Bluesky Tests Private Dislike Button for Discover

Bluesky’s new “dislike” button experiment is an opt-in, private vote designed to improve what appears in its Discover section. Bluesky is targeting the root cause of one of the network’s most disliked features by adding a negative signal to the feedback loop. This teaches the recommendation system what not to show as often and does not affect public feedback metrics.

What Bluesky Is Actually Testing With Private Dislikes

The Bluesky product team is testing a new private control that lets users signal content they want to see less of. The signal reduces the presence of that content in Discover and will not aggregate as public preference points.

It feeds into personalization, especially in Discover, and modestly influences reply ranking by steering toward less controversial outcomes further down the stack.

In general, it’s a “dislike/not like this” signal that reduces exposure for you and for people with similar tastes. By keeping it private, Bluesky believes it can collect valuable feedback without introducing visible negative counters or public judgment.

Why the Discover feed needs a reset to curb engagement bait

Since launch, Discover performance has tended toward engagement bait—short, spiky posts that generate quick clicks but don’t reflect users’ real interests—because of weak signals and feedback loops. Power users have noted these dynamics.

How other platforms already use similar feedback

A private dislike acts as a counterweight to the ranking system. If you hide celebrity drama, sports memes, or crypto charts, you should see fewer of those patterns moving forward. Because Bluesky is built on the AT Protocol and encourages custom feeds, better core signals should also help third-party feed builders create better filters and more interesting topic streams.

This has been tried before: X has a reply downvote that isn’t publicly visible and uses it as a ranking signal. YouTube removed public dislike counts but still uses the action as an internal signal. Instagram and TikTok both feature “Not Interested” buttons that tell the algorithm to avoid certain posts or topics. Reddit remains the only major public social network with upvotes and downvotes, although norms around that feature are mixed.

Research on ranking systems shows that negative signals are strong predictors of future dissatisfaction. Even a single “hide this” action can carry more weight than many likes. By capturing this signal directly, platforms can remove unwanted content and reduce the chance of users churning from key surfaces like Discover. Bluesky adds a twist by focusing on personal use and discouraging public judgment; private dislikes also reduce social pressure and encourage honest feedback.

Potential risks and safeguards for weighted negative signals

Still, there are risks. If negative signals are weighted too strongly, filter bubbles can form that shield provocative or challenging content. Platform teams must also mitigate coordinated misuse, as some groups may attempt to drive down scores en masse for particular subjects.

Common mitigations include rate limits, anomaly detection, and signal blending.

What private dislikes mean for creators and conversations

Creators will not be harmed as heavily by invisible downvotes, but they may notice shifts in their audience. Posts that elicit private dislikes receive less distribution in Discover, which reduces growth potential. Specificity is rewarded: content that resonates with a particular group performs better than broad prompts aimed at harvesting likes.

A light touch is crucial—not so heavy that it suppresses experimentation, and not so light that it makes room for low-effort posts. Proper calibration treats private dislikes as one more signal in a current mix of social links and improvement indicators for creators. It will not erase low-effort content altogether.

How private dislikes fit into Bluesky’s broader strategy

Bluesky, now a decentralized social network in which customers construct or purchase feeds, has pitched itself on stronger core signals—positive and negative—powering a healthier ecosystem. The experiment is a modest first step, but if the dislike test works, feed builders could use it to deliver more tailored experiences.

For example: no spoilers for film geeks, fewer trading posts for art lovers, and more control over their feeds for professionals. The experiment also aligns with a broader industry push to give customers clearer, faster feedback on why recommendations were made. Developers will be able to offer more granular ways to say “less like this” while building feeds, rather than assuming that a single favorite implies “show me more like this.”

What to watch next as Bluesky rolls out private dislikes

Pay attention to rollout details:

• Whether the dislike appears across all feeds or only Discover
• How quickly it affects the content mix
• Whether Bluesky shares transparency notes about how the signal is weighted

If engagement quality improves and Discover feels less repetitive, this small button could punch above its weight. For a network competing with larger incumbents, sharpening personalization without amplifying toxicity is a delicate balance. A private dislike—with guardrails—might be the precision control Discover needs.