On 24 September 2025, Mark Zuckerberg and Adam Mosseri announced that Instagram had passed 3 billion monthly active users. That single number disposes of the literal reading of the headline. Something is genuinely dying, though, and it is worth being precise about what: the public-feed playbook most brands still budget for.
Key takeaways
- 3 billion monthly active users announced on 24 September 2025, so the platform is not shrinking in audience terms.
- Socialinsider’s benchmark study, published 20 February 2026 across 35 million posts, puts engagement per follower at 0.48% for Reels, 0.50% for carousels and 0.33% for images.
- Mosseri has repeatedly named the same signals: average watch time, likes per reach and sends per reach.
- Widely quoted “engagement collapse” figures disagree with each other because they measure different bases. Check the denominator before quoting one.
What the benchmark numbers actually say, and why they conflict
Organic engagement per follower is falling, consistently and across formats. The clearest public dataset comes from Socialinsider, published on 20 February 2026, built on 35 million Instagram posts from 447,613 active pages over January to December 2025. It measures engagement as likes plus comments divided by follower count.
On that basis, carousels lead at 0.50%, Reels sit at 0.48% and single images trail at 0.33%. Year on year, Reels moved from 0.52% to 0.50% and images from 0.45% down to 0.37%, while carousels held flat at 0.55% before the more recent quarterly readings.
Now the part that matters for anyone about to put a slide in front of a client. That same report carries an explicit caveat: the values labelled 2026 are in fact 2025 data, because there was not enough of the new year available at publication. Meanwhile other widely circulated reports describe engagement falling from 7.3% to 5.4%, a completely different order of magnitude. Neither is lying. They are dividing by different things, one by followers, the other typically by reach on a narrower influencer sample.
| Format | Engagement per follower (Socialinsider, 2025 data) | What it is still good for |
|---|---|---|
| Carousel | 0.50% | Depth with people who already follow you |
| Reel | 0.48% | Reaching accounts that do not follow you |
| Single image | 0.33% | Brand consistency on the profile grid |
The practical instruction is dull but useful: never quote an Instagram engagement benchmark without naming its denominator, its sample and its period. Half the panic in this category is two studies talking past each other.
The app reorganised itself around DMs, and most plans did not follow
The most revealing product decision of the last two years was not an algorithm tweak. It was Instagram moving the content upload button out of the bottom navigation bar and putting a direct link to Direct Messages in its place, with Mosseri describing the app’s focus as DMs, Reels and recommendations. Product real estate is a stated priority, not a guess.
The ranking guidance points the same way. Mosseri has told creators to watch average watch time, likes per reach and sends per reach, and has been explicit that sends correlate with overall reach more than anything else, because a send is a proxy for content someone wants to talk about with a friend. He has also drawn a useful split: likes matter slightly more for reaching your existing followers, sends matter slightly more for reaching people who do not follow you.
That distinction should change how briefs are written. A post engineered for applause and a post engineered to be forwarded are different objects. The first ends in a like, the second ends in a conversation you cannot see and largely cannot measure.
Five changes worth making to a 2026 plan
- Report sends per reach and average watch time alongside the usual engagement figure. If your dashboard does not surface them, that is the first fix, because they are the signals the platform says it uses.
- Stop treating follower count as a proxy for distribution. Reach is decided post by post, and a large account with weak send rates now behaves like a small one.
- Write for the forward. Ask of every asset whether a specific person would send it to a specific friend, and who those two people are. A vague yes means no.
- Split objectives by audience type. Carousels for depth with followers, Reels for reaching non-followers, and separate success criteria for each rather than one blended engagement target.
- Date every benchmark you cite. Ranking behaviour and reported rates both shift within a quarter, and a figure without a date is a liability in a client deck.
Two caveats we would keep on the record. Nothing above guarantees a result, and any agency promising a specific lift from these changes is selling certainty that this platform does not offer. And if your audience skews young, the consent and transparency rules around advertising to minors vary by jurisdiction, so targeting decisions belong with your legal counsel rather than in the media plan alone.
Quick answers to what clients keep asking
Should we move budget to another platform?
Not on the basis of engagement rate alone, since the metric is falling as a structural consequence of feed dilution rather than as a signal that your audience left. Move budget when your own attributed conversions move, not when a benchmark chart does.
Is organic still worth staffing?
Yes, but with different outputs. Organic on Instagram now earns its keep through sends, saves and DM conversations rather than through visible vanity counts, which means the reporting line has to change before the headcount question can be answered honestly.
Do we need to post more often?
Frequency is not among the signals Instagram has named. Publishing more of what does not get forwarded simply produces more posts that do not get forwarded. We would fix the send rate on a smaller volume first.
What about AI-generated content?
Instagram has signalled that it intends to weight authentic human content more heavily going into 2026. We would treat that as a direction of travel rather than a settled ranking rule, and we have not seen a public dataset that quantifies the penalty. That is a genuine unknown, not a secret.
The honest summary is that “Instagram is dying” is a headline about a marketing playbook, not about a platform. Three billion people are still there. They are just in the inbox, and the work now is producing things worth sending. Our look at the psychology behind scroll-stopping ads covers the creative side of that problem in more depth.
Wondering why the younger half of that audience never reacts?
The send-not-like behaviour shift is sharpest among under-25s, and it has consequences well beyond Instagram.
Sources: TechCrunch report of the Instagram 3 billion monthly active user announcement by Mark Zuckerberg and Adam Mosseri, 24 September 2025, including the bottom navigation change; Socialinsider Instagram benchmarks published 20 February 2026, covering 35 million posts from 447,613 pages over January to December 2025, engagement measured per follower; Instagram and Social Media Today reporting of Adam Mosseri’s stated ranking signals (average watch time, likes per reach, sends per reach). Benchmark figures vary substantially between studies depending on the denominator used. Platform behaviour in this category changes quickly. Updated August 2026.
