Netflix does not wait long to form an opinion about its users. Within minutes of watching a single title on a brand-new account, the platform's recommendation engine begins reshaping itself, genre by genre, actor by actor, to reflect what it thinks a viewer wants to see next.
Our group set out to examine how that system works by starting completely from scratch. Instead of using an existing account, members created a new profile with no previous watch history, allowing us to track how Netflix builds its suggestions from the ground up.
Our group paid close attention to the "Gems for You" section, a feature designed to personalize recommendations, especially for users who have not yet accumulated much viewing history.
Throughout the audit, we deliberately watched content across a wide range of genres at random to observe how the algorithm responded. Selections included action, drama, sitcoms, sports dramas, romantic comedies, family movies, and general comedy.
Titles included "Trash," "All American," "The Lincoln Lawyer," "Young Sheldon," "Breaking Bad," "XO, Kitty," "Jumanji," "Anaconda," "Joe's College Road Trip," "Spider-Man: Into the Spider-Verse," and "Better Call Saul." For each title, we logged runtime, genre, IMDb rating, release date, and lead actors to identify patterns in subsequent recommendations.
Algorithm Responds Quickly to Viewing Shifts
The findings show that Netflix adapts fast. Recommendations changed almost immediately after each new viewing. After watching action-heavy films such as "Trash," the homepage is populated with high-intensity titles including "Scream," "Deepwater Horizon," "Blood Red Sky," and "Lucy."
When viewing shifted to serious drama series, specifically "Breaking Bad" and "Better Call Saul", the recommendations pivoted to crime dramas and serialized shows such as "Ozark," "Fear the Walking Dead", and "Southland." Lighter, family-friendly content produced a corresponding shift toward films such as "Jumanji."
The data suggest that Netflix relies heavily on genre and recent viewing behavior when shaping its recommendations. Whatever a user watches most recently appears to carry the greatest weight in determining what appears on the homepage next.
Popular actors also surfaced repeatedly in the algorithm's suggestions. Names such as Dwayne Johnson, Jack Black, and Bryan Cranston appeared across multiple recommended titles, indicating that star recognition may factor into what the platform chooses to promote.
A large share of recommended content was also newer or currently trending, suggesting that Netflix is not only personalizing suggestions but also steering users toward titles it has a commercial interest in promoting.
Users Describe Mixed Experiences With Personalization
Two Netflix subscribers shared their experiences with the platform's recommendation system, offering perspectives that align with the group's findings.
Lucas Rivero, an occasional Netflix user, said the platform's suggestions have become almost entirely genre-locked for him. "I'd say 90% of the shows I watch are anime, and when I'm looking for something else, I have to search for it," Rivero said.
His "Gems for You" section was filled almost exclusively with anime titles at the time of the interview. Rivero said the shift did not happen immediately; it developed gradually after he watched several series in that category, at which point the platform stopped surfacing content from other genres entirely.
"I don't think Netflix's recommendation system is very accurate. I watched one drama series, and my whole recommendation section was dramas, which I won't watch."
- Dilan Sulemen, Netflix subscriberHowever, he added that after recently diversifying his viewing habits, the recommendations have started to better match his interests.
System Shows Strength and a Notable Weakness
Our group's findings suggest that Netflix's recommendation algorithm is highly responsive but prone to overcorrecting based on limited viewing data. While the system does an effective job of quickly identifying viewing patterns, it may struggle to maintain a balanced content feed unless users consistently watch across multiple genres.
For users whose habits skew heavily toward one category, such as anime or drama, the algorithm appears to narrow its suggestions sharply, potentially reducing discovery of content outside that lane.
Netflix has not publicly disclosed the full details of its recommendation methodology. The platform's internal weighting of factors such as viewing recency, genre, actor popularity, and content promotion remains proprietary.
