The recent INMA Generative AI Master Class offered an opportunity to explore the “various ways in which this AI revolution is changing everything,” Sonali Verma, lead of the INMA Generative AI Initiative, said.

The key for news media companies is to learn how to prepare for these new news experiences and learn to serve audiences more effectively — opening the door for the opportunity to monetise them.

During the master class, media leaders from Dotdash Meredith, The New York Times, United Daily News, The Wall Street Journal, Newsday, and Ippen Digital shared how AI is enabling their companies to improve revenue and increase performance around business goals.

Dotdash Meredith leverages cookieless targeting

Dotdash Meredith uses a tool called D/Cipher, an “intent targeting tool,” that allows for users to be addressable without needing to be identifiable.

Joetta Gobell, senior vice president of data strategy and insights, said if someone is reading an article on TripSavvy about family travel, “we don’t need to know their age, gender, dietary restrictions, or health conditions to provide the information they need. We know in that moment of need why they’re on a Dotdash Meredith site and what they’re hoping to understand or achieve.”

Simply by visiting that site and reading that content, a user is providing information to D/Cipher. The tool then scans the entire Dotdash Meredith portfolio, using OpenAI to “understand the position in that vector database and multidimensional space” and find content areas across all sites that cover very similar things.

D/Cipher then looks at all of the family travel content and how users have interacted with it, so it’s not about where an individual user goes next — it’s about where anyone is likely to go next. They learn, for instance, that someone reading about planning a family trip is very likely to also read about tablets and gaming devices on LifeWire.

“We don’t make that assumption,” Gobell said. “Our users tell us this through millions of interactions every day.”

Because of Dotdash Meredith’s large and varied portfolio, D/Cipher also allows them to understand user behaviour across the “premium open Web,” which is of great value to advertising partners.

The bottom line, she said, is finding ways to use AI to “help us unlock the value we have in our existing expert content and make sure it’s actually aligning with users in their moments of need.”

The New York Times matches advertisers to the right audiences

“Every marketer knows how hard it is to capture the nuances of a brand and a target audience and how to translate those nuances into actionable targeting criteria,” Valerio Poce, executive director of ad product marketing at The New York Times, said.

BrandMatch was created as a solution to this problem, employing GenAI to bridge the gap between a brand’s unique identity and the practical implementation of targeting in advertising campaigns.

The BrandMatch process begins with a campaign brief provided by the advertiser, which contains the details and attributes of the brand's target audience and marketing goals.

BrandMatch then utilises GenAI to interpret these briefs deeply, moving beyond mere keywords to grasp the semantic and thematic layers of the content. The AI then scans through The New York Times’ extensive library of articles to identify those that align semantically with the concepts in the brief.

Once relevant articles are identified, BrandMatch analyses the readership data of these articles to construct a custom audience segment. This segment isn’t just based on basic demographic or psychographic data, but on the readers’ interactions and engagement with content that shares deep thematic connections with the advertiser’s brief.

“For click-through rate, we are finding that over half of all the campaigns that ran at least one BrandMatch line item, BrandMatch was the top CTR driver in the entire campaign,” Poce said, adding that when paired with native display formats, BrandMatch boosted CTR by over 32%.

AI powers contextual ads at United Daily News

When the likes of Google, Facebook, and other global tech platforms began stealing advertising dollars from United Daily News in Taiwan, they turned crisis into opportunity.

Anson Mok, head of data development, said that with a vast archive of content and ongoing news production, they knew they needed to leverage contextual and content-based targeting. They trained AI models to understand their content, identify topics, keywords and even the sentiment of certain articles.

“This allowed advertisers to place ads in contextually relevant articles,” Mok said. “If a piece was about hiking trails, for example, an outdoor gear ad could automatically appear there. Contextual advertising might sound old school, but AI gave it new life by making it far more precise and scalable.”

They’re also able to combine multiple points of data like demographics and behaviour to get a customised segment. Advertisers found these segments incredibly valuable, as well as the game changer for United Daily News which is their library of AI-driven predictive models.

These AI-driven ad campaigns are showing an average click-through-rate that’s more than 200% higher than their previous campaigns. Today, about 45% of direct ad sales come from AI-driven targeted advertising, an area that was netting zero for them until they launched this type of advertising.

“The beautiful part is that every campaign feeds back more data into our system, creating a cycle of improvement,” Mok said. “Our model learns and our team learns as well.”

The Wall Street Journal connects branded content with readers

For many companies, creating content isn’t the real challenge; getting it seen is. The Wall Street Journal parent company Dow Jones has put AI at the heart of a client-facing innovation to reshape how branded content connects with readers.

Thematic AI is a turnkey, scalable solution designed to dynamically match brand content with the most relevant reader at the most relevant moment. And it was built upon Dow Jones’ unique strengths: “We have an incredible platform and deep expertise in our own audiences,” Katie Weber, senior vice president of commercial strategy and head of financial services, explained.

The heart of Thematic AI is Dow Jones’ most powerful asset: first-party data. With more than 30 years of paywall experience, the company has a tremendous amount of knowledge about its users.

Since its launch, Thematic AI has powered more than 30 campaigns, each tailored to unique client goals. Some, like Barclays, focused on broader brand content. Others, like Global X and Invesco, used it to drive interest in specific financial products such as ETFs.

Each time, however, the campaigns surpassed expectations, with clients reporting higher engagement and a deeper understanding of what resonated with their audiences.

Newsday reduces push alert churn

Utilising machine learning has proved to be a worthwhile strategy for maximising user engagement and reducing churn around push alerts, John Callegari, senior digital editor at Newsday in the United States, said.

According to Callegari, Newsday initially limited their number of push alerts to not “overwhelm our users and cause them to unsubscribe from receiving alerts.”

Testing this theory, they increased the average daily number of alerts by 175%, resulting in a 38% increase in traffic to content. Alternatively, nearly four times the number of users unsubscribed than in the previous year.

Thankfully, Newsday’s alerts platform has a built-in predictive AI tool. This tool assisted their strategy by helping distill information from the largest audience subgroup regarding how likely they were going to unsubscribe and determining about half of that segment was considered a low risk for unsubscribing.

In 2024, more alerts were sent out, with a particular focus on the new subgroups based on data from predictive AI, Callegari said: “And we were able to continue to increase the number of alerts that we sent while also almost halving the number of unsubscribes we had in 2024.”

In challenging their assumptions about their initial strategy, Callegari said the company learned the importance of trusting machine learning and AI. To avoid falling into familiar patterns, newsrooms need to be open-minded and utilise predictive analytics in a thoughtful way to produce constructive outcomes, Callegari said: “Sometimes we just have to trust its suggestions. And in this case, we were able to really come up with a winning strategy for us.”

Ippen Digital improves workflows with agentic AI

Markus Franz, chief technology officer and IT chief architect at Ippen Digital in Germany, said it’s time for media brands to rethink how to collaborate with intelligent machines. Franz discussed how AI agents can help improve workflows for news media companies.

“We are building an exocortex,” Franz said. “It’s an external second brain, like your co-worker, for example, with you working with that and that runs on an AI-powered publishing substrate. It may be your own publishing infrastructure. It may be a collaboration with OpenAI, Anthropic and so on.”

From Franz’s perspective, the machines get “better and better” when agentic workflows are built: “You always have your task, but the task will be divided into smaller prompts. And we are chaining this kind of prompt to generate that kind of output.”

For example, if you are planning to do deep research on a topic, you would start with a variety of smaller prompts, and then combine the output (responses) to form a summary of the overall research.

The example Franz shared involved a special interest article that the workflow produced about a Netflix announcement. Training one agentic workflow produced an announcement that a human would have likely created. The agentic workflows, when trained properly, can improve editorial workflows and support content development and production.

“And it’s software, right? So, it’s really software,” Franz said. “And if you have that software, we can extend that kind of software and that kind of capabilities to get a better and better story with a less amount of time.”

The agentic workflows, when trained properly, can improve editorial workflows and support content development and production.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .