Página 12 uses AI to predict headlines’ appeal

By Paula Felps


Nashville, Tennessee, United States


To improve its click-through rates, Página 12 wanted to find a way to give editors more information about how specific articles were performing — and learn how to present them in a way that would be more attractive to readers.

To make that happen, it was important to create a tool that could predict the probability of each article’s success, said Mariano Blejman, head of digital strategy at the Buenos Aires-based newspaper, which was founded in 1987 and is owned by Grupo Octubre.

Pagina 12 began using AI to help editors determine which headlines would get the greatest amount of attention.
Pagina 12 began using AI to help editors determine which headlines would get the greatest amount of attention.

To begin the journey, the company adopted a seven-step discovery process for working with the editors:

  1. Present their hypothesis.
  2. Adapt hypothesis based on any new information they discovered.
  3. Conduct trials using different data sets and test the hypothesis.
  4. Redefine the scope of the trial based on results.
  5. Find the best results.
  6. Review the results with the editor.
  7. Learn from the results.

Paloma Urtizberea Garcia, a freelance data science consultant who worked with Página 12, helped the company create a model that would predict the click-through rate (CTR) in headlines. During the first trial, she explained, they collected articles from the last two years and extracted words from Google Analytics to see how they performed. From there, they gained a better understanding of what title they wanted to be part of the model.

That helped create a model in which they achieved a 76% rate of precision, “which meant 76 of 100 headlines could predict if it would fail or be successful,” Garcia said. Then, they set up an online application that allowed them to play with the headlines and track their performance.

“Once they were published, they were matched with a historic base to see if they had low or high success probability,” Garcia said. “We had to manually track how the articles were doing.”

Then, they compared the CTRs to understand how headlines were helping the stories perform. Being able to give feedback to editors and bring in historical information using the CTRs of similar headlines was helpful in determining what their next steps should be.

Now, they use the platform taitel.news, which is available online and uses Artificial Intelligence (AI) to assist with headlines. Blejman said it has been an important step in changing how editors select headlines — and how readers respond.

 This case study originally appeared in the INMA report The Benefits and Risks of Media Data Democratisation.

About Paula Felps

By continuing to browse or by clicking “ACCEPT,” you agree to the storing of cookies on your device to enhance your site experience. To learn more about how we use cookies, please see our privacy policy.