X shifts algorithm to reward dwell time over likes, forcing brands to rethink post length and format

On October 20, 2026, Elon Musk announced two updates to X's timeline recommendation system: a new AI ranking model built to better predict how long users will spend on content, and a retrieval model that favours posts people actually read to the end.

The change marks a deliberate move away from traditional engagement signals such as likes, reposts and replies, toward what X believes is a better proxy for content quality: whether someone actually finishes reading it. The underlying logic shifts from rewarding light engagement to rewarding deep reading, with the algorithm now directly predicting and allocating traffic toward posts with high dwell time and giving priority to content that users are likely to read in full.

For brands and social media teams, the immediate consequence is tactical. Posting strategies optimised around quick-hit engagement, likes and shares, or formats designed to game legacy signals, are now working against the system rather than with it.

What actually changed in the feed

The retrieval model pulls in posts with high completion, meaning posts that users tend to read all the way through rather than abandon halfway, while the new ranking model's job is to predict dwell time, the industry term for how long someone lingers on a piece of content.

X shipped major updates to its ranking model in January, May and August 2026, with the October changes targeting two specific levers: predicting dwell time and retrieving posts with high completion rates. This is the fourth major adjustment to the recommendation system this year, each successive update tightening the platform's focus on retention metrics rather than surface-level interaction.

Alongside the algorithm announcement, Musk shared a chart showing daily user active seconds on X, claiming that the metric is growing exponentially and reaching new highs almost every day. The chart suggests that time spent on the platform has increased, but the figures haven't been independently verified, the available data doesn't establish exponential growth, and it's unclear how much of that increase can be attributed to the latest recommendation models.

The change aligns X with the retention-focused algorithms already dominant on TikTok and YouTube. The X algorithm in 2026 has added a critical new variable: time spent on platform, with posts that cause users to scroll to replies, click through to threads, watch embedded videos, or visit profiles getting significantly more distribution than posts that simply get liked and scrolled past.

Why dwell time breaks lightweight content

Dwell time, the number of seconds a user spends looking at a post before scrolling past, carries a weight of +10 in the ranking system, the same weight as a bookmark and half the weight of a retweet. Dwell time on long-form content exceeding two minutes is an even stronger positive signal.

The implication is structural. Content designed to accumulate quick likes without holding attention, posts built around sensational hooks that lead nowhere, or formats optimised for virality rather than substance, now trigger the opposite of what they were designed to achieve.

Tweets that used sensational hooks to attract clicks but had low information density and relied on emotional engagement may now receive reduced traffic support under the new algorithm due to high drop-off rates. If people stay longer on a post, it signals quality to the ranking engine, while short dwell times suggest users are not engaged, leading to algorithmic suppression and a drastic reduction in reach.

If a user views a post for less than 3 seconds before scrolling, X records a negative quality signal known as Dwell Time Decay, and if average dwell time falls consistently, an account's global Quality Multiplier drops by 15 to 20 per cent.

The penalty is compounding. Accounts that consistently produce low-dwell content not only see individual posts suppressed but also suffer lasting damage to their account-level distribution potential.

Long-form posts now outperform threads

X now supports posts up to 25,000 characters for Premium subscribers, essentially full blog posts in a single post, and long-form posts get a single URL, accumulate all engagement in one place, and avoid the fragmentation problem that plagued threads where most people never made it past tweet 3.

If the first post in a 10-part thread does not generate enough engagement to clear the distribution threshold, posts two through ten are never seen, while long-form posts are evaluated as a single piece of content with aggregated dwell time, and the two-minute dwell time bonus gives them a significant scoring advantage.

Long-form single posts are generating 40 to 60 per cent more impressions than comparable thread content on average. That shift is not anecdotal. The format consolidates engagement, concentrates dwell time into a single measurable unit, and removes the drop-off risk that threads carry between each successive post.

For brands, the format decision is no longer aesthetic. Format changes outcomes, and in a controlled experiment, a single long-form X post reached a 7.03 per cent engagement rate, compared with 4.72 per cent for a thread on the same topic.

Our breakdown of YouTube's conversational AI shows how other platforms are also shifting discovery toward retention signals, making this a cross-platform trend rather than an X-specific experiment.

What brands should do differently on Monday

Stop writing for the like. Write for the person who stays.

Convert thread ideas into long-form posts, use clear headers, line breaks, and numbered lists within a single post to maintain readability, front-load the most interesting insight so the For You feed preview hooks readers, and if repurposing older thread content, combine the key points into one cohesive long-form piece instead of re-posting the thread.

Craft posts that hook the reader within the first second, using formats that are visually easy to scan, and present stories or open-ended questions that naturally keep users scrolling and reading, as creators need to master curiosity-driven writing, strong hooks, and well-structured content to keep attention past the crucial 3-second threshold.

Native video gets the strongest boost in the current algorithm, with short videos under 2 minutes 20 seconds performing best for initial distribution, so keep it native and don't link to YouTube, instead recording directly in the app or uploading the video file. Even 15-second clips outperform text-only posts if they hold attention, as watch completion time is a direct ranking signal.

Multi-image posts with 2 to 4 images often outperform singles because they increase dwell time. Visual formats that require users to swipe, expand or linger register as higher-quality content than static single-image posts that can be absorbed in under a second.

Snapchat's approach to blocking fully AI-generated videos from recommendations shows how platforms are becoming more selective about what qualifies as engagement-worthy content, a trend X is now reinforcing through dwell-time weighting.

Premium subscribers keep their distribution advantage

Premium posts receive a scoring bonus in the algorithmic feed. That advantage has not been removed by the dwell time shift. It has been layered on top of it.

The algorithm does highlight content from smaller accounts, but largely prioritises those with X Premium, meaning small creators can still go viral and grow valuable audiences, but they generally need to be verified subscribers to signal credibility to the ranking AI.

For brands running organic strategies without Premium, the dwell time update makes an already difficult distribution environment significantly harder. The penalty for posts containing external URLs was measurably tightened in early 2026, with analysis showing that posts with links in the main body now receive approximately 30 to 50 per cent less initial reach than equivalent posts without links, though the workaround of posting the link as the first reply remains effective.

TikTok's image search shift demonstrates how discovery mechanics are evolving across platforms, with each network building systems that prioritise native, high-retention content over external traffic drivers.