What are the benefits of personalization, and what are the risks of taking it too far?

ProcrustesPR

Personalization can filter large amounts of information to surface content that better meets users’ needs, reducing the effort of searching and choosing. But if a system repeatedly caters only to known interests, content may become monotonous, other needs may be overlooked, and users may even become reluctant to provide further feedback. Recommending content that matches someone’s interests is not the same as understanding the whole person. In its 2019 research overview, NN/g noted that people’s interests are multifaceted and change over time. The examples below reflect observations from that research, not necessarily how the platforms in question operate today.

ProcrustesPRProcrustesPRDoes my website need a speed score of 100?

Personalization can reduce noise, but browsing is not just about finding the right answer

When there is too much information to review item by item, filtering stories, products, or other information by interest can ease the burden on users. The benefit is not that less content is always better, but that useful information becomes easier to find.

However, looking for specific information and casually browsing call for different mixes of content. NN/g uses web search as an example: with a nearly limitless amount of information available, prioritizing relevant results helps reduce users’ exposure to low-value results. But when people are browsing content or following updates, focusing solely on matching every recommendation to known interests may limit opportunities for discovery. Designers need to weigh this trade-off rather than assume that recommendations should be narrowed in every context.

Relying too heavily on past interactions can leave other interests undiscovered

NN/g’s article offers a hypothetical example: someone likes both cats and dogs but only likes or shares posts about cats. If the system then recommends only content about cats, the recommendations are relevant, but they still overlook her interest in dogs and offer no opportunity to discover other interesting animals. A lack of interaction history does not mean a lack of interest.

This limitation has costs for both users and service providers:

  • Users repeatedly encounter the same types of content, with fewer opportunities to broaden their interests.
  • Service providers have fewer opportunities to learn about users’ other needs.
  • If old records continue to dominate recommendations, the recommended content may fail to reflect preferences that have changed.

For example, one Amazon user in the study said that her recommendations still included products she had viewed or purchased years earlier, which made the recommendations feel outdated. This was one participant’s experience and cannot be generalized to the entire platform, but it clearly illustrates the problem of recommendations that are not updated sufficiently.

Excessive personalization can cause content fatigue and may also affect engagement and trust

Some social platform users in the study said that the content they saw became increasingly similar, causing them to gradually lose interest. Some even stopped using the platforms. Another participant continued scrolling in search of interesting content despite finding Facebook boring. Continued browsing alone, therefore, does not necessarily mean that users are satisfied with the recommendations.

Users’ feedback may also become less complete. One user worried that liking posts would make future recommendations increasingly narrow and did not want to be defined by a single preference. Based on this, the NN/g author cautions that excessive personalization may change how users interact, which in turn affects the information collected about their interests.

Using information across contexts raises another question: are users comfortable with their information being used this way? Some study participants felt uneasy when content they had browsed on their phones appeared in ads on their computers. This reflects users’ perceptions, not technical proof of a tracking mechanism. NN/g recommends carrying information about a user’s activity across contexts only when users expect and welcome it. Users should also be able to flag annoying or inappropriate recommendations.

ProcrustesPRProcrustesPREight SEO essentials for beginners: Website speed

Personalization: perspectives from across the web

Does personalization necessarily hinder discovery?

Not necessarily; the key is how recommendations are designed. Serving content strictly based on past interactions may overlook interests users have not yet expressed. But another author argues that, while deep personalization may lead people to skip their usual discovery stage, it may also inspire them to explore new brands and experiences. Both outcomes are possible, so the degree of personalization alone cannot determine the result.

How might excessive personalization affect public discussion?

If platforms continually push content that aligns with existing beliefs, they may create echo chambers, reduce exposure to different opinions, and fuel polarization. NN/g also notes that when The New York Times announced experiments with personalized news in 2017, some readers worried about losing the shared experience of receiving the news. These are risks and reader concerns from that time, not a claim that all personalized news causes polarization.

How can relevance and content diversity be balanced?

Interest-based recommendations can be retained while broadening overly narrow user segments and deliberately mixing in novel, popular, or untargeted content. NN/g recommends adjusting the mix by channel: email and text messages can provide content aligned with known interests, while social feeds can make more room for discovery-oriented content. This is design guidance that still needs to be adapted to the context of use, not a rule that applies uniformly across channels.

Are standard personalization methods appropriate for sensitive topics?

The same algorithm should not be applied indiscriminately. NN/g recommends first identifying whether targeted content involves sensitive topics, reassessing the personalization approach, and taking a more conservative approach when necessary. Users should also have ways to provide feedback and some degree of control so they can flag problematic or annoying content. The level of sensitivity and the appropriate approach still need to be assessed in context.

ProcrustesPRWhat Ten Years of SEO Taught Me About Its Real Purpose

References

  1. The Dangers of Overpersonalizationwww.nngroup.com · 2019-01-20T17:00:00+0000 — Discusses: Does personalization necessarily hinder discovery?; How might excessive personalization affect public discussion?
  2. Google Is Becoming A Personalizing Mirror Before You Even Type A Querywww.searchenginejournal.com · 2026-06-18T11:00:31+00:00 — Discusses: Does personalization necessarily hinder discovery?

Related articles

Does my website need a speed score of 100?Eight SEO essentials for beginners: Website speedWhat Ten Years of SEO Taught Me About Its Real PurposeHow Do RWD and AWD Work, and Which Should You Choose for Your Website?What are the benefits and development challenges of SPAs?How do UTF-8 and Big5 differ in their effects on search rankings and user experience?What are single-page websites best suited for, and what are their pros and cons?Keep different topics on one site or split them up? Start with audience overlap and maintenance capacity
ProcrustesPR
ProcrustesPR