Instagram hits record user time as AI boosts Reels engagement

MUMBAI: Instagram is giving users more reasons to keep scrolling, and its AI is learning how to make every swipe count. The platform recorded an all-time high in global time spent per user in the third quarter of 2026, with year-on-year engagement growing at a double-digit rate across Reels and Feed as Meta fine-tuned its recommendation systems.

The gains suggest that Instagram’s battle for attention is increasingly being fought behind the scenes, through algorithms deciding not just what users see, but how quickly relevant content finds its way to their screens. Meta said upgrades to its recommendation technology helped increase time spent on Reels by more than 4 per cent during the quarter.

One change is particularly telling. Instead of evaluating each Reel in isolation, Instagram’s newer model architecture assesses sequences of videos, helping the system better understand what users are likely to watch next. The company has also incorporated richer watch-time signals to sharpen its predictions and improve the match between content and individual interests.

In other words, the algorithm is looking beyond a single video to understand the viewing journey. If a user watches several videos around a particular interest, the system can use those signals to make subsequent recommendations more relevant. The aim is to keep the feed feeling personal rather than like an endless stream of unrelated clips.

Freshness is becoming another important part of that strategy. Nearly 45 per cent of recommended Reels on Instagram were less than a day old in Q3, an increase of more than 11 percentage points from the previous quarter, according to Meta. The figure points to a greater emphasis on getting newer videos in front of users instead of relying solely on content that has already built up substantial engagement.

For creators, that could mean newer videos have more opportunities to enter recommendation feeds and find audiences. However, Meta did not quantify how the change affected individual creators’ reach or distribution, so it remains unclear whether the shift has translated into broader or more predictable exposure.

The company is also using AI to improve Instagram search, where the challenge is not simply finding content containing a particular word, but understanding what a post is about and whether it matches what someone is looking for.

Meta said it processes public Instagram posts daily through large language model-based content-understanding systems. Signals generated by these systems improved engagement with Instagram search results by 1 per cent through the end of Q3. While the increase may appear modest, it offers an indication of how improvements in content interpretation can influence the way users discover posts.

Meta also began shifting its content-understanding models to its own Muse model family during the quarter. In offline evaluations, the newer models outperformed previous production models, while reducing per-unit content-processing costs by more than 80 per cent across Instagram, Threads, Facebook and advertising, the company said.

That combination of better performance and lower processing costs matters as Meta applies AI across a vast volume of content. Understanding posts more accurately can help improve recommendations and search, while lower costs make it more practical to deploy those systems at scale. The company did not provide a separate breakdown of the financial savings attributable to Instagram alone.

The engagement push also has a commercial dimension. More time spent on Meta’s platforms creates additional opportunities to serve advertising, although keeping users engaged does not automatically translate into higher revenue without effective ad delivery and advertiser demand.

Meta reported that advertising revenue across its Family of Apps reached $63.1 billion in Q3, up 26 per cent year on year. Ad impressions across its services increased by 10 per cent, while the average price per ad rose by 14 per cent. The company attributed its advertising performance partly to stronger engagement and improvements in ad delivery and performance.

Instagram’s standalone advertising revenue was not disclosed in the earnings call, meaning the platform’s precise contribution to the Family of Apps total remains unclear. Nevertheless, its record time spent per user and double-digit engagement growth across Reels and Feed form part of the broader engagement picture underpinning Meta’s advertising business.

The figures also underline a change in how social platforms compete for attention. The contest is no longer just about who can produce the most videos or attract the biggest creators. It is increasingly about which platform can identify what a user wants to watch, surface it at the right moment and keep recommendations relevant as interests change.

For Instagram, the latest quarter suggests that AI is becoming central to that process, from recommending sequences of Reels to bringing newer videos into feeds and improving search results. The commercial test will be whether those improvements can sustain engagement and support advertising growth without making the experience feel repetitive.

For now, Instagram’s feed appears to be getting both smarter and fresher. The real measure of success, however, will be whether Meta can keep users coming back without leaving them feeling that the algorithm knows them a little too well.