply restrictions could fragment conversations, making it harder for new voices to participate in discussions. This creates tension between user control and open discourse.

Bluesky's small team structure affects its development capacity. The company is actively hiring and describes itself as "nimble," but competing with well-resourced platforms requires significant technical expertise. Building effective machine learning models for content ranking and toxicity detection typically requires large datasets, specialized talent, and substantial computational resources.

The company's commitment to transparency regarding experimental features contrasts with competitors who deploy algorithmic changes without public disclosure. This openness aligns with growing demands for platform accountability, though it also exposes Bluesky to criticism when experiments fail or produce unintended consequences.

Implementation timelines remain vague. The announcement indicates that dislike testing will begin "soon" and refinements will continue "over the next few months." This gradual rollout allows for iteration based on user feedback but also means the full impact of these changes won't be clear for some time.

Timeline

• October 31, 2025: Bluesky announces testing of social proximity ranking, dislikes beta, improved toxicity detection, reply context changes, and reply settings refresh
• September 2024: Bluesky reaches 21 million users following growth surge
• Prior to announcement: Bluesky implements followers-only reply settings, mod lists, and detachable quote posts
• Coming months: Bluesky plans continued refinement and measurement of new features' impact

Summary

Who: Bluesky, a Public Benefit Corporation building an open social media platform on the AT Protocol, announced new features through its official team blog.

What: The platform is testing five major updates: social proximity ranking that prioritizes replies from connected users, a private dislike button for personalization, enhanced toxicity detection to down-rank problematic content, a reply button modification requiring users to view threads before responding, and a redesigned reply settings interface with better visibility.

When: The announcement was published on October 31, 2025, with features rolling out gradually. The dislike beta will begin "soon," with continued refinements planned over the following months.

Where: These changes affect the Bluesky social media platform and its apps. As a protocol-based network, the updates apply across all implementations of the AT Protocol that Bluesky uses.

Why: Bluesky aims to improve conversation quality by addressing what it calls "a root flaw" in social platforms—systems optimizing for attention and outrage rather than genuine interaction. The company wants to create "fun, genuine, and respectful exchanges" without relying on engagement-driven metrics or advertising incentives.