Instagram Engagement Rate Benchmarks and the Denominator Trap

A brand manager opens a shortlist of three Instagram creators. The media kits say 6.2%, 3.8% and 1.1%. She ranks them in that order, sends the brief to the first one, and two months later the campaign has produced almost nothing. When she finally pulls the numbers apart, it turns out the 6.2% was calculated against reach, the 3.8% against views, and the 1.1% against followers. Recomputed on the same basis, the ranking reverses completely. The creator she rejected was the strongest account on the list.

This happens constantly, in both directions. Creators lose deals because they quoted the conservative number while a competitor quoted the flattering one. Brands overpay because they compared a reach-based percentage against a follower-based expectation. Agencies report a rate to a client in January, switch analytics tools in March, and spend a week explaining a "collapse" that was a formula change. None of these are data problems. Every one of them is a denominator problem.

Instagram makes this worse than other platforms for a specific reason: the platform removed organic impressions in April 2025 and replaced several metrics with a single unified views figure. So the three denominators people were already confusing became two denominators plus one that no longer exists, while thousands of guides written before that date kept circulating with formulas that point at a metric your account no longer reports.

This guide does one job properly: it makes engagement rate benchmarks usable. You will get the published 2026 figures with their formulas attached, an explanation of why three credible studies report 0.30%, 0.48% and 5.46% for the same platform without any of them being wrong, benchmarks by account tier and by content format, a diagnostic sequence for a rate that is falling, and a mechanical account of what purchased engagement does to each term of the fraction. If you are weighing growth tools, including the Instagram service range on this site, knowing which term you are moving is the difference between a deliberate purchase and an expensive accident.

One position stated up front so nothing later reads as a pitch: buying engagement does not raise the number a serious brand audit computes. In the most common case it lowers it, arithmetically and permanently. That is not a caveat buried at the end of this article. It is one of its main findings.

The three engagement rates every Instagram account has

Your account does not have an engagement rate. It has at least five, and they all come from the same post.

Take a working example. An account with 12,000 followers publishes a carousel. Instagram reports reach of 3,400 accounts and 5,900 views. The post collects 240 likes, 14 comments, 41 saves and 26 shares.

Formula	Calculation	Result
Likes and comments, over followers	254 / 12,000	2.12%
All four interactions, over followers	321 / 12,000	2.68%
Likes and comments, over reach	254 / 3,400	7.47%
All four interactions, over reach	321 / 3,400	9.44%
All four interactions, over views	321 / 5,900	5.44%

One post, five defensible numbers, a spread of more than four times between the lowest and the highest. Nobody in that table is cheating. They are answering different questions.

The follower-based numbers answer "how alive is this account's audience?" The reach-based numbers answer "how compelling was this content to the people who actually saw it?" The views-based number answers "how much interaction did each display produce?" Those are three genuinely different business questions and they deserve three different answers.

The reason this causes so much damage is that the word "engagement rate" is used for all five without qualification. A percentage without its formula is like a price without a currency. The habit that fixes it is small and slightly tedious: never say a rate without saying its numerator and denominator in the same breath. "Our 30-day median engagement rate, counting likes, comments, saves and shares against reach, is 9.4%." Longer, yes. But the person who talks that way has just told a professional counterpart that they know what they are doing, and they have made their number impossible to misread.

If you want the cross-platform version of this argument, including how the same trap plays out on TikTok and X, the companion piece on social media metrics and engagement rate formulas covers the general case. This guide stays inside Instagram, where the specific numbers live.

Why 0.30%, 0.48% and 5.46% are all correct

Three independent 2026 studies measured Instagram engagement at large scale. They published wildly different headline numbers. Here they are side by side with their methodology attached, which is the only way these figures are worth anything.

Study	Numerator	Denominator	Sample	Data period	Instagram median
Rival IQ / Quid, 2026 benchmark report	likes, comments, favorites, reposts, shares, reactions, video interactions	followers	150 randomly selected companies per industry, 18 industries, from a database of 200,000+ companies	2025	0.30% per post
Socialinsider, 2026 Instagram benchmarks	likes and comments only	followers	35 million posts across 447,613 active profiles	January to December 2025	0.48%
Buffer, State of Social Media Engagement 2026	likes, comments, shares and saves	reach, meaning unique accounts that saw the post	9.6 million Instagram posts across more than 200,000 accounts	January 2024 to December 2025	5.46%

Buffer's figure is roughly eighteen times Rival IQ's. That gap is the product of two separate multiplications, not a disagreement about reality.

The first multiplication is the numerator. Buffer counts saves and shares; Socialinsider counts neither. On a well-performing carousel, saves and shares can be a meaningful share of total interactions, so simply widening the numerator lifts the rate before you touch the denominator at all.

The second multiplication is the denominator, and it is the larger of the two. Reach is a fraction of followers. Socialinsider's separate reach study put the average Instagram reach rate at 3.50% for the year to May 2025, down 12% year over year. When your denominator shrinks to a small fraction of your follower count, your percentage grows by the inverse of that fraction. That single substitution is responsible for most of the eighteen-fold gap.

Now look at the more instructive comparison, the one people miss. Rival IQ and Socialinsider both divide by followers. Rival IQ counts more interaction types than Socialinsider does. By rights Rival IQ should report the higher number. It reports 0.30% against Socialinsider's 0.48%, roughly 40% lower.

The explanation is population, not arithmetic. Rival IQ samples corporate brand accounts across 18 industries and reports a median of medians; Socialinsider samples 447,613 active profiles, a pool that includes a large body of creator and personal accounts, which engage at higher rates than brand pages. Both are honest measurements of different populations wearing the same label.

The practical rule that falls out of this: a benchmark is only usable if the study's population resembles your account. A local bakery comparing itself to the Socialinsider average is comparing a brand page against a pool full of creators. A creator comparing against Rival IQ is holding themselves to a corporate median. Both will draw the wrong conclusion, and both will feel it is the data's fault.

Reading the methodology fine print before you quote a number

Four details decide whether a published benchmark means anything for you. They are almost never in the headline.

Median or mean. Rival IQ and Socialinsider both report medians. This matters enormously, because engagement distributions have long right tails. A handful of viral posts pull a mean far above the typical experience. If a source does not say which one it used, assume mean and treat the number as optimistic.

The year the data covers versus the year on the cover. Socialinsider's "2026" report analyses January to December 2025 data, and says so in its methodology. Rival IQ's 2026 report, published in March 2026, likewise covers 2025. Their previous edition, covering data through early 2025, put the all-industry Instagram median at 0.36% with a top quartile around 1.02%. The move from 0.36% to 0.30% is a 17% year-over-year decline, not a difference of opinion between two reports. If you cite the industry table from the older edition alongside the 0.30% headline from the newer one, you have silently mixed two years.

Per post or per period. "0.30% per post" and "0.30% for the month" are different statements about an account posting 3.69 times a week, which is what Rival IQ found for the median Instagram brand in 2025.

Whether the formula changed under you. This is the sharpest one. HypeAuditor moved to an engagement rate defined as likes plus comments divided by views in 2025, falling back to the older follower-based formula for accounts without enough reels. Any chart that spans that switch has a discontinuity in it that has nothing to do with the accounts being measured. The same is true of any account whose analytics tool changed its Instagram integration after April 2025.

The takeaway is not that benchmarks are useless. It is that a benchmark is a sentence, not a number, and quoting only the number throws away the part that made it true.

Converting between denominators with the reach rate bridge

There is a clean identity connecting the two main rates, and it is worth internalising:

Follower-based ER = Reach-based ER x Reach rate

where reach rate is reach divided by followers. Check it against the example from earlier: the reach-based rate on all four interactions was 9.44%, and reach rate was 3,400 / 12,000 = 28.3%. Multiply them and you get 2.67%, which is the follower-based rate of 2.68% to rounding.

This identity is the most useful thing in the article for day-to-day work, because it separates two things that constantly get blamed on each other. Your follower-based rate can fall for exactly two reasons: your content got less compelling to the people who saw it (reach-based rate fell), or a smaller share of your audience saw it at all (reach rate fell). Those two causes have completely different fixes, and the follower-based number alone cannot tell you which one you have.

Now the warning, because this identity gets abused. It only works inside one account with one set of numbers. You cannot take Buffer's 5.46% reach-based median, multiply it by Socialinsider's 3.50% average reach rate, and expect Socialinsider's 0.48% to pop out. It does not, and the mismatch is not an error in either study. The numerators differ, the account populations differ, and the periods differ. Two studies are not two measurements of one quantity.

Try it on the format figures and you can see the trap directly. Dividing Socialinsider's follower-based reels rate (0.52%) by Buffer's reach-based reels rate (3.31%) implies a reels reach rate of about 16%. Doing the same for single images implies about 8%. Neither implied figure matches Socialinsider's own published overall reach rate. That is not a contradiction to resolve; it is a demonstration that cross-study arithmetic produces confident nonsense. Run the bridge on your own account, where reach, followers and interactions all come from the same source, and it is exact.

April 2025 broke the impressions denominator for good

If you are following a formula that divides by impressions, on organic Instagram content that formula now has no denominator.

Meta announced the change on 8 January 2025 and it took effect on 21 April 2025. Organic impressions and plays were retired across the platform and replaced by a single unified views metric covering feed posts, reels, stories, carousels, photos and live video. Impressions still exist in the advertising surface. They no longer exist for your organic posts.

The distinction that survives, and the one you should build on, is this. Reach counts unique accounts that saw the content, each person once. Views count displays, and Instagram's own help documentation is explicit that views may include multiple views by the same accounts, which is what makes them different from reach. So views are always greater than or equal to reach, and the gap between them is repeat viewing.

Three consequences follow, and each one causes real confusion:

Any engagement rate calculated against views is now structurally lower than the same rate calculated against the old impressions figure, because repeat viewing inflates the denominator. Accounts that saw their reported rate sag in the second quarter of 2025 without any change in their content were often watching this and not a performance decline.

Comparisons across April 2025 are invalid. A 2024 views figure and a 2026 views figure are not the same measurement. If you keep a long-running spreadsheet, put a marked break at that date and stop drawing a trend line through it.

Reach became the more robust denominator by default. A single enthusiastic viewer can inflate your views; nobody can inflate your reach by rewatching. That is one practical reason to prefer reach-based rates for internal work, and there is a second reason coming in a moment.

The measurement side of this change deserves more room than it gets here. The dedicated guide on Instagram views, reach and impressions works through what each metric now counts and how to rebuild reporting around the surviving ones.

Instagram's own recommended metrics are reach-based

In January 2025, Adam Mosseri named the three signals that matter most for ranking: watch time, likes per reach, and sends per reach. He added that likes weigh slightly more for distribution to your followers and sends slightly more for distribution to people who do not follow you. No numerical weights have ever been published, and any guide quoting a conversion like "one send equals fifteen likes" invented it.

Read the phrasing carefully, because it is doing something the whole industry ignores. Instagram's head of product told creators to look at per reach figures. The platform's own diagnostic vocabulary is reach-denominated. Meanwhile the benchmarks everybody quotes, and the numbers brands ask for in media kits, are follower-denominated.

That mismatch is worth sitting with. The number that best predicts how Instagram will treat your next post is not the number the market uses to price you. Both are legitimate. They simply serve different masters: reach-based rates for understanding distribution, follower-based rates for commercial comparison. Keeping both, and never confusing them, is the whole discipline. The mechanics behind those signals are laid out in the guide to how the Instagram algorithm distributes content.

Benchmarks by account tier, and the trap in tier tables

Tier tables are everywhere and they are the most frequently misused artefact in this field. Here is a representative version of the ranges you will see quoted, with the column that is normally missing added back:

Tier	Followers	Commonly quoted range	Denominator usually implied	Reality check
Nano	1K to 10K	4% to 8%	followers, small and self-selected audience	plausible for a genuinely tight audience, rare for a business page
Micro	10K to 100K	2% to 4%	mixed, often unstated	above the published brand medians by several times
Mid-tier	100K to 500K	1.5% to 3%	mixed	compare against creators, not brands
Macro	500K to 1M	1% to 2%	followers	still well above the all-industry median
Mega	1M+	0.5% to 1.5%	followers	absolute volume is the real story here
Brand pages, any size	any	0.3% to 1%	followers, measured at scale	this is what the large studies actually found

Look at the bottom row against the rows above it. The creator tier tables and the large-scale brand studies are describing the same platform and they differ by roughly an order of magnitude. That gap is not evidence that creators are ten times better at content. It is population plus denominator plus survivorship: tier tables are typically assembled from creators who are already being considered for paid work, which is a filtered sample by construction.

Two rules make tier tables safe to use. First, only compare within a tier, never across one. Rival IQ's own framing is that the same 2% rate is unremarkable at 5,000 followers and outstanding at 500,000. Second, treat these ranges as orders of magnitude rather than targets. If you are at 0.9% on a brand page, you are normal. If a tool tells you that you are at 15%, either your audience is tiny and unusually loyal, or something in the measurement or the account needs explaining.

The absolute numbers underneath the percentages

Percentages hide the thing that actually pays. Socialinsider's tier data gives the raw averages per post, and they are sobering in a useful way.

An account with 1,000 to 5,000 followers averages around 3 comments on a reel, 1 save, and 580 views. An account with 100,000 to 1 million followers averages around 60 comments on a reel, 96 saves, and 16,035 views. The small account has the far better percentage. The large account has 28 times the views.

This is why "small accounts have better engagement" is true and useless as a business statement on its own. A percentage tells you about audience intensity. An absolute number tells you about commercial scale. A brand paying for reach cares about the second; a brand paying for trust and conversion cares about the first. That is precisely why budget has been shifting toward smaller creators: recent industry benchmark surveys report nano creators receiving the largest share of allocated budget, with more than half of brands expanding work with nano and micro tiers.

Benchmarks by industry, and why yours may not be listed

Industry moves Instagram engagement more than almost any other variable except format. From Rival IQ's edition covering data through early 2025, where the all-industry median stood at 0.36%, the spread ran like this: higher education led at 2.10%, roughly 5.8 times the median, with sports teams at 1.30%, influencers at 0.576% and nonprofits at 0.561%. Health and beauty sat at the bottom across every platform measured, with fashion, technology, retail and home decor below the median.

The mechanism is not mysterious. Higher education and sports have audiences bound together by identity, and identity produces comment threads. Nobody forwards a bank's post to a friend. Retail and beauty audiences are transactional: high purchase intent, low interest in conversation. That combination produces a low engagement rate on an audience that is commercially excellent, which is the clearest available proof that engagement rate is not a measure of audience value.

If your vertical is not in a published table, or your business sits at the intersection of two, the honest answer is to build your own comparison set. That process is covered at the end of this guide, and it is worth the afternoon it costs. A benchmark drawn from eight accounts that genuinely look like yours beats any published median for the specific decision you are about to make.

Format moves the number more than tier does

This is where the denominator question stops being theoretical and starts changing decisions.

Format	Follower-based, likes and comments (Socialinsider, 2025)	Reach-based, all interactions (Buffer, 2025)	Rank
Carousel	0.55%	6.90%	first on both measures
Reels	0.52%	3.31%	second by followers, last by reach
Single image	0.37%	4.44%	last by followers, second by reach

Reels and single images swap places depending on which denominator you use. That is not noise. It is the single most instructive fact in the whole benchmark landscape.

The reason is reach itself. Buffer's earlier reach analysis found reels earning roughly 1.36 times the reach of a carousel and 2.25 times the reach of a single image, and Socialinsider puts the average reels reach rate above 30%. Reels are the distribution engine, so they arrive at the reach-based formula with an inflated denominator. They spread wide, and a wide audience is on average less committed than a narrow one, so the interactions per person seeing them are lower. Meanwhile in the follower-based formula, reach never appears, so all that extra distribution shows up as pure upside.

Both facts are simultaneously true and they point at different tactics:

If your objective is discovery, reels are the correct choice even though their reach-based engagement rate is the weakest of the three. You are buying audience, not ratio.

If your objective is a defensible media kit number or depth with an existing audience, carousels win on both measures at once. Carousels also lead on saves by a wide margin; one large 2026 study of more than 24 million posts found carousels earning roughly nine times the saves of a single image.

If your reporting mixes formats into one blended average, you have destroyed both signals. Split by format before you split by anything else. The reasoning behind why saves and shares behave so differently from likes is unpacked in the piece on Instagram saves and shares as ranking signals.

A diagnostic tree for a falling engagement rate

Most people respond to a declining rate by changing their content. Roughly half the time the content was never the problem. Work through this in order and you will know which half you are in within twenty minutes.

Step one: compute both rates for the same window. Take your last twelve posts. For each, record followers at time of posting, reach, and each interaction type. Compute follower-based and reach-based rates. Take the median of each, not the mean.

Step two: compare the two trends. If the reach-based rate held steady while the follower-based rate fell, your content is fine and something happened to the denominator or to your reach share. If both fell, the content or the audience fit is genuinely in question.

Step three: split the reach. Instagram Insights separates reach from followers and from non-followers. A falling non-follower share means recommendations narrowed. A falling follower share means the connected side is delivering to a smaller slice of your audience, which usually tracks posting cadence and recent performance rather than any penalty.

Step four: look at numerator composition, not just its total. Likes holding while saves and shares collapse is a specific diagnosis: the content is still scannable but has stopped being useful or worth passing on. That pattern usually precedes a reach decline rather than following it, because those are the signals Instagram says it weighs most.

Step five: check for a break event. Did you run a giveaway, change topic, migrate analytics tools, or cross the April 2025 metric change? A denominator event looks exactly like a performance collapse on a chart and is fixed by annotating the chart, not by rewriting your strategy.

What you observe	Most likely cause	Check this first	Do not do this
Follower rate down, reach rate flat	denominator grew: real growth, a giveaway, or an acquired batch	followers gained in the window versus interactions gained	rebuild your content strategy
Both rates down together	content fit or audience mismatch	saves and shares per reach, by format	simply post more often
Views up, reach flat	repeat viewing by the same people	the gap between views and reach	report views growth as audience growth
Reach down across every format at once	distribution shift, seasonality, or recommendation eligibility	Account Status, which shows recommendation eligibility directly	assume a shadowban and start deleting posts
Likes steady, saves and shares near zero	depth problem, not a reach problem	save rate on your best-performing older posts	add more "like this post" calls to action
A vertical step in followers, then a flat line	an acquisition event with no engagement behind it	whether the date of the step matches the date the rate broke	present that follower count in a media kit unqualified

That fourth row deserves a note. Reach falling across all formats at once is the pattern most often labelled a shadowban, and Instagram does maintain a recommendation eligibility layer that can remove content from Explore and suggested feeds while leaving delivery to your existing followers intact. Account Status inside the app reports that state directly rather than leaving you to guess. The full picture, including what the term does and does not mean, is in the guide on how to detect and recover from an Instagram shadowban.

How purchased engagement changes each term of the fraction

This is the mechanical core, and it is worth being precise rather than moralistic. Every panel service touches one specific term of one specific formula. Once you can name which term, the effect stops being mysterious.

Service	Term it touches	Follower-based rate	Reach-based rate	Visible from outside
Followers	denominator	falls	unchanged	yes, directly
Likes	numerator	rises	rises	yes
Reel or post views	denominator of any views-based rate	unchanged	unchanged	yes, as a views to likes mismatch
Comments, generic	numerator	rises	rises	yes, and the text itself is the giveaway
Saves and shares	numerator	rises	rises	no, only in your own insights
Story views, profile visits	none, these appear in no engagement rate formula	unchanged	unchanged	no

Read the first row again, because it is the finding that matters most and it works against the interests of anyone selling followers.

Followers are the denominator. Adding followers lowers your follower-based engagement rate by definition. If an account at 12,000 followers with 321 interactions per post is running 2.68%, adding 5,000 followers who never interact takes it to 1.89% with identical content. Nothing about the content changed. The division changed.

That is decisive, because follower-based is precisely the formula an outside evaluator uses. Reach is invisible externally, so a brand assessing you has no choice but to divide by followers. Buying followers to look better to brands moves the exact number they compute in the wrong direction. It is one of the few places in marketing where the counterproductive effect is not a matter of opinion but of arithmetic.

The second row is the mirror image. Bought likes raise both rates at once, which is why they are effective at the one thing they can do (crossing a visual social proof threshold on a specific post) and why they leave the most characteristic pattern. A real audience produces uneven results: some posts land, others do not. A purchased top-up produces an unnaturally even line. The distribution of your likes carries more information than the level. The narrow cases where a like top-up is defensible, and the much wider ones where it is not, are worked through in the guide to Instagram likes strategy and the like ratio.

The third row surprises people. A view service inflates a denominator. If you or anyone else measures interactions against views, buying views lowers the resulting rate. Buying views to look popular and then reporting a views-based engagement rate is self-defeating in the same way as buying followers.

The last row is the one nobody mentions: story views and profile visits do not appear in any engagement rate formula at all. Whatever else they may do, they cannot move this number.

The ratio mismatches an audit catches first

The classic red flag is 50,000 views against 40 likes. Work out why it is a flag. Forty interactions against 50,000 displays is 0.08%. Typical reach-based engagement sits in the low single digits of percent, and views run higher than reach, but not by two orders of magnitude. A gap that size says one of three things: the views did not come from people who could engage, the audience was completely mismatched to the content, or the two numbers came from different sources.

But there is a 2026 correction to make here, and it cuts against a heuristic that circulated for years. The old rule of thumb was "100,000 followers and 200 likes means bought followers." Against a median that has fallen to 0.30% follower-based, 200 likes on 100,000 followers is 0.2%: below median, but no longer outside the range of a real, aging, disengaged audience. Organic engagement has declined far enough that the absolute level has lost most of its diagnostic power.

What retains diagnostic power is internal consistency. Auditors and the tools they use look for contradictions between numbers that should move together:

Flat like distribution. Real audiences produce variance. Near-identical like counts across every post is among the most reliable signals available, and it is the signature of a subscription service delivering a fixed quantity to each new post.

Like to comment ratio far outside the norm. Socialinsider's tier data gives an anchor: accounts at 10,000 to 50,000 followers average roughly 12 comments on a reel and 10 on a carousel. Thousands of likes with two comments is a mismatch that survives no explanation.

Audience geography that does not match the content. English-language content aimed at a specific market with an audience concentrated somewhere with no path to buy the product is the pattern brands open first, because the cheapest source accounts in this industry are concentrated in a small number of infrastructure countries.

Vertical steps in the follower graph followed by flat lines. A rapid jump with no corresponding jump in interactions is what growth-graph analysis is built to find.

Comment text quality. One 2026 study scoring 100,000 accounts across twelve indicators found comment quality to be the highest-accuracy fraud signal in its set, ahead of commenter authenticity and engagement anomalies. Generic praise unrelated to the post is not subtle to a classifier and it is not subtle to a human either.

Notice the pattern: none of these are about the level of your engagement rate. They are all about whether your numbers are consistent with each other. An account with a modest but internally coherent set of metrics passes an audit that an account with impressive but contradictory metrics fails.

Audience quality scoring: how brands actually check

By 2026 audience auditing is routine rather than exceptional at the brand end. Industry benchmark surveys report that roughly 60% of brands have encountered influencer fraud in campaigns, and that 56.5% of reported fraud and quality issues trace specifically to fake or bot followers, with only about one team in ten reporting no such issues at all. That is not an environment where an inflated media kit goes unexamined.

The two most commonly used scoring approaches work differently and you should know both.

HypeAuditor's Audience Quality Score runs from 1 to 100 and combines four components: engagement rate, the share of the audience judged to be real people, follower and following growth patterns scanned for anomalies, and engagement authenticity, meaning whether recent likes and comments come from people outside engagement pods and tag-to-win giveaways. It also classifies followers into categories, and one of them matters more than people expect: mass followers, meaning real humans who follow more than 1,500 accounts and are therefore statistically unlikely to see any given post. Those are not bots. They are real accounts that behave like dead weight in your denominator. The company's own guidance is that the score alone is not enough to decide on and should be read alongside the detailed report.

Modash takes a network-graph approach, scoring an account's behaviour against typical behaviour across the wider graph, flagging missing or generic profile pictures, abnormal following-to-follower ratios, account age, posting activity, empty bios, and unnatural growth curves. Its published thresholds are the most useful public reference point available:

Larger creators normally show 20% to 30% fake followers. Creators under 50,000 followers normally show 10% to 20%. Below 25% is broadly acceptable, above 50% is a category to avoid entirely. And the line that should reset expectations for anyone panicking over an audit result: a perfect score is itself unusual. Bots follow accounts on their own; a completely clean audience is not realistic and its absence is not evidence of wrongdoing.

There is one structural limitation worth knowing. If a creator's follower list is private, audience quality cannot be computed accurately at all. That does not read as clean, it reads as unverifiable, and in a shortlist process unverifiable and disqualified are frequently the same outcome.

The commercial logic underneath all of it was put well in one platform's 2026 pricing analysis: a creator with 100,000 followers and 5% engagement is worth more than one with 500,000 and 0.5%. Pricing follows engagement and audience quality, not follower count. Which means the follower number is a filter to pass, not a value to maximise, a point developed further in the guide to Instagram follower quality and what buying actually delivers.

The United States rule that turned fake followers into legal exposure

English-language readers operate under the strictest regime in the world on this specific question, and most of them do not know the rule exists.

The Federal Trade Commission finalised its Trade Regulation Rule on the Use of Consumer Reviews and Testimonials on 14 August 2024. It was published in the Federal Register on 22 August 2024 and took effect on 21 October 2024. Alongside its better-known provisions on fake reviews, including AI-generated ones, review suppression and paid reviews, the rule addresses fake indicators of social media influence directly.

The provision prohibits both selling and buying fake indicators of social media influence, meaning followers or views generated by bots or by hijacked accounts, where the buyer knew or should have known the indicators were fake and where they are used to misrepresent influence for a commercial purpose. Civil penalties run to $51,744 per violation.

Three points of precision, because overstating this would be as unhelpful as ignoring it.

The knowledge standard and the commercial purpose element both matter. This is aimed at commercial misrepresentation of influence, most obviously a creator inflating a media kit to win paid work or a business inflating its apparent authority to sell. It is not framed as a rule about a personal account with an ego-driven like count.

It applies to the buyer, not just the seller. That is what is genuinely new. Prior enforcement targeted vendors: the FTC's action against Devumi produced a $2.5 million settlement, a separate $50,000 settlement with the New York Attorney General, and a permanent ban on selling social media influence indicators. The New York action was described as the first law enforcement finding that selling fake followers is deceptive and that generating fake activity with stolen identities constitutes illegal impersonation. That was a case against a seller. The 2024 rule reaches the purchaser as well.

It is jurisdictional. Readers in the European Union operate under a different instrument: the Omnibus Directive amendments to the Unfair Commercial Practices Directive blacklist publishing fake consumer reviews outright, with penalties reaching 4% of annual turnover in the relevant member state, though that regime targets reviews specifically rather than follower counts. The FTC rule remains the most specific regulation anywhere naming social media influence indicators.

Set against the platform layer, the picture is consistent. Meta's Spam standard already prohibits "attempting to or successfully selling, buying, or exchanging for engagement, such as likes, shares, views, follows, clicks." What changed in October 2024 is that in the United States the buyer's exposure moved from a terms-of-service question to a civil penalty question when influence is misrepresented commercially.

Geography changes what normal looks like

The benchmark studies quoted throughout this article are dominated by English-language brand and creator accounts, mostly in North America and Europe. Instagram's actual population is not shaped like that, and English-language readers are the group most likely to assume otherwise.

India is Instagram's largest market at roughly 481 million users, growing 22.9% year over year, the fastest expansion of any large market. Its Instagram audience is approximately 69.7% male and 29.9% female, the most gender-skewed of any major market measured. The United States sits at about 182 million users, roughly 52.3% of the population, with an audience skewing female at about 54.5%. The United Kingdom is at 35.5 million, roughly 50.9% of the population, similarly female-leaning.

The received wisdom that "Instagram is a female-skewed platform" is therefore false in the single largest market on the platform and true in the two markets that generate most English-language commentary about it. That has three practical consequences for anyone reading benchmarks:

Your comparison set has to share your market. Engagement norms, comment behaviour and posting cadence differ substantially between markets. A benchmark drawn from US brand pages is not a benchmark for an account whose audience is primarily elsewhere, regardless of the language you publish in.

Audience geography is an audit field before it is a targeting field. Brands open the country breakdown early precisely because a geographic mismatch between content, product availability and audience is one of the strongest fake-follower signals available. This also means a legitimate account with a genuinely international audience should be ready to explain its distribution, because the pattern looks the same from the outside.

Country-targeted follower services do not solve this. Country targeting in this industry is real but shallow: it reflects the registration infrastructure of the source accounts, meaning the SIM or proxy country used when they were created, not the demographics of an actual person in that country. Paying a premium for country-matched accounts buys a better-looking geography column, not an audience.

One more piece of English-language context worth stating plainly, because it affects platform allocation decisions: in the United States, Instagram's Meta-reported ad reach of roughly 182 million sits below the self-reported audience figures for several other platforms. Those figures come from different methodologies and are not directly comparable, so no ranking claim should be built on them. But the confident assertion that Instagram is the dominant platform in the US market does not survive contact with the numbers.

Where a panel service honestly fits in an engagement rate strategy

Everything above should make the boundary obvious, but it is worth stating without hedging.

A panel service can do exactly one thing that relates to this article: help a specific piece of content clear a social proof threshold, the point at which a new visitor reads a post as active rather than abandoned. The psychological effect is real and well documented in the research on bandwagon cues; people evaluate an identical post differently at 40 likes and at 4. That is the whole of the legitimate case.

Here is what it cannot do, stated as flatly as it deserves:

It cannot raise your engagement rate if you buy followers, because followers are the denominator. This bears repeating because it is the most common purchase made for exactly the wrong reason.

It cannot produce customers. Purchased interactions are not fans. They will not reply, will not return, will not buy, and will not recommend you. If your funnel problem is at the profile-visit-to-enquiry stage, no metric purchase touches it. The realistic conversion arithmetic is worked through in the guide to turning an Instagram business account into sales.

It cannot pass an audience quality audit. If it leaves an inconsistent pattern, that pattern is precisely what an audit is built to detect, and the cost of failing one is usually not a single campaign but a note in the file at an agency that works with many brands.

It cannot be guaranteed. Meta's Spam standard prohibits buying and selling engagement, and the enforcement range runs from silently removing the purchased interactions through warnings, loss of recommendation eligibility, reduced distribution and monetisation restrictions. Drop is a normal outcome, not an anomaly: a platform-wide cleanup over roughly six hours on the night of 6 to 7 May 2026 removed enough inactive and automated accounts that very large accounts lost followers in the millions and small to mid-sized accounts commonly lost 2% to 5%. No supplier can prevent an event at that layer. Refill guarantees only exist on services explicitly flagged for them, they send a fresh batch rather than restoring the same accounts, and on services sold without a guarantee drop is not refunded. The mechanics of drop and what a refill window really covers are set out in why followers drop and how refill guarantees work.

It cannot fix weak content. This is the part that no amount of spending changes. A cosmetic layer over a good offer can accelerate a first impression. A cosmetic layer over an empty shop window is a decorated empty shop window.

If, having read all of that, a threshold purchase is still the right call for a specific launch, then it should be a deliberate decision made against a named metric. Reading what an SMM panel actually is and how it works before checking the live service catalogue and its pricing is a cheaper sequence than the reverse, because the catalogue answers "which is cheapest" and the concept page answers "which number am I trying to move." Only the second question has a right answer.

Agencies carry an additional obligation here. Any figure reported to a client has to be traceable to its source, and the most damaging mistake teams using panel infrastructure built for agencies make is presenting purchased volume inside an organic performance report. That reconciliation fails in month three, when the client asks why enquiries did not move with the chart.

Building a benchmark you can actually defend

Published medians answer a general question. The decision in front of you is specific. Building your own comparison set takes an afternoon and produces something no report can give you.

Pick 8 to 12 comparable accounts. Same vertical, same rough follower tier, same primary market. Competitors are ideal. Aspiration accounts three tiers above you are useless for this purpose.

Take the last 12 posts from each. Twelve is enough to survive one viral outlier and short enough to reflect current conditions rather than last year's.

Use likes plus comments over followers. Not because it is the best formula, but because it is the only formula computable from outside. You cannot see anyone's reach. Being consistent matters more than being ideal.

Take medians, never means. One breakout post moves a mean enough to make the whole exercise lie to you.

Split by format. Carousels and reels behave differently enough that a blended average tells you nothing actionable, as the format table above shows.

Record the date and the formula in the same file. Six months from now, that note is what makes the comparison valid. Without it you have a number with no provenance, which is where this article started.

Run your own numbers on both denominators. Track the reach-based rate for content decisions and the follower-based rate for external comparison. When they diverge, the reach rate bridge tells you which one moved and why.

Then write the formula into your media kit next to the number. Very few creators do this, which is exactly why doing it reads as competence. A brand that has spent a week comparing incompatible percentages will notice immediately that yours is the only one they can verify.

Frequently Asked Questions

What is a good Instagram engagement rate in 2026?

There is no single figure, and any answer that does not name a denominator is unusable. Measured against followers at large scale, the 2026 studies put the Instagram median between 0.30% and 0.48% depending on which population and numerator they used. Measured against reach with a wider numerator, Buffer's median is 5.46%. For a brand page, anything near or above 1% follower-based is strong; Rival IQ's top quartile sat around 1.02% in the previous edition of its report.

Why does my analytics tool report a different rate than my agency?

Almost always because one of you divides by followers and the other divides by reach or views, and often because one counts saves and shares while the other counts only likes and comments. Ask both parties to state numerator and denominator, then recompute on a single basis. In the example used in this article, the same post produced results from 2.12% to 9.44% depending on those two choices alone.

Should I use followers or reach as my denominator?

Both, for different jobs. Use reach-based rates to judge content, because that is what Instagram's own guidance points at when it tells creators to watch likes per reach and sends per reach. Use follower-based rates for brand deals and competitor comparison, because reach is invisible from outside and follower-based is the only rate an external party can compute.

Does buying followers lower my engagement rate?

Yes, mechanically and immediately. Followers are the denominator in the follower-based formula, so adding followers who do not interact reduces the ratio with no change to your content. An account at 2.68% adding 5,000 non-engaging followers to a base of 12,000 lands near 1.89%. Since follower-based is precisely the rate a brand evaluator computes, buying followers to attract brand work moves the number they look at in the wrong direction.

Can brands actually tell if engagement was purchased?

They can detect inconsistency reliably, which in practice amounts to the same thing. Audit tools flag near-identical like counts across posts, like-to-comment ratios far outside tier norms, audience geography that does not match the content, vertical steps in the follower graph followed by flat lines, and generic comment text. One 2026 study scoring 100,000 accounts found comment quality to be its single highest-accuracy fraud indicator. What passes an audit is coherence between your numbers, not an impressive level.

What percentage of fake followers is considered normal?

Modash's published thresholds are the clearest public reference: 20% to 30% is normal for larger creators, 10% to 20% for creators under 50,000 followers, below 25% is broadly acceptable and above 50% is a category brands avoid. Their own note is worth remembering, that a perfect score is unusual, because bots follow accounts unprompted and no organically grown audience stays completely clean.

Why did my engagement rate drop in 2025 when my content did not change?

Check the date against 21 April 2025. Instagram retired organic impressions and plays that day and replaced them with a single views metric that counts repeat viewing. Any rate calculated against that denominator became structurally lower overnight. Separately, follower-based rates fall arithmetically as an account grows, since the denominator expands faster than distribution does. Compare your reach-based rate across the same window to see whether anything real changed.

Do reels really have a lower engagement rate than photos?

It depends entirely on the denominator, and this is the clearest example in the data. Measured against followers, reels beat single images (0.52% against 0.37%). Measured against reach, single images beat reels (4.44% against 3.31%). Reels earn far more reach, so they carry a much larger denominator into the reach-based formula. Carousels lead on both measures at once, which is why they remain the safest format for a media kit number.

Is buying followers or views illegal in the United States?

The FTC's rule on consumer reviews and testimonials, effective 21 October 2024, prohibits both selling and buying fake indicators of social media influence such as bot-generated followers or views, where the buyer knew or should have known they were fake and used them to misrepresent influence for a commercial purpose, with civil penalties up to $51,744 per violation. It is aimed at commercial misrepresentation rather than personal vanity metrics, but the commercial fact pattern, inflating a media kit to win paid work, is squarely covered. Separately, buying engagement violates Meta's Spam standard regardless of jurisdiction.

Does a low engagement rate mean I have been shadowbanned?

Not by itself. Work the diagnosis in order: if reach held steady and only the rate fell, the issue is in your numerator or your follower count, not in distribution. If reach fell across every format simultaneously, check Account Status inside the app, which reports recommendation eligibility directly. Instagram restricts recommendations to non-followers under its guidelines while stating that it does not limit distribution to your existing followers, so a genuine eligibility problem shows up as collapsing non-follower reach with follower reach roughly intact.

Conclusion

The single habit that separates people who read Instagram data well from people who argue about it is this: when you see a percentage, ask what was on top and what was underneath. Nearly every expensive misunderstanding in this field, the mispriced creator, the panicked strategy pivot, the client report that unravels in month three, comes from comparing two numbers that were never measuring the same thing.

Once you can do that, the benchmarks stop contradicting each other and start being useful. Rival IQ's 0.30% tells you where brand pages sit against followers. Socialinsider's 0.48% tells you where a broader, creator-heavy population sits on a narrower numerator. Buffer's 5.46% tells you what happens when you switch to reach and count saves and shares. Three sentences, three tools, no conflict.

And when the question turns to whether to spend money on any of this, the arithmetic gives you a straight answer that no vendor is going to volunteer. Followers sit in the denominator. Buying them lowers the exact percentage a brand will compute about you. Whatever a purchased metric is for, it is not for that. Deciding what you are actually trying to move, and on which denominator, is the work that has to happen before you open any service catalogue and its pricing, not after.