Bluesky Crash Course: Labelers

Bluesky Crash Course: Labelers
Bluesky's prior logo for the AT Protocol, an @ symbol backed by a blue sky with clouds

[DISCLAIMER: Because it keeps coming up somehow, neither myself nor this blog are affiliated with Bluesky PBLLC]

In the beginning, online content moderation was created. This made a lot of people very angry and has been widely regarded as proof that All Mods Are Bastards.

On March 20th, 2024, Bluesky PBLLC launched app version 1.73, which introduced the Community Labeler system. This has made a lot of Very Normal people upset and has been met with mixed emotions, primarily because end users still do not understand what, exactly, this system does or how it works. This is going to attempt to demystify community labelers, what their capabilities and use cases are, and how they might evolve as the platform does.

What is a labeler, really?

I'm glad you asked! A community labeler is a specially designated account that can utilize Bluesky's open-source Ozone software, apply labels to accounts, and perform other basic moderation functions, with some exceptions. Here is Bluesky's own documentation on how to set up a labeler service account. Note that this process is very tech-heavy and you WILL need to have a decent amount of know-how in order to get one running, i.e. while everyone can set up a labeler, few will have the know-how to do it.

What makes a labeler account different from a normal account on Bluesky?

For one thing, labeler accounts have special features that regular accounts do not, but are also restricted in some other ways. We'll look at the labeler account XBlock Screenshot Labeler, run by, for an example of what I mean by this.

The profile page for the XBlock Screenshot Labeler with the account headings Labels, Posts, and Replies

For one thing, you'll probably notice that there isn't a listed follower count, just a number of users who have 'liked' the labeler, and instead of a Follow button, you will instead be greeted with a Subscribe button in the top right. These features are available to labelers, they're just hidden to everyone except the account holder. Here's what the labeler sees:

The interface for the Aegis Blue labeler, whose account is run by yours truly.

This gives a labeler a realistic perspective on how many users are utilizing their service at any given time, however, this is also likely much lower than the actual number of users on the service, since you can subscribe without following the labeler directly (the Follow button is in the contextual menu with the three dots).

The most important difference between a labeler and a standard account, though, is their ability to apply labels to accounts or content via Ozone, which we will cover next:

What are labels, and what do they do?

Labels are descriptive tags which can be applied at various levels, including at the post level or the account level, with varying effects and interactions with both users and the affected account/user. You can access what labels you're currently subscribed to using your Moderation tab on the sidebar, under Advanced. We'll use this label from Bluesky's own moderation layer as an example.

The Intolerance label as described by Bluesky Moderation with options for Off, Warn, and Hide

If you set a label to OFF, it will not be applied to your view on any content. This is the equivalent option to not using the label at all.

If you set a label to WARN, it will be obscured behind a toggleable Content Warning flag, like this:

A warning for content described as Racial Intolerance (an label)

If you set a label to HIDE, it will not appear on your feed at all. This is the equivalent of blocking without the block actually taking place (i.e. if the content is set to hide for you, it will only affect your visibility of it, it still exists for other users).

The labeler service can also set the label to Show Badge, which is usually applied at the account level. This will appear on the profile page as well as anything the account posts. These can, in theory, be used to indicate anything from an improvised Verified badge, to a tag for an account that is impersonating someone else.

An example of a label badge, this one for Impersonation

How do post labels and account labels differ?

One of the early sticking points with how labels function is that the UI is currently (at time of writing this, the app is on v1.77) really bad at distinguishing between when a label is applied to an account or a specific post. One of the most common ways this happens is when an account that has had a label applied to it posts something innocuous, say, a picture of a cat, when the labeler has applied, for instance, a Racial Intolerance label to the account, the picture of the cat will look as if the labeler is saying "this picture of this cat is racist" when in actuality, what's happening is that when a label is applied at an account level, everything that account posts will look as though it has that label applied, when in actuality, it's because of the account it's originating from, not the content itself.

It's for this reason that labelers focused on activities like moderation will tend to label more at the post level vs. the account level, as the post level says "this post was [X]" vs the account level which says "the user controlling this account is inherently [X]."

As far as differentiating between the two, a post label will ONLY appear attached to the post itself, an account label will be visible on the profile page itself as well on everything the account has or will post.

Who sees the labels applied to content?

One of the interesting features of labelers, by design, is that their custom labels will only show up when subscribed to the labeler in question, with the sole exception of Bluesky's moderation service. This means that a post or account could have a label applied to it, but the affected user will have no way of knowing this unless someone who's subscribed to the labeler views their profile. There are experimental services like this label scanner that can be used to manually scan profiles or posts for any labels applied to them, and by which service. This is also tied into the Appeals system, to be mentioned later.

How do you find labelers to subscribe to?

Well, see, that's the fun part: you don't. Not one that's immediately apparent, anyways. However, two methods are commonly used to find a complete list of active labelers:

  1. This user list by, under the "About" tab.
  2. This website by

Additionally, all accounts are by default utilizing Bluesky's built in moderation service, however, it is in fact entirely possible to disable all of the built in moderation filters that Bluesky provides.

Why might someone use a labeler service?

There are a wide range of possible use cases for labelers that have come up in just the past month of their existence. Here's some notable ones:

  1. As an additional layer of moderation/friction in addition to Bluesky Moderation. These can be region-specific mod services, mod services for certain communities or language clusters, or general use moderation services. The best way to think of this is like using an additional latch on the "door" to your timeline - sure, the lock on the door provided by Bluesky might work, but sometimes depending on the person, you might want an additional layer of protection, in case the first one fails or doesn't do what you want it to. The most-utilized service of this type next to Bluesky's built-in service is Aegis
  2. As a means to better categorize and filter the content they see on their timeline. Some labelers, like XBlock mentioned before, are well suited to preventing a specific kind of content from showing non-labeled, such as screenshots of posts from other social media platforms. Others are specifically designed to prevent spoilers from, for instance, Star Wars, from showing up without proper labels or hiding spoilers from showing up entirely. Other labelers might try to prevent phobia-triggering content from showing unlabeled, to better assist users with a fear of things they might be exposed to, for instance, spiders.
  3. As a joke. This is Bluesky, after all, and we have fun here. Labelers like's Stop The Beans labeler, is entirely focused on labeling pictures of beans. Anti-Alf Aktion, while it does dabble in other categories, was originally created to label content specifically featuring the penis of sitcom character ALF, comedically rebranded to Fuck ALF, or Sensual ALF.
  4. For literally any reason they might want. The entire point of composable moderation is that it's customizable to each and every user, whether they use labelers or not, which ones they use, and how they are configured.

As a final note, as of the time of writing, there is a current upper limit of 10 labeler services that can be subscribed to at any one time. The service will not allow a user to subscribe to more than this to cut down on potential spam/malicious use of the services.

How do you use a labeler service?

As far as configuring how you want labels to show up, these can all be adjusted via your Moderation tab under Advanced once you subscribe to a labeler. You can also configure Bluesky's official moderation settings in there and adjust for their presets.

To report a post or an account to a labeler, click on the three dots on the post or profile page, click the report post button, and select the primary labeler you wish to file the report for.

The option to send a report to either Bluesky's Moderation Service or Aegis, a community moderation network, as an example

Next, you'll be prompted with a report reason - select whichever is most appropriate, or Other if none of them quite fit.

A list of report reasons for a post including Spam, Unwanted Sexual Content, Antisocial Behavior, Illegal and Urgent, and Other

Finally, you will have the option to select additional labelers to send the report to if it's relevant. Make sure to put something in the text box so the labeler knows some of the context (if needed - you can also link other posts in the text box for the labeler to search for additional context). Also, be sure to ONLY include the other labels if the report is going to be relevant to something they handle (more on this in the next section)

An example of the context/additional labeler screen

From there, just wait for the labeler's response. Currently, there is no way for labelers to communicate bidirectionally with reporters, so your confirmation will probably come in the form of the content being labeled...or not. This can also be affected by labeler response times, staff, deliberation, or other factors. In other words, just cross your fingers and wait, resist the urge to spam reports.

When it comes to reporting, there are some criteria to keep in mind:

When to use Bluesky Moderation (i.e. the default labeler)

Bluesky's official mod service is best utilized when you are trying to accomplish the following:

  • Report an account or post for any of Bluesky's specific label criteria (for a full list see their profile here )
  • Report content which is illegal (i.e. doxxing, actionable threats, CSAM, extremist content) or that violates Bluesky's Terms of Service or Community Guidelines.
  • Petition for an account to be taken down or suspended (currently, ONLY Bluesky Moderation can suspend or take down accounts - see upcoming sections for more info on exceptions to this)

When to use a community labeler

Community labelers, mod services and the like, can serve the same functions as Bluesky's mod service, but with noted limits to their capabilities:

  • Community labelers cannot shadowban, suspend users, take down accounts, or apply any of Bluesky Moderation's default labels (namely the adult content flags) unless they ALSO declare these as custom labels. The only way to apply these is for the labeler to have their own replicated labels, or to report the content to Bluesky directly.
  • The exception to the above is in the scenario (not exactly possible as of time of writing this) wherein the labeler in question, an App View, AND a Personal Data Server, or PDS, that the content is hosted on, are controlled by the same entity. In this scenario (which is very tech heavy and/or expensive, so likely out of the affordability of an average user) it is hypothetically possible for the community labeler to have moderator authority over any content posted to that specific PDS. However, this remains to be seen if it is viable or whether or not this occurs.

With this in mind, here's when to use a community labeler:

  • When the content being reported aligns with the labeler's declared label criteria (for a list of these, consult with the labeler service you are utilizing and see what their label descriptions advise)
    • As an example, if you were subscribed to both Aegis and Stop the Beans Labeler Service, do NOT include Stop the Beans on a report for transphobic content (unless you somehow found a photo of transphobic language superimposed on an image of beans, but that's unlikely). This allows labeler operators to do what they need to without unnecessary reports flooding their queue. You should, however, in the above case, include Bluesky's moderation service in the report, as transphobia is against Bluesky's Community Guidelines, section 2B.
  • When the content being reported IS NOT: illegal, urgent, or requiring a suspension or ban. These cases should ONLY be reported to Bluesky moderation as no labeler will be able to act on this in the necessary manner.

How to Appeal a Labeler Decision

Sometimes, labelers can mess up. Shit happens. Let's use a hypothetical case wherein your account has been wrongly labeled by a labeler service. How do you appeal this decision?

Well, for one thing, you need to know the label is even there in the first place. If you are not subscribed to the labeler that applied the label, you will not see this label. You can, however, check any profile or post for a label using this experimental website.

Once you've confirmed the presence of the label, here's the interesting part. You need to subscribe to the labeler service that labeled you. While this seems odd, it's because the only way to send in reports to a labeler is to be a subscriber (as of app v.1.77), and since an appeal is a type of report, you have to subscribe to appeal a decision. You can, of course, unsubscribe after a verdict is achieved.

Once you've subscribed, your profile will look like this:

A testing account which has intentionally been mislabeled by Aegis for testing.

Click or tap on the popup indicating the labels being applied. On a post, this will appear under your display name, near the top of the post.

Example misapplied labels by Aegis, with matching Appeal buttons.

Once there, you will see which labels are present, which labeler applied it, and an Appeal button. Click that button. You will then be prompted with a text box to appeal the labeler's decision - you will have to individually appeal each label if there are multiple. Note that the recipient of this appeal is going to be a human being, and also the only one who can remove this, so it's probably a good idea to keep the appeal very civil. Once finished, hit send.

Here comes the difficult part. Currently, there is no way for a labeler service to have bidirectional communication with an appellant, OR reporters. This means that currently (as of v.1.77) the only way to know that your appeal was accepted is the label being removed or adjusted. In other words, if it is denied, you will have no way of knowing for sure.

Frequently Asked Questions

Q: Will a labeler be able to see the account that sent in the report?

A: Yes. However, these will not be publicly accessible, as all reports occur on the backend. In other words, a report is confidential between you and the labeler operators. Neither the user you report, nor any other users, will know you submitted this report.

Q: Does blocking a labeler prevent that labeler from labeling my content?

A: No. Because the most common tactic to avoid moderation action is to block the account applying it, all labelers operating on Ozone have a built-in block override function when parsing reports. This is crucial to them functioning as intended, to ensure they can view the full thread context for a reported post without the thread being cut short by a block. In other words, the only thing that blocking a labeler does, functionally, prevent a labeler's posts from showing on your timeline.

Q: Are any community labelers officially sanctioned by Bluesky or assist in its moderation decisions?

A: No. Every labeler that exists is made by someone in the community, and none are endorsed officially by Bluesky itself.

Q: Are we required to use the labeler system?

A: No. Labelers, including Bluesky's own moderation service, are completely voluntary. You could have labels on your account, never know it, and function more or less alright. You can even disable every single one of Bluesky's own labels for a completely unfiltered experience, save for suspensions or takedowns of profiles or content. It's how the system is built.

Note: This is a live-edited list as newer and prominent labelers come into play that may be useful to users interested in exploring the system. These will also be all English language services, though a full list of labelers can be found here


Designed by the girl who wrote this blog, Aegis is the product of one of the longest-running moderation services on Bluesky, colloquially known prior as the Contraption, though now purpose-built for the task. Operating as a collaboratively run project, Aegis' moderators use a hybrid model of moderation to ensure the greatest amount of protection for their users. When an account is tagged with an Aegis label, it is automatically added to a corresponding moderation list, which the user can choose to mass-mute or block, to ensure the offender cannot continue to leave harmful comments towards users who might be affected.

XBlock Screenshot Labeler

One of the first labeler accounts to be published, XBlock has one goal in mind: control the flow of third or first party screenshots, and allow its subscribers to either hide them entirely, or reduce their visibility on their timeline, so they're not constantly bombarded by reminders of the shittier social media platform that they left to get here. I'm not clarifying which one that is, but we're all thinking it.

AI Imagery Labeler

Not a fan of AI-generated images or art? This is the labeler service for you. This labeler also utilizes some basic automation, auto-flagging certain users or posts containing specific hashtags.

Crypto Labeler

Similar to the AI imagery labeler, if you'd rather not have to deal with posts from proponents of cryptocurrency or similar spammy posts about it, this is the labeler for you. A very new addition to the labeler roster by

IFTAS Labeler

Similar to Aegis, IFTAS is a moderation-oriented labeler with its roots in Mastodon and the Fediverse, which I am going to pretend to know what those are, for the sake of this summary (sorry gang). If it's something serious, like a coordinated harassment campaign, account takeover, impersonation, and more, IFTAS is a solid pick. More information can be found on their website.

Phobia Labeler

Got a phobia or several? Check out this labeler for whatever icks you, and use its services to completely wipe that content from your timeline. Simple, straightforward, and to the point.

Bad ALT Text Labeler

Do you hate it when folks neglect to add in alt text? Do you use a screenreader and want none of these posts in your timeline? This is the labeler for you. Also includes provisions for unicode abuse, which can also interfere with screenreaders.

Stop the Beans Label Service

Sometimes, you need a labeler that does one thing. Sometimes, you really don't want to see pictures of beans. This labeler can help. This is literally all it does, I'm not kidding.

Taurus Shield

Similar to Aegis, Taurus is a labeler service that is primarily oriented towards the LGBTQ+ and furry communities on Bluesky. It has its own custom declared labels to service those two communities for a wide array of content that they might not want to see. This is also one of the only labelers with a feed designed off of one of its content labels, since it uses some labels to strictly categorize the network instead of moderation solely. It's also experimenting with labelers' ability to add self-generated Verified badges, in its case, for News Services.

Spoiler Droid

Hate seeing spoilers on your timeline? Spoiler Droid proposes an alternative: what if you simply never had to see them? A smaller labeler, this service hopes to prevent users, particularly those using the Star Wars feed, from accidentally having their favorite series ruined.