Why the Internet Needs Content Labels

In the early 1900s, an American chemist by the name of Harvey W. Wiley led the famous “Poison Squad” experiment; volunteers ate meals laced with chemical preservatives to test their effects on human health. His studies revealed the dangers of unregulated food additives and sparked public awareness about the importance of knowing what was in the food they consumed. But it wasn’t until almost 90 years later that the Nutrition Labeling and Education Act made it mandatory for companies to disclose what was really inside the boxes, cans, and bags that lined grocery store shelves in the U.S. Suddenly, consumers had clear information to help them make healthier choices.

Fast-forward to today, and it’s hard to imagine navigating a supermarket without these labels. Yet when we browse the internet—another essential marketplace in our lives—there is little guidance about what we are “consuming” with each click. We’re absorbing vast amounts of digital information, often without any idea of its emotional or mental “nutritional value.” But what if every webpage came with a content label, offering insight into its potential emotional impact, knowledge value, and practical usefulness?

Just like nutrition labels help you make better choices about what enters your body, content labels can help you take control of what enters your mind, potentially reducing harmful behavior such as doom-scrolling and time wasted on mindless browsing.

Read More: How to Stop Doomscrolling and Find Meaning on Social Media

What should “content labels” list? Science indicates that individuals care about three key features when searching for information: how practical the information is in guiding decisions and actions, its ability to enhance understanding, and how it will impact mood. On average, people select to engage with information that is useful, most likely to increase knowledge and improve their mood. But just as certain individuals tend to make nutritional choices that are unhealthy for their body, certain individuals tend to browse the web in a way that is unhealthy for their mind.

A study we conducted, published today in Nature Human Behavior, revealed that individuals with worse mental health tended to seek out more negative and fear-inducing content when they went online, which in turn exacerbates their symptoms, creating a vicious feedback loop. The study involved over a thousand participants who gave us access to their web browsing history. Using natural language processing techniques, we assessed the emotional tone of the websites they visited. Not only did we find a correlation between people’s mental health symptoms and the type of content they consume online, but we were able to verify a causal relationship. We did this by running a study in which we manipulated the websites people consumed. We show that making people consume negatively valanced webpages led to worse mood, which in turn led them to browse more negative content.

To break this cycle, we introduced content labels that provided users with emotional “nutrition information” about the websites they were about to visit. Would this webpage make you feel better or worse? Was it full of negativity or likely to brighten your mood? When users were shown these labels in the search-engine result page, they were far less likely to click on negative content. In turn, this led to improved mood after browsing.

But you would not want people to make decisions based solely on the emotional tone of a website. It is often crucial to be knowledgeable about the world, even if it makes you feel sad or anxious.

And so, we developed a plugin that provided users with not only labels regarding the emotional tone of a website, but also with scores based on how practical and helpful the information on a webpage is, andits ability to enhance understanding. The scores, which are presented on a search engine result page, are calculated using a machine learning algorithm that was trained on thousands of human scores. Of course, what is helpful to me might not be helpful to you, and what makes me sad might make you happy. But the data shows these scores do reflect the human experience on average.

Anne-Linda Camerini, who studies digital media and mental health, noted that while much of the focus on mental health and technology has been about limiting screen time, the type of content we consume is just as important. With platforms like X and Instagram are under fire for amplifying negative emotions, a content labeling system could offer users a much-needed tool to make healthier digital choices.

Read More: Misinformation Is Exhausting. Listening Helps

In fact, U.S. Surgeon General Dr. Vivek Murthy recently proposed warning labels for social media platforms, much like cigarette packages have, to alert users to the risks of overuse and exposure to harmful content. While Murthy’s approach is bold, it may not be enough. We don’t just need warnings—we need guidance. Content labels provide that, offering a way for users to proactively steer their digital diets, rather than simply be told what to avoid.

This shift—from focusing on screen time to focusing on content—could change the conversation around mental health and technology. Instead of framing the internet as inherently harmful, content labels offer a way to curate a healthier digital experience. Just as nutrition labels empowered people to make better food choices, content labels could empower users to make smarter, healthier choices online.

As these tools become more sophisticated, they might soon be integrated into our browsers, social media platforms, and search engines. Just as we now glance at nutrition labels without a second thought, we could one day check content labels before diving into a news article or blog post.

It’s time to put content labels on our digital diet, guiding us toward better mental health and a more mindful way to browse.

Why the Internet Needs Content Labels

Why the Internet Needs Content Labels

Why the Internet Needs Content Labels

Why the Internet Needs Content Labels

Why the Internet Needs Content Labels