Why Attention is Harder to Hold Than it is to Get

The scroll that almost stopped

You saw it and almost stopped. It had all the right ingredients: the image was sharp, the caption started well, and for a quarter of a second your thumb hovered. However, it kept moving. The post disappeared upward, replaced by the next one, and then the next, and you cannot remember what any of them said. Sounds familiar?

This experience is so ordinary that most people would not call it an experience at all. But it is the central event in modern marketing, and most brands misunderstand what happened in that fraction of a second. What happened is that your attention was captured but not held. These are two different cognitive operations, and the distance between them is where most social media strategies quietly fail.

Attention is harder to hold than it is to get because capturing it relies on automatic, low effort reflexes that every scroll triggers, while sustaining it demands continuous emotional and intellectual investment that the vast majority of content never earns. While the first is nearly free, the second is extraordinarily expensive and the gap between those two costs is growing wider every year.

Gloria Mark, Chancellor’s Professor of Informatics at the University of California, Irvine, has tracked this shift across two decades of observational studies. In her 2023 book Attention Span, she documents a decline that puts the problem into uncomfortable focus.

Gloria’s research found that the average time a person sustains focus on a single screen has declined from roughly two and a half minutes in 2004 to just 47 seconds. This was measured not in laboratory conditions, but in people’s real working environments, surrounded by the same interruptions and interfaces that surround your audience right now.

Two systems, one scroll

The reason capture and retention feel so different is that they are powered by different neural systems. The orienting response is an automatic, involuntary snap of attention toward anything novel, sudden, or visually salient. It evolved to detect threats. It requires almost no cognitive effort.

A bright thumbnail, a face mid-expression, a first line that opens with conflict: all of these trigger orienting, and they do so before your audience has made any conscious decision to engage. This is the system that most social media content is designed to activate, and it works. It is also nearly useless on its own.

Sustained attention is a different operation entirely. It runs on executive control, the brain’s frontoparietal network that must actively suppress competing distractions, maintain a goal, and decide, moment by moment, that the current input is worth continuing to process. This system fatigues. It is resource intensive. It is the system that decides whether someone reads to the end of a caption, watches past the first three seconds of a video, or returns to a page they bookmarked yesterday.

A 2023 study in the Journal of Marketing by Jonah Berger, Wendy Moe, and David Schweidel examined this gap directly. Analyzing over 600,000 reading sessions across 35,000 pieces of content, they found that what holds attention is not what captures it.

Processing ease and high arousal emotions sustained reading, while content that evoked sadness or cognitive difficulty caused readers to disengage even when they had already chosen to click. The implication is significant: the qualities that earn a click and the qualities that earn sustained engagement are not merely different in degree. They are different in kind. A content strategy optimized for the first may actively undermine the second.

A 2025 study published in Nature Human Behaviour reinforced this asymmetry from the social media side. Studying attention allocation patterns on social platforms, the researchers found that attention gained from highly appealing expressions is easier to acquire than to sustain, because it is highly sensitive to changes in expression modes over time.

The attention that feels effortless to earn is the attention that vanishes fastest. This is the distinction that separates content designed for directed attention from content designed merely for impression volume, which we covered in one of our previous pieces. One requires understanding how cognition works and the other, only a budget.

The retention economy

The platforms themselves have noticed the asymmetry, and they have responded by restructuring their algorithms to reward the expensive kind of attention.

Buffer’s 2026 analysis of over 52 million posts found that Instagram’s median engagement rate dropped significantly year over year, falling from the highest in its dataset to third place behind LinkedIn and Facebook, as the platform shifted its algorithmic weight from likes toward saves, sends, and watch time.

That decline did not happen because Instagram became less popular, but because the platform redefined what counts. As a consequence, likes, the original currency of social engagement, now carry less algorithmic weight than the signals that indicate someone actually stayed, processed, and valued the content enough to act on it. The platform is explicitly telling creators and brands that surface level interaction no longer earns distribution. Retention does.

The same pattern is visible across other major platforms. Sprout Social’s 2026 metrics guidance reports that engagement signals like dwell time and saves are now weighted more heavily than simple likes across multiple platforms, a shift that rewards depth of engagement over speed of reaction.

In parallel, TikTok has implemented a two stage distribution model that first tests content with a creator’s existing followers and only scales distribution if completion and rewatch rates meet threshold levels.

What these shifts mean in practice is that a brand can achieve enormous initial reach and still fail to generate any meaningful engagement. The content that stops a thumb is not the same content that holds a mind, and the algorithms are increasingly capable of measuring the difference.

This is the environment in which the old model of content production without content thinking becomes not just inefficient but actively counterproductive. Volume without depth now triggers the very algorithmic penalties that reduce future reach.

Why virality is a trap

A 2025 study in Scientific Reports, analyzing content timelines from over 1,000 European news outlets on Facebook and YouTube, found that most viral events do not significantly increase engagement and rarely lead to sustained growth.

The study employed a Bayesian structural time series model to evaluate what actually happens after a post goes viral. The results challenge a deeply held assumption in social media marketing: that viral reach is the highest form of success.

What the researchers found is that virality does not guarantee lasting engagement. When a viral moment follows a period of already growing attention, it typically represents the final burst of that growth cycle, not the beginning of a new one. Engagement afterward settles at levels lower than the phase that preceded the spike.

Only when virality arrives unexpectedly during a period of declining attention does it temporarily reactivate audience engagement, and even that effect is transient. The study also found that faster emerging viral effects fade more quickly, while slower, steadier growth patterns produce more persistent engagement over time. In roughly half of the cases studied, the impact either faded within the first week or never materialized at all.

The brands that chase viral moments are optimizing for the cognitive reflex that costs their audience nothing to give and, consequently, produces almost nothing in return.

Consistent, depth-oriented content that earns retention over time produces a different growth curve altogether: it is less dramatic, less shareable in screenshots, and far harder to point to in a monthly report. Nevertheless, it builds the kind of audience relationship that compounds: people who return not because an algorithm surfaced a post but because they remember the last one and want more.

This is the difference between a brand that trends and a brand that matters. While the first requires luck and timing, the second requires editorial intelligence and strategic patience.

What holding looks like

Let’s go back to that scroll. The post you almost stopped for was competing against everything your brain has learned to expect from a feed: novelty, brevity, disposability.

The post that would have held you is the one that offered something the next post could not replicate: not a louder hook, but a deeper reason to stay.

This is a design problem, not a volume problem. It cannot be solved by posting more frequently, by A/B testing thumbnails, or by chasing the format the algorithm rewarded last month. It can only be solved by building content that earns the sustained, effortful, cognitively expensive attention that audiences protect and platforms increasingly measure.

Because the brands that will matter in the next decade are not the ones that mastered the scroll stop, but the ones that gave people a reason to stay.