Instagram's 2026 purge took 1,260 of my followers

Pooya · 26 Jul 2026

I run the Instagram account for a small woodworking shop. On the morning of 7 May it had 14,200 followers. That evening it had 12,940.

Nobody told me. I found out because a client I had pitched two days earlier wrote back to ask why the number in the media kit I had sent did not match the number on the profile. That is an embarrassing way to learn something about your own account.

My first assumption was that I had been hacked. My second, once I saw other people posting the same complaint, was that Instagram had broken something. Both were wrong, and the second one cost me a week of doing nothing while I waited for the number to come back. It was never coming back.

What happened on 6 and 7 May

Over roughly six hours, Meta ran a platform-wide sweep of bot, spam and inactive accounts. It was not aimed at me and it was not a fault. Meta described it as part of a routine process for removing inactive accounts, and said active followers were unaffected.

The coverage was mostly about the large accounts, because they lost the most in absolute terms — Inc and Yahoo both listed celebrity losses in the millions. Most creator accounts saw drops of 2 to 5 percent. Mine lost 8.9 percent, which told me something about my audience that I did not enjoy learning.

One detail is worth keeping, and it is the reason I now measure things myself. The published figures for individual celebrities do not agree with each other. One outlet put Kylie Jenner's loss at 7.2 million. Another put it at 14 to 15 million. Same event, same week, double the number. If reporting cannot pin down the biggest accounts on the platform, be careful with any single figure quoted about yours.

The measurement that changed my mind

I had been reading the follower count as the health of the account. That is the habit this taught me out of. When I finally sat down and measured properly, the two days looked like this:

Twelve hundred and sixty accounts left and my likes moved by five. That is the whole story. The followers that disappeared were not reading anything, were never going to buy a table, and had been quietly making every number I reported to clients worse.

The engagement rate went up because the denominator got honest. A rate is likes and comments divided by followers, so removing followers who do neither can only push it upward. Mine moved from the bottom of the ordinary band for an account this size into the middle of it, and I did nothing to earn that.

Where mine came from

I did not buy followers. I want to be clear about that, because my first reaction to losing 1,260 accounts was defensive, and I suspect that is most people's.

Two things I did do. In 2024 I ran a giveaway — tag two friends, win a cutting board — which brought in around 900 accounts in a week, almost none of which ever appeared again. And for about eight months I followed back everyone who followed me, which is how you end up with a following list of several thousand and an audience assembled out of reflex rather than interest.

Neither of those is buying followers. Both produce the same shape, and the audience quality signals cannot tell them apart: engagement below the band for the account size, comments thin against likes, a following list out of proportion to the audience. When I ran my own account through that check before the purge it scored 61, and I decided the score was being harsh. It was not being harsh. It was early.

What I would do in the first week after a drop

Take the count and write it down. Not a screenshot of the app, which caches and can lag by hours — read the live follower count and note the time next to it. You need one fixed number to measure the next one against.

Compare engagement before and after, not followers. If likes and comments held while followers fell, nothing left that was ever there. If likes fell by roughly the same proportion as followers, the accounts that went were real, and that is a different problem with a different answer.

Re-run the audience check. A score that improves after a sweep is telling you the sweep found what it was looking for.

Re-issue any rate you have already quoted. I had a media kit out with 14,200 on it. Rather than hide the drop I sent a corrected one with the old and new counts side by side and the engagement rate underneath, which had gone up. The per-post range barely moved, because that range is driven by engagement as much as by size. The client signed. I think the correction helped more than the original number would have.

What I changed afterwards

I stopped running giveaways that ask for tags. I unfollowed about four thousand accounts over two months. And I stopped putting the follower count at the top of a media kit — it goes second now, under the engagement rate, because the engagement rate is the number that survived.

Both of those decisions had consequences I did not see coming, and I have written them up on their own: what happened when a brand told me 3.2% was low, and why my public follower count then stopped moving altogether.

Eleven weeks on the account is at 13,410. That is slower than before, and every one of them arrived on their own.