The Instagram we knew is gone. Viewers rewrote it. Median engagement fell by more than a quarter year on year, and the ranking factors the platform now highlights came from viewer decisions, not from an engineer.
Three independent sources support this thesis. The analytics platform Socialinsider, in this year’s benchmark, found that the median engagement rate on Instagram fell from 7.3% in 2024 to 5.4% in 2025, a drop of more than a quarter in twelve months.
Buffer, in its State of Social Media Engagement 2026 report, says Instagram performance deteriorated in 2025 across brand sizes, with engagement down roughly 24% year on year. Follower growth slowed at the same time, according to the analysis. A third measurement comes from Social Media Today, which reports a 28% year-on-year decline. Three different methodologies share one conclusion.
What drives the change
The numbers alone are not enough. What lies behind them is more interesting. The audience has become more selective. People finish less content, follow fewer accounts and quickly stop paying attention to anything that does not make sense to them. The algorithm simply copies the aggregated choice.
The London marketing community Pretty Little Marketer, which is followed by 369,670 users on LinkedIn, described the observation in the same way. “The Instagram you knew is gone, and it's not the algorithm's fault”, it wrote in a post from 20 April 2026. In the body of a thirteen-slide carousel, it explained how this change in habits shows up at feed level.
Pretty Little Marketer has a clear reading of this. “We got more selective, we stopped finishing content that didn't earn it, and we stopped following accounts with no clear point”, the post says. Socialinsider and Buffer figures show exactly how much that selectivity has affected engagement.
What Mosseri confirms
Adam Mosseri, head of Instagram, confirmed new ranking signals in January 2025. They are watch time, likes per reach and DM shares. Instagram moved watch time to first place and at the same time reformulated it. The algorithm now favours total time spent consuming content, not just the percentage of a video that is completed.
Each of these three signals is purely behavioural. Watch time measures how much time a viewer is willing to give something. Likes per reach measures what percentage of the people reached responded. DM shares indicate who finds content valuable enough to pass it on by private message. All three variables arise in the viewer’s mind, not in the platform’s code.
There is also a bonus for original content. According to Buffer analysis, original content gets 40 to 60% more distribution than reposts. Accounts with more than ten reposts within thirty days drop out of recommendations altogether. In doing so, the platform is listening to unspoken demand, which viewers express by skipping a second copy of the same video.
Your Algorithm as a formal admission
In December 2025 Instagram launched the Your Algorithm feature and is rolling it out globally during 2026. It is the biggest structural change to Reels recommendations in recent years. The name says it all. Everyone writes their own recommendation system by tuning their preferences.
It is the first official admission of what data and analyses have been describing for a year. The platform no longer tries to persuade viewers what they should see, but opens an editor where they can tell it themselves.
From lottery to editing behaviour
The old Instagram rewarded volume. Post often, post widely, and something will work. The new Instagram rewards completion, depth and spread between people. Each of those three signals stems from a viewer’s choice.
This is the shift that has been taking place on Instagram over the last twelve months. Data from Socialinsider, Buffer and Social Media Today show how deep it went. Mosseri’s update to the ranking signals codified it. Pretty Little Marketer captured it in the language of the community. The Instagram we knew is not gone because of a platform decision. It is gone because of the decisions of individual viewers added up into a mass.
Comments