illustration: DALL-EHeadlines on the internet have become longer, more negative, and surprisingly similar to clickbait. Research conducted by Pietro Nickl, Mehdi Moussaïd, and Philipp Lorenz-Spreen from the Max Planck Institute for Human Development shows that since the beginning of the 21st century, there has been a significant change in how online media headlines are constructed.
To analyze around 40 million headlines from four major English-language sources - The New York Times, The Guardian, The Times of India, and ABC News Australia, researchers used NLP methods, sentiment analysis, and regression models to detect language changes. The results were published in the article "The Evolution of Online News Headlines" in Nature Humanities and Social Sciences Communications, revealing a clear trend toward more clickable, emotional content.
Longer and more emotional headlines
Over two decades, online media headlines have increased in length by an average of 25%. In 2000, a headline in The New York Times typically took the form of a short sentence fragment, like "Shell`s future in Nigeria in doubt." Today, they more often take the form of a complete sentence, similar to clickbait from Upworthy: "I`m No Supreme Court Expert, But I Kinda Think You Shouldn`t Be Able To Pay For This?".
The report shows that full sentences and questions are replacing traditional short headlines. This phenomenon is driven by growing competition for readers` attention online. Algorithms favor content that triggers reactions, such as clicks or shares, giving an advantage to longer and more emotional headlines.
Clickbait not just in tabloids
Interestingly, these changes are not limited to low-quality media. Researchers observed similar trends in high-reputation outlets like The Guardian and The New York Times. Negative headlines are now more common than positive ones, further emphasizing the dominance of emotion over facts.
- The number of headlines with a negative tone has increased by 30% over the past 20 years.
- More headlines include personal pronouns and questions to pique the reader`s curiosity.
- The number of headlines containing words like "why" and "how" is rising, indicating an attempt to draw readers into the narrative.
Algorithms and A/B strategies
Media are increasingly using A/B testing to optimize headlines. An example is Upworthy, which experimented with different variants to maximize click-through rates. From 2013 to 2015, this model dominated Facebook until the algorithm changed to favor more engaging content rather than mere clicks. As a result, clickbait-style headlines became somewhat less popular, although they are still widely used.
Although clickbait styling offers tangible benefits in terms of higher click numbers, it also raises ethical and social issues. It undermines trust in media and contributes to misinformation. Moreover, promoting negative content encourages social polarization, especially in media with extreme political sympathies.
| Headline Features | Increase from 2000 to 2025 |
|---|---|
| Length (number of words) | +25% |
| Negative wording | +30% |
| Personal pronouns | +20% |
| Questions and interrogative pronouns | +15% |
Research results indicate that these changes occur regardless of journalistic quality or political sympathies. This means that even reputable sources are adapting to the rules governing the online attention market, where the most engaging headlines prevail.
Can it be changed?
Experts believe that changing the algorithms that promote content on social media could solve the problem. An example is The Guardian, which placed a "Most Read" section next to "Most Shared" on its website, allowing users to make more informed choices.
A list of potential solutions includes:
- Increasing algorithm transparency on social media platforms.
- Media literacy education to help users distinguish reliable content from clickbait.
- Promoting socially valuable content rather than just visually appealing material.
Headlines on the web have ceased to be just information. They have become a product designed to sell emotions and clicks. While this trend may seem irreversible, it is essential to remember that changing algorithms could bring a more balanced approach to creating and consuming content online.
The research report "The Evolution of Online News Headlines" is available at
https://www.nature.com/articles/s41599-025-04514-7
COMMERCIAL BREAK
New articles in section Media industry
Who trusts articles written by artificial intelligence? Results of a study from China
Krzysztof Fiedorek
Young readers trust machines more than human writers. When they consume news, they prefer the cool objectivity of algorithms over human emotions. They believe that articles created by AI are objective. As they interact with these texts more frequently, they rarely notice any harmful biases in them.
Who really controls the news on the internet? A study from Spain
Krzysztof Fiedorek
Analysis of more than thirteen thousand articles reveals behind-the-scenes mechanisms of spanish digital media. Left-wing portals are responsible for 50.33 percent of texts on lobbying. The energy lobby generates 63.41 percent of publications. Only one daily newspaper evaluates pressure groups positively.
Children and artificial intelligence. The generation of a global experiment is growing
Joanna Sokołowska, UNICEF Polska
A UNICEF analysis based on new data from 10 countries shows that at least 20 million children have already used artificial intelligence (AI). Many of them are ahead of adults, adopting these technologies at a pace more than three times faster.
See articles on a similar topic:
Print Advertising in Poland. Analysis by Polskie Badania Czytelnictwa
BARD
Analysts at Polskie Badania Czytelnictwa (Polish Readership Surveys) have compiled statistical data highlighting the reach, engagement, and opinions regarding print advertising campaigns. These insights are valuable for marketers aiming to plan and optimize advertising budgets.
Repression Against Media: Committee to Protect Journalists Report for 2024
Krzysztof Fiedorek
In 2024, at least 361 journalists worldwide were imprisoned, often for exposing the truth. In China, reporters are tracked using advanced facial recognition systems, in Israel, Palestinian journalists are jailed without trial, and in Myanmar, journalist Shin Daewe received a life sentence for... a drone.
Journalism in the age of AI. Why people prefer humans over machines
Krzysztof Fiedorek
Only 12% of people accept news created solely by AI, while 62% prefer those written by humans. At the same time, only 19% notice labels indicating the use of artificial intelligence, while younger audiences ask AI to explain the content to them. These are the findings of the Reuters Institute report on artificial intelligence in media.
Artificial intelligence in newsrooms. Three realities of the AI era in media
Krzysztof Fiedorek
According to a report by the European Broadcasting Union, many newsrooms already use AI but still do not fully trust it. Audiences do not want "robotic" news, and the technologies themselves though fast can be costly, unreliable, and surprisingly human in their mistakes.




























