Follow & unfollow alerts: see who crypto KOLs just followed

Quick answer: Follow-graph alerts tell you when a tracked account starts or stops following someone. Follows arrive quickly; unfollows are deliberately held back. That asymmetry is the important thing to understand: X's following lists paginate inconsistently, so an account can appear to vanish from a list and reappear on the next pass. Firing on the first apparent disappearance produces batches of invented unfollows. X-Relay requires three consecutive scans to agree before reporting an unfollow, which costs about fifteen minutes and removes false alarms essentially entirely.

Why the follow graph is worth watching

A follow is a cheap, quiet action that usually precedes a loud one. Before a KOL posts about a project, they very often follow it first — to read it, to keep an eye on it, or because someone showed it to them. That gap between following and posting is the signal, and it exists because following is not a public statement in the way a post is. Nobody drafts a follow.

The reverse is quieter and sometimes more informative. A developer unfollowing the project they built, a team member unfollowing the founder, an advisor unfollowing a protocol — these are things people do when they have decided something and have not announced it. The unfollow is rarely proof of anything on its own, but it is the kind of detail that makes you look harder at something you were already unsure about.

How this is detected, and why that shapes everything

There is no notification from X saying "this person followed someone". The only way to know is to read the account's following list and compare it against what it looked like last time. Everything else follows from that fact:

• It is a scan, not an event. Steady-state re-scans run roughly every five minutes per tracked account, so a follow surfaces within minutes rather than instantly.
• Large following lists are slower. An account following many thousands takes many pages to read. Beyond a page cap, those accounts fall back to a much slower cadence — around half an hour — because burning a full pagination every few minutes for one account would starve everything else.
• A new entry is unambiguous; a missing entry is not. Something appearing that was not there before is a real follow. Something absent might be an unfollow, or might be a page X served inconsistently.

The pagination problem, and the fifteen-minute answer

This is the part worth understanding even if you use a different tool, because a tool that has not solved it will lie to you regularly.

Reading a following list of a thousand-plus accounts means requesting it page by page. Those pages are not perfectly consistent — under load, an entry present on one pass can be absent on the next and back on the third. Nothing changed on X; the list was simply served differently. A naive differ treats every one of those as an unfollow, and the result is batches of twenty-plus false unfollow alerts arriving on a schedule.

The fix is confirmation. X-Relay requires three consecutive full scans to agree that an account is gone before it reports an unfollow. At a five-minute cadence that is a confirmation window of roughly fifteen minutes, and the probability of the same pagination glitch occurring identically across three independent scans is effectively zero.

There is a second guard on top. If a single scan suggests more than five simultaneous unfollows, that scan is treated as suspicious — real people do not unfollow six accounts in five minutes, but a bad page looks exactly like that — and the confirmation requirement is raised further for it.

The trade is explicit and, we think, obviously correct: a real unfollow arrives up to fifteen minutes late, and a false one never arrives at all. An unfollow is a slow-burn signal. Nobody trades in the first fifteen minutes of a developer quietly stepping back, and an alert stream that cries wolf twenty times a day is worse than no alert stream, because you stop reading it.

What is in the alert

The alert names the account that was followed or unfollowed — display name and handle, not a numeric ID — so you can judge it without going and looking anything up. That sounds obvious and is not: the following list arrives as IDs, and the human-readable name has to be carried along deliberately or it is lost.

What to do with it depends on which direction it went:

• A follow of a brand-new account — the highest-value case. Check its age and bio. A KOL following a two-day-old project is a lead.
• A follow of an established account — usually nothing. People follow things all day.
• Several tracked accounts following the same new thing — the strongest signal available from this method. One follow is noise; four independent follows in a day is a cluster.
• An unfollow inside a project team — worth a look at whatever else is happening around that project.

Honest limitations

• Follows are not endorsements. People follow scams to watch them, follow competitors, and follow things a friend asked them to. Treat a follow as a pointer, never as a call.
• Volume varies enormously. Some accounts follow twenty things a week. Watch the graph on people who are selective, or the channel becomes unreadable.
• Unfollows are up to fifteen minutes behind by design, and longer on accounts with very large following lists.
• Very large following lists are scanned less often, so an account following tens of thousands gives coarser coverage than one following a few hundred.
• Private accounts are invisible. Only public following lists can be read.
• A rename is not an unfollow, though it can look like one to a tool matching on handles rather than IDs.

Combining it with the other signals

The follow graph is at its best as a confirmation step rather than a primary source. Its natural partner is bio-keyword discovery: a bio hunt surfaces accounts that just declared themselves in your niche, and most are nothing. A follow from someone credible turns one of them into something worth twenty minutes of your attention. One method generates candidates, the other filters them.

It also pairs with profile-change tracking. A team unfollowing each other and a project quietly editing its bio in the same week is a pattern, and neither half means much alone.

Setting it up

1. Open @x_relay_bot on Telegram, or add X-Relay to a Discord server.
2. Add the accounts whose graph you care about. Prefer selective followers — founders, developers, funds — over accounts that follow back everyone.
3. Enable the follows and unfollows activities, plus profile changes if you want the rug-pull side of it.
4. Route these to their own channel. They are a research feed, not an interrupt, and mixing them with time-critical alerts devalues both.

FAQ

Detect rug pulls via profile changes → · Track an X List → · Guides and comparisons → · Pricing →