Bluesky Like Bots: I Measured 1.1 Million Likes to Find Them
Bluesky like bots exist, but they're a small slice of the likes you get. In 1,138,670 likes I pulled from Bluesky's firehose on September 29–30, 2026, only 2.4% came from accounts liking 100 or more times in 15 minutes, and most of those looked like ordinary, established users. The blatant ones did stand out: nine throwaway accounts on one small independent server made 4,629 likes, one of them 1,010 in about ten minutes, and all nine were suspended by the time I checked.
If you search "bluesky like bot" today you get Fiverr gigs, bot-building libraries and very little else. None of the pages I found say how common like bots are or what they look like in the data. Likes on Bluesky are public, so I counted them.
(Disclosure up front: I make Agent Sky, a Bluesky growth tool. It follows people for you on a paced schedule. It doesn't like anything. More on it at the end.)
How I Measured It
I replayed Jetstream, Bluesky's public stream of every record created on the network, and kept every like in six 15-minute windows, roughly six hours apart, from 10:08 UTC on September 29 to 14:53 UTC on September 30, 2026. That's 1,138,670 likes from 208,045 accounts (between 155 and 285 likes a second, depending on the hour), plus 14,118 unlikes.
For each account I counted likes per window and the time gaps between them. Then, at 16:14 UTC on September 30, I looked up the current profile of every account that had liked 50 or more times in a window, plus random samples of 400 lighter likers for comparison. Everything below is aggregate. I'm not naming anyone, because a fast liker is not automatically a bot.
How Many Bluesky Likes Come From Bots?
Very few, going by pace alone. The median account liked twice per window, and nine in ten accounts liked fewer than ten times:
Likes per account in a 15-minute window. Six windows, September 29–30, 2026, 1,138,670 likes:
| Likes in 15 minutes | Share of accounts (per window) | Share of all likes |
|---|---|---|
| 1 | 40.3% | 9.2% |
| 2–9 | 49.2% | 42.4% |
| 10–49 | 10.1% | 40.7% |
| 50–99 | 0.36% | 5.3% |
| 100–199 | 0.06% | 1.7% |
| 200 or more | 0.01% | 0.7% |
A hundred likes in 15 minutes works out to one every nine seconds, which is fast but possible for a person working through an art feed. So I also checked signals that are much harder for a human to produce:
Accounts showing machine-like liking, same sample:
| Signal | Accounts | Share of all likes |
|---|---|---|
| 100+ likes in a single 15-minute window | 158 | 2.41% |
| Most gaps between likes under one second (30+ likes) | 43 | 0.70% |
| 50+ likes in at least 4 of the 6 windows, around the clock | 3 | 0.12% |
| One network of throwaway accounts on an independent server | 9 | 0.41% |
Even on the loosest definition, fewer than 1 in 40 likes came from accounts liking at machine pace. On the stricter ones it's under 1 in 100. When a post of yours gets a like, it almost always came at a human pace.
Most of the 158 fast likers didn't look like bots. Their median profile had 1,125 followers, 672 follows and 495 posts, and only 10 of the 158 (6.3%) had been suspended. Nine of those ten were the same network. (For comparison, the median account that liked fewer than ten times had 383 followers.)
What Does a Bluesky Like Bot Actually Look Like?
The clearest like bots I found were disposable accounts that liked hundreds of posts in one burst and were suspended by the next day. The network used nine accounts, each on its own subdomain of one small independent server (Bluesky lets anyone run a server, and this one takes open sign-ups). Their IDs were the did:web kind tied to that server's domain, not the kind Bluesky's own sign-up gives out. They all followed the same routine:
- A fresh account comes online. For five of the nine, the firehose logged the account going active in the same window it started liking.
- It likes hundreds of posts within minutes. The busiest made 1,010 likes in about ten minutes, and 99% of the gaps between its likes were under a second. Nobody tapping by hand can go that fast.
- It spreads the likes thinly. The nine accounts' 4,629 likes landed on 3,938 posts by 3,532 different people, so 95% of the people they hit got one or two.
- It never showed up again in my other windows. Each account liked in exactly one window, and a different one turned up in five of the six windows.
- It gets suspended. When I checked, the public API answered "Account has been suspended" for all nine.
That fifth step answers "do like bots get caught?", at least for the obvious kind: all nine were suspended by the time I checked, between 6 and 30 hours after each one's burst.
One more thing I didn't expect: a lot of unliking looks automated too. 22.9% of unlikes removed likes more than 30 days old, and just ten accounts made 21% of all unlikes, the busiest over 350 in a single 15-minute window. That looks like bulk like-deletion tools. Another 29.6% of unlikes came within a minute of the like (probably mis-taps).
How Can You Tell If a Bluesky Like Came From a Bot?
Tap the account and look at when it joined, what it has posted, and where its handle lives. None of these proves anything alone, but together they're a good sign:
- It only just appeared. For five of the nine accounts above, the firehose logged the account coming online minutes before its burst of likes.
- It has posted little or nothing, and has no bio. (I couldn't check this for the network above, because their profiles were already suspended. But an account that exists only to like has no reason to post.)
- Its handle is on a domain you've never heard of, rather than
.bsky.socialor a recognisable personal or company domain. (Custom domains are normal on Bluesky, so this is a hint, not proof.) - Its profile won't load, or shows it as suspended. That's moderation having got there first.
If a like came from an account like that, you don't need to do anything. The network above spread its likes across 3,532 people, so it almost certainly wasn't about you.
How Do You Build a Bluesky Like Bot That's Allowed?
Make it opt-in: have it like only posts that tag it. That's the one kind of like bot Bluesky's docs allow. The outline:
- Create a separate account for the bot and an app password for it, so your code never holds your main password.
- Label it as a bot. Bluesky's bot guide asks bot accounts to add the
botself-label to their profile "as a best practice." - Use Bluesky's own SDK. The bot starter guide walks through a TypeScript bot built on
@atproto/api, including login and session handling. - Like only posts that mention the bot, for example to confirm it logged a request. Never like posts it found by searching or scrolling.
- Stay well under the rate limit of 1,666 likes an hour. An opt-in bot rarely comes close.
There are also tools that sell auto-liking for your main account. One Chrome extension, SkyFollow, lists "auto like" among its features. Those like strangers' posts on your behalf, which is the kind of liking the next section is about.
Are Like Bots Allowed on Bluesky?
Not if they like strangers' posts. Three Bluesky documents cover it:
- The bot rules. Bluesky's bot starter guide welcomes bots that post on a schedule, then says: "If your bot interacts with other users, please only interact (like, repost, reply, etc.) if the user has tagged the bot account. It must be an opt-in interaction, or else your bot may be taken for spam."
- The community guidelines. "Do not artificially manipulate features or social signals to gain unearned reach or mislead users, including engagement metrics, follower counts, or other measures of community interest" (Community Guidelines).
- The rate limits. Bluesky's rate-limit docs cap a Bluesky-hosted account at 5,000 points an hour, with each like costing 3, so 1,666 likes an hour and 11,666 a day at most. Bluesky says it set that ceiling with "the most active human users on the network" in mind, and that the limits "shouldn't affect typical Bluesky users." The docs say the limits "will affect prolific bots, such as the ones that follow every other account or like every post on the network," and that "bulk or spammy interactions are against the Community Guidelines."
A bot that likes random posts to get attention is exactly what the rules describe as spam. The busiest account in the network above liked about once every 0.6 seconds, far faster than anyone taps. That's still under the hourly cap of 1,666, so the rate limit alone wouldn't have stopped it. Moderation did. For the wider rules on automation, see are bots allowed on Bluesky.
Should You Use an Auto-Like Bot to Grow?
No. Setting the rules aside, it doesn't get you much:
- Likes are public. Every like is a record in your account's public repository, visible to anyone who looks. Hundreds of likes an hour on posts you obviously didn't read is easy to spot and doesn't make you look good.
- Likes aren't what earns attention back. In my study of 14,400 random posts, what went with getting likes was posting images and replying to accounts near your own size. I couldn't measure liking itself there, but a bare like from a stranger is easy to miss among everything else in someone's notifications.
- It's the same idea as bought engagement. Either way, the likes don't come from anyone who read the post.
What people usually want from an auto-liker is to get in front of the right people without hours of doing it by hand. That's what I built Agent Sky for. It follows accounts that are active in your niche at a human pace, well below the rate limits, and checks for a live follow-back before it unfollows anyone. It doesn't like, repost or reply for you. $9 a month, free to start. It won't make a bad post good. What it saves you is the hours of finding and following people in your niche by hand.
To see how your own posts are doing, our free Bluesky analytics tool takes any handle and shows your engagement rate against the benchmark for your follower bucket.
Honest limits on these numbers. This is 90 minutes of likes spread over 29 hours, not a full day, so rare patterns could be missed. My thresholds (100 in 15 minutes, sub-second gaps) catch fast bots, not slow ones: a bot liking once a minute looks like a person here. Profiles were checked once, right after sampling, so some accounts may be suspended later. And "machine pace" is not proof of a bot. Several of the fastest likers were long-standing accounts with thousands of followers.
Quick FAQ
Is there a like bot for Bluesky? Yes. Bluesky's API lets any account create likes programmatically, Bluesky's own SDK makes a like bot a short script, and freelancers sell custom Bluesky bots. But they're a small part of Bluesky's likes. In a sample of 1,138,670 likes from September 29–30, 2026, only 2.4% came from accounts liking 100 or more times in 15 minutes, and most of those accounts looked like ordinary, established users.
Are there bots on Bluesky? Yes, and many are allowed. Bluesky's developer docs welcome bots that post on a regular schedule. What isn't allowed is automated engagement: the same docs say a bot should only like, repost or reply when a user has tagged it, and Bluesky's community guidelines ban artificially manipulating engagement metrics. In my sample, the clearest like-bot network was nine throwaway accounts on one small independent server, all of them suspended when I checked.
Are auto-like bots allowed on Bluesky? Not for liking strangers' posts. Bluesky's bot documentation says a bot should only interact (like, repost, reply) when the user has tagged the bot account, because it must be an opt-in interaction, or the bot may be taken for spam. The community guidelines separately prohibit manipulating engagement metrics. A bot that likes posts from people who mentioned it is fine; a bot that likes hundreds of random posts an hour is not.
How many likes can you give on Bluesky in an hour? Up to 1,666 an hour and 11,666 a day on a Bluesky-hosted account. Bluesky's PDS allows 5,000 points per hour and 35,000 per day, and creating a like costs 3 points. Bluesky says it set that ceiling with its most active human users in mind, so typical users won't hit it. Accounts on other servers can have different limits. The fastest liker in my sample made 1,010 likes in about ten minutes, which is under the hourly cap but at a pace no person could keep up.