Why You Keep Losing Followers, and How to Slow It Down

A falling follower count is the most visible number on an account and one of the least informative. It moves for four unrelated reasons, only one of which is feedback on what you post, and each leaves a different trace in your reach. This page sets out the four, how to tell them apart, and what the sellers of followers say in their own price lists about how long what they sell lasts.

The short answer. You are losing followers because Meta removes fake and automated accounts, because bought followers leave with them, because some people decide to stop following, or because something you posted pushed them. Check reach alongside the count: if reach and interactions held while the count fell, the accounts that left were not seeing your posts anyway.

Is Meta removing accounts that followed you?

Often, yes. Meta takes down fake and automated accounts continuously and at scale. Its transparency reporting describes detection technology that “helps us block millions of attempts to create fake accounts every day and detect millions more, often within minutes after creation”, and notes that blocked attempts are not counted in the published figures Meta Transparency Center. What reaches your follower count is the residue: accounts that were created, followed things, and were caught later.

What it looks like: a fall that often arrives in one step, with reach and interactions where they were. It is not a verdict on you; it is a correction to a list.

What to do: take the improvement. If interactions stayed put while the count shrank, the accounts that went were not taking part, and your engagement rate is now both higher and closer to the truth.

Are bought followers leaving in the same sweep?

If any were bought, by you or by a previous owner, this is the likeliest cause of a large step. Buying engagement breaks the rules rather than merely being unwise: Meta’s spam policy names buying follows among the behaviours it acts on, and Instagram announced in 2018 that it would remove “inauthentic likes, follows and comments from accounts that use third-party apps to boost their popularity” Instagram. Accounts kept to supply follows are the population such removals are aimed at. What an SMM panel is explains the shops that sell them.

What it looks like: a fall much larger than any change in interactions would explain. If eight thousand followers vanish and your average likes do not move, those eight thousand were decoration.

What to do: nothing that helps. A refill draws on the same pool and goes the same way, and the sellers’ own price lists, read below, expect losses. If you inherited an account in this state, the follower audit in our Instagram followers guide covers how to assess it and the three options that follow.

What do follower listings themselves promise about drops?

The people best placed to know whether bought followers stay are the people selling them, and their catalogues are public. Three times in September and October 2026, on 2 September, 13 September and 2 October, we saved the public service lists of fifteen SMM panels for our panel price survey; nothing was bought and no account was opened. For this page we sorted every Instagram follower listing by what its own name says about loss: a promise that nothing will drop, an admission that something will, a refill window, or nothing at all.

On 2 October, 11 of the panels listed Instagram followers in US dollars: 2,370 listings at a median $2.16 per 1,000, with the middle half between $1.10 and $3.99. A name can carry several terms, so the rows overlap.

Prices are the panels’ asking prices on the day, per 1,000 followers. The last column counts listings on adderpanel.com, one of two panels operated by the company that publishes Novamya; the other, smmlaunch.com, had no Instagram follower listing on 2 October.

Sellers expect followers to leave

Listings that expect some loss outnumber listings that promise none by two and a half to one: 547 against 220 (three smmcost.com listings say both “Non Drop” and “Low Drop” and are counted on each side). Of the 547, 320 say “low drop” or “less drop”, 219 put a number on it, and 9 say “high drop”, “may drop” or “drop possible”; one listing says two of these. Where a number is given, the typical ceiling is 5%, and 138 of the 219 say exactly that. Seventeen state a ceiling of 20% or more, and ten of 50% or more, up to listings on smmcost.com that read “Real & Bot Accounts … Drop 70-95% … No Refill” at $0.75 to $0.80 per 1,000.

These admissions are concentrated. smmcost.com, the largest catalogue in the sample, carries 509 of the 547, so the two-and-a-half-to-one balance is one seller’s habit of labelling: across the other ten panels the balance reverses, 86 non-drop listings against 38 that expect loss. What does hold across panels is that both kinds are sold side by side: in seven panels a “non-drop” listing sits beside one that expects loss.

Most “non-drop” listings also sell a refill

Of the 220 listings promising no drop, 153, or 69.5%, also name a refill period in the same line, for example “Non Drop | Refill & Cancel Enabled | 30 Days ♻️” at $0.27 per 1,000, the cheapest non-drop listing on the day. A refill window is an undertaking to top an order back up after it falls. Selling one beside a promise that the order will not fall tells you how the seller weighs the promise. Our refill comparison measures what the refill itself adds to the price.

Does a non-drop promise cost more?

Not reliably. Compared inside each catalogue, non-drop listings had a higher median than listings that expect loss in 5 of the 7 panels that sell both, and the median ratio across panels was 2.82. But the spread runs from 0.60 to 16.83 times, and the figure moved from 5.19 on 2 September to 2.82 a month later while the listings’ wording barely changed, which suggests it tracks what else each panel happens to sell more than the promise itself. In smmcost.com, the one panel with large numbers of both, non-drop listings were the cheaper kind: a median $1.17 per 1,000 against $1.95.

Holding more fixed gives a smaller answer. Pairing non-drop and loss-expecting listings in the same panel, under the same category heading and with the same refill term leaves 11 comparable groups in 3 panels, 9 of them on smmcost.com. The median non-drop listing asked 1.10 times its neighbour, dearer in 7 groups and cheaper in 4. On these catalogues, “non drop” adds little to the asking price.

Did the promises change between readings?

Almost never. 2,286 Instagram follower listings appear under the same service id in all three readings. Of the 220 among them that promised no drop on 2 September, 219 still did on 2 October. Only 6 of the 2,286 changed any of their retention wording, while 790 changed price. The one that withdrew a non-drop promise, a South Korea listing on smmkings.com, replaced it with “30 Days Guarantee” and fell from $51 to $34 per 1,000.

The publisher’s own panels

adderpanel.com and smmlaunch.com belong to the company behind this site, and they are counted rather than left out so that a reader can see where they sit. Neither made a non-drop promise on any Instagram follower listing in any of the three readings. On 2 October adderpanel.com had 72 such listings: 54 name a refill period, 18 say “No Refill”, and one, at $0.37 per 1,000, says “Low Drop”. smmlaunch.com had two in September and none in October. Because neither has a non-drop listing, every price comparison above is the same with or without them; without them, non-drop listings are 9.6% of the remaining 2,298 and loss-expecting ones 23.8%.

Is Instagram different from other platforms?

The balance varies by platform. On 2 October, among 1,520 TikTok follower listings, 13.9% promised no drop and 24.0% expected loss, close to Instagram’s pattern. X follower listings made the promise least often: 2.8% of 358, against 14.5% that expected loss. Facebook follower listings ran the other way, 41.2% non-drop against 7.6% of 568, and so did YouTube subscriber listings, 28.3% against 3.6% of 276. More promises on a platform is not evidence that its followers stay; it is evidence of how that platform’s listings are written.

What these figures cannot tell you

Every figure here is what a listing says about itself. A “non drop” name is a seller’s promise, not a measurement of drop, and “drop 5%” is the seller’s estimate, not an observed rate: we placed no order and measured no follower count over time. Names were read as they appear; five of the panels hide service descriptions behind a button, so wording in those descriptions is not counted anywhere. Every count and price here is produced by follower-drop.py and stored in follower-drop-stats.json; the script also writes out each listing it set aside, with the reason, and halts when a recount of its own totals disagrees. Both files are published with the survey data, next to the saved catalogue pages.

What the catalogues do show is enough for the question this page asks. If you bought followers and they are going, the sellers wrote that down in advance, sometimes with a percentage, and the “non drop” alternative was usually sold with a refill attached in case it did.

Are people unfollowing on purpose?

Some always are. Every account sheds a few followers who decide it is no longer for them, and a single unfollow means almost nothing on its own. What repays attention is the rate.

What it looks like: steady, low, and proportional to how many followers you have. Convert it to a percentage, or every period of growth will make the losses look like a crisis when they are only scaling with the account.

Did something you posted cause it?

This is the only one of the four that is feedback. What it looks like: a step that follows one post or one change within hours or days, rather than a drift over weeks. The usual candidates:

• Posting outside your subject. The most reliable cause by a distance. People followed one thing; anything else is an unfamiliar account wearing a familiar name.
• Reach that arrived for the wrong reason. A post that travels beyond your usual audience collects people who wanted that post. Over the following weeks they find the rest of the account is not it, which is why a follower spike is so often followed by a slump.
• Advertising as a share of the grid. Audiences absorb the occasional commercial post. When a third of what you publish is paid, the account has changed job.
• A jump in posting frequency. More posts means more chances to be the thing someone removes while tidying up.
• Returning after a long absence. A burst of posts reminds people the account exists, and some use the reminder to leave.

How do you tell the four causes apart?

Followers and reach mislead separately and diagnose together. Read them side by side:

• Followers fall in one step, reach and interactions unchanged. Removed accounts, or bought ones going with them. Nothing to act on.
• Followers decline slowly, reach declines at the same rate. Real people leaving, for reasons that trace back to what you post.
• Followers flat or rising, reach falling. The urgent case: the audience is still listed but is not being reached or has stopped responding.
• Followers rising, reach flat. The new arrivals are not taking part. Look at what brought them in.

The third pattern deserves the attention and rarely gets it, because nothing visible has happened. The count catches up later. Why Instagram reach drops works through the causes, most of which are not penalties, and which engagement denominator to use covers measuring the response.

What is the follower number actually for?

As an input to distribution it matters less than people assume. Instagram describes Explore as “photos and videos that we go out and find for you from accounts you don’t yet follow”, and says that on Reels “the majority of what you see is from accounts you don’t follow” Instagram. A system built to serve strangers is not mainly reading your total.

As social proof it matters more, and it is read as a ratio. Someone deciding whether to follow weighs the count against what the posts appear to be getting. Losing ten thousand accounts that never engaged moves that comparison in your favour, which is the sense in which a drop can leave you better off, and why a smaller account with a strong ratio is usually worth more than a large one with a weak one.

Why do unfollower apps disagree with each other?

Because none of them is reading a list of your followers. Instagram’s API exposes followers_count, the “total number of Instagram users following the user”, and no field or edge on the account returns who those users are Meta for Developers. A tool claiming to name the person who unfollowed you is either working outside the sanctioned interface or comparing two snapshots it took itself, which is why two such tools on one account give two answers.

What is officially available is aggregate and late. The insights API offers a follows-and-unfollows metric defined as “the number of accounts that followed you and the number of accounts that unfollowed you or left Instagram in the selected time period”, does not return it below a hundred followers, and warns that data “may be delayed up to 48 hours” Meta for Developers. That definition folds deleted accounts in with real unfollows, so the metric alone cannot separate a removal from churn; reach is what separates them. And giving a third-party app your login for a number Instagram already publishes is a poor trade.

What slows the avoidable losses?

1. Hold the subject steady. The largest single factor. If you need to cover something else, say why, or run it from a second account.
2. Find the frequency your audience tolerates. Increase gradually, check the unfollow rate after each step, and stay below the point where it moves.
3. Be selective about what you go wide with. Reach outside your subject imports people who leave and weaken your ratio on the way out.
4. Keep paid posts to a share you would defend out loud, and mark them as paid.
5. Publish on a rhythm people can predict. An absent account gets removed during a tidy-up; a present one merely gets skipped.
6. Check where new followers come from. Fast acquisition with fast churn buys the same position as patient growth, with a worse ratio to hold it.

What does not help?

• Buying followers to cover the shortfall. It hides the trend, worsens the ratio, and removes any way of telling whether your next fix worked. The catalogues above show the purchase is expected to shrink too. The buyer’s guide sets out what a purchase does and does not deliver.
• Paying extra for “non drop”. Matched like for like, non-drop listings asked about the same as their neighbours (a median 1.10 times, across only 11 small groups, cheaper in 4), so the label is not a reliable sign of a different product, and most listings that carry it also sell a refill window against the drop they say will not happen.
• Publishing more to make up for falling reach. Frequency is often what caused the fall.
• Deleting old posts. No published mechanism connects an archive to current distribution.
• Follow-for-follow. It brings people who followed back by reflex and leave within days.
• Blocking the accounts an unfollower app named. The list is a guess, and acting on it costs you real followers.

When is there nothing to fix?

If reach and interactions are steady and only the follower number moved, nothing is wrong. A count that falls while reach holds is shedding accounts that were not seeing you, a cleanup you would otherwise have had to pay for.

The pattern that warrants real attention is falling reach, with or without a change in followers. The social media growth guide explains how distribution is decided and which numbers repay watching.