Bluesky shows you two numbers on your profile: followers and following. It never shows you the interesting part, the overlap. Who followed you and left. Who you followed at a conference and who never looked back.
Here are the three ways to get that list, from most tedious to least.
1. By hand (fine up to about 200 follows)
Open your own profile, then the Following tab, and scroll. For each account, Bluesky shows a “Follows you” badge if they follow you back. Anyone without the badge is a one-way relationship.
This genuinely works, and for a small account it takes ten minutes. It falls apart around a few hundred follows: the list is paginated, the badges only render as you scroll, and you have no way to sort by anything useful, like “biggest accounts that ignore me” or “people I followed a year ago”.
2. With the API (fine if you write code)
The AT Protocol exposes the whole graph publicly, no authentication needed:
GET https://public.api.bsky.app/xrpc/app.bsky.graph.getFollows?actor=you.bsky.social&limit=100
GET https://public.api.bsky.app/xrpc/app.bsky.graph.getFollowers?actor=you.bsky.social&limit=100
Both are cursor-paginated at 100 records a page. Pull both lists, put the follower DIDs in a set, and subtract. Compare identities by DID rather than display name. A public, unauthenticated response does not provide your own authenticated viewer state; finding the record to change requires an authorised request.
Two traps worth knowing. First, getFollows returns accounts in reverse
chronological order, so the first page is who you followed most recently, not
who matters most. Second, deactivated or suspended accounts can be missing from lists while still contributing to counters. A difference in totals does not prove that someone stopped following you.
3. In one click
That is what the free analyzer does: type a handle, and it pulls both sides of the graph, compares them, and shows you the counts plus the biggest accounts that don’t follow back. No sign-in, nothing to install, it only reads what Bluesky already publishes.
If you want to act on the result, connecting the account in the app adds filtering, sorting by follower count or by when you followed, a protected list that bulk actions never touch, and unfollowing in bulk inside a safe rate budget.
What to actually do with the list
Do not mass-unfollow everyone on it. A one-way follow is not an insult, and some of them are the whole reason you are on the network.
A better pass, in order:
- Protect first. Friends, people whose posts you actually read, accounts you follow for information rather than for reciprocity. Mark them and forget them.
- Look at the big accounts. Any account with 100k followers is not going to follow you back, and that is fine, but ask whether you read their posts. If not, they are just noise in your feed.
- Look at the oldest. People you followed a year ago and have not thought about since are the easiest wins. Sorting by “followed longest ago” surfaces them.
- Leave the dormant ones for last. An account that stopped posting six months ago costs you nothing and might come back.
Why the number matters, a bit
Your following list is your feed. Every account on it that never posts, or posts things you scroll past, dilutes what you see. Cleaning it up is not about the ratio, it is about the timeline you get tomorrow morning.
Two related questions come up straight after this one: how to grow without follow-for-follow, and how to tell a bot from a quiet human among the accounts that never followed back.
Start here: check any handle, free and without an account.
Present relationships and recorded history
A comparison shows who follows you in the data read, not who never followed you. Detecting a departure requires an earlier observation. Blue Horizons records changes from your refreshes onward and cannot reconstruct an unobserved past. Check timestamps and partial-read notices before acting. Bulk actions and protection require Pro or Max; Basic supports individual actions. An upgrade cannot restore historical records already removed under the retention rules.