Binge-watching by design: how Netflix and streaming apps keep you watching longer

Netflix coined the term "binge-watching" in a 2013 press release. The company was describing behaviour it had observed, but the framing also normalised something that Netflix's own design had made significantly easier to sustain than to stop. Some of what drives binge-watching is what viewers actually want. But a meaningful portion of it is what happens when the path of least resistance leads directly to the next episode, regardless of whether starting it was a considered choice.

On-demand TV streaming appears explicitly in the Flayelle et al. (2023) taxonomy in Nature Reviews Psychology as one of six digital contexts where platform design features have documented relationships with loss of control and compulsive use. It tends to get left out of conversations about addictive design, probably because the harms feel softer than gambling debt or adolescent depression. The design mechanics, though, are closely related to those in both of those contexts.

Autoplay: now with quantified effects

Until recently, most writing about autoplay was either critical commentary or platform marketing. In 2025, researchers at the University of Chicago produced one of the first experimental studies to actually measure its effect.

The study assigned 76 Netflix users to two groups. One group kept autoplay enabled. The other disabled it. Viewing behaviour was tracked over several weeks.

Results from disabling autoplay (University of Chicago, 2025):

• Sessions were around 18 minutes shorter on average
• Users took longer between episodes
• Participants described viewing decisions as more deliberate

A separate diary study by researchers at the Institute of Art, Design and Technology in Ireland, involving 101 participants, found that switching off autoplay significantly reduced binge-watching frequency, while having no meaningful effect on how mindful viewers felt about their watching habits.

The Chicago researchers classified autoplay as an "attention capture damaging pattern," a term from dark pattern taxonomy for design choices that extend platform sessions beyond what users would choose under normal conditions. The core observation was simple: autoplay does not deceive viewers about what they are watching. It removes the natural gap where the decision to keep watching would otherwise occur.

Netflix made autoplay optional in 2020 following sustained user complaints and coverage in the New York Times. It remains on by default.

Recommendations built for watch time

Streaming platforms describe their recommendation systems as helping users find content they will enjoy. That is accurate, but it leaves out the underlying calibration. These systems are primarily optimised for watch time, meaning the total minutes a user spends on the platform, rather than for satisfaction with what was watched or whether the user intended to watch that much.

The MDPI (2025) study on AI-driven platforms describes how this produces feedback loops over time. A user who watches a specific genre late at night will be served more of that genre at that time the following day. The algorithm learns preferences that include situational context, time of day, and emotional state, not just content format. A system optimising for watch time will learn to serve its most compelling content at moments when users are tired or emotionally susceptible, because that is when sessions tend to run longest.

Flayelle et al. (2023) specifically identifies personalised recommendations from content users did not actively seek out as a documented feature associated with uncontrolled use in streaming contexts.

Release schedules and the removal of natural pauses

Weekly broadcast television came with a structural gap: seven days between episodes, during which a storyline was left unresolved. That gap was an accident of broadcast scheduling, but it also functioned as a natural stopping point. Full-season drops replaced it with immediate access to the next episode at the moment of maximum narrative tension.

Research on binge-watching has documented how this interacts with cliffhanger endings. Unresolved narrative tension increases viewing compulsion independently of content quality. When that tension combines with zero-friction access to the resolution, stopping requires more active effort than continuing.

Some platforms have reintroduced weekly episode releases for selected series. The decision reflects marketing considerations as much as ethical ones: weekly drops generate sustained social discussion across a seven-day gap, while full-season releases concentrate cultural conversation into a shorter window and produce significantly more watch time in the first week.

Progress bars and the completion drive

Streaming platforms display a thin progress bar on episode thumbnails showing how far through a series a viewer is. The information is practically useful. It also activates what behavioural researchers describe as the completion drive, grounded in the Zeigarnik effect: the documented tendency to feel discomfort around unfinished tasks and to remember incomplete activities more vividly than completed ones.

"Continue watching" rows, viewing history, and profile-level progress tracking all serve a related function. They reduce the cognitive effort required to return to something started, making continuation the easier choice and starting something new comparatively harder.

A note on whether "addiction" is the right frame

Most people who watch more episodes than they intended are not experiencing anything clinically resembling addiction. They feel mild regret about the sleep they lost. Aagaard's (2020) paper on technology habits argues that applying the addiction label to this behaviour can pathologise something ordinary while leaving the design conditions that produced it untouched.

The more actionable question, he suggests, is whether platforms are designed with user wellbeing as a genuine constraint on product decisions, not just a statement in terms of service. Looking at default settings, recommendation calibration, and the deliberate removal of stopping cues, it is difficult to argue that wellbeing is currently a primary design constraint.

What the intervention research shows (Discover Mental Health, 2025 systematic review):

• Personalised, real-time screen time feedback tools were among the most effective interventions
• Effectiveness depended almost entirely on tools being set as defaults
• Opt-in versions of identical tools had negligible uptake

At FairPatterns, we believe that fundamental human rights like freedom, dignité and privacy, should not « dissolve in the digital world ». Addictive design is a predatory practice, preying on the people these platforms pretend to serve. That’s why we did 3 years of R&D to create the concept of “fair patterns” (interfaces that empower users to make their own, free and informed choices” and built a multimodal AI that scans sites, apps and social media to find and fix dark patterns and addictive design.
We’re building the Human Safety Tech architecture that’s now indispensable to protect humans online and when interacting with AI.

Sources: Flayelle et al., Nature Reviews Psychology (2023); Schaffner et al., University of Chicago (2025); Aagaard, Phenomenology and the Cognitive Sciences (2020); MDPI Youth study (2025); Discover Mental Health systematic review (2025); Netflix autoplay diary study, Institute of Art, Design and Technology (2024)