Here’s a look at a smattering of GenAI use cases our peers are working on in the news media business — both in the B2B space and in the B2C space.

Some are pure engagement plays, whereas others are for paying subscribers. All of them point the way towards showing value to our audiences through reliable tools that can be built quickly or inexpensively.

The trend is clearly towards personalisation. In each case, the user finds exactly what they are looking for in a format that is tailored to their needs — rather than a one-size-fits-all product that is handed to them and is identical to that handed to every user.

Take what you need, make what you want

Dow Jones’ risk and compliance division created a product that lets its clients build an investigative due-diligence report from several sources within minutes. 

The product “extracts and summarises vast volumes of information into intuitive reports that are quick and easy to digest. The platform significantly reduces false positives, saving compliance professionals valuable time and resources,” according to the company. 

“The outputs are fully sourced and auditable, with links to the original articles and records for further interrogation.” 

The Washington Post is working on a product that lets its readers interrogate its archives. What is particularly interesting is that it is partnering with a university to develop this — two doctoral students have started a year-long research and development effort, with three Virginia Tech faculty members supervising.

This is a common practice at some other news organizations as well, such as The Globe and Mail in Canada, Schibsted in Norway, and Mediahuis and DPG Media in Belgium. It gives them access to innovative research, often at a relatively low cost.

Yahoo is using AI to rewrite click-bait headlines if they are misleading or unduly sensational.

The Associated Press is using AI-powered search for its videos, letting clients find individual moments within a video clip, even if they have never been tagged or captioned. “Rather than a traditional metadata search, the tool understands descriptive language and produces search results based on the description a user provides,” the company said.

Outside Magazine built a chatbot called Scout to ask specific questions, such as: “Can you recommend a hiking trail in San Francisco under four miles with a view of water?” It is powered by ChatGPT, but draws only from articles written by journalists at Outside Magazine, Backpacker, Climbing, and other publications in the Outside network over the past 20 years. It can answer questions about trip planning, outdoor gear, or meal planning while exploring nature. 

Scout was initially created to serve only subscribers, but it now appears on the top navigation bar of Outside’s Web sites and serves as the primary search engine across the network. The next step is to build new GPS-related features: “Imagine asking Scout for a mountain biking route and receiving a turn-by-turn Trailforks map, GPS coordinates, and a destination article from Pinkbike,” the largest mountain bike community in the world, the company said.

Hearst Newspapers created a quiz-generation engine called Emcee. “Emcee (M.C. = multiple choice) is built on the back of our newsrooms’ reported, edited and vetted work. No hallucinations here. We scrape our most popular stories, use GenAI to turn the stories into a multiple-choice format, spin up one of the DevHub’s interactive Quiz templates, and alert the editor, whose job is to read through the questions and answers, edit for taste and accuracy, and then publish the quiz manually,” said Tim O’Rourke, vice president of content strategy at Hearst.

“No direct publishing to the Web; a human is in the loop and in control of every piece of content.” Hearst’s newsrooms are using this technology to create a weekly programme to engage new readers. 

ICYMI: During a recent Webinar, The Times UK and Der Spiegel shared with INMA smart ways in which they are using GenAI to help their newsrooms.  

News publishers’ content, UX gets personal through AI

Harvard Business Review is thinking about “hyper-personalisation” — about the extent to which an editor curates an issue of a magazine versus the reader creating their own issue based on themes that are relevant to them. 

“We see hyper-personalisation driven by AI as a game changer. It will allow us — alongside editorial curation — to deliver highly relevant and individually tailored content and experiences at scale, whether it’s bespoke content recommendations or personalised learning pathways for our subscribers,” Sarah McConville, co-president of Harvard Business Publishing, told INMA.

The publication is also working on a bot that helps readers get feedback on difficult conversations and how to prepare for career transitions. 

“Readers turn to HBR for proven ideas to help them run their companies and their careers more effectively. By providing helpful, on-demand answers to leadership questions, HBR’s bot goes beyond search to offer a new way for subscribers to interact with, learn from, and tap into HBR’s rich well of content going back decades — including its flagship magazine, content on HBR.org, video, and podcasts,” McConville said.

The Harvard Business Review already has an app that offers some personalisation, which the user can direct, and a tip of the day that provides a summarised version of management advice.

When it comes to consuming content in a form that is personalised to the user’s needs, the reader is spoilt for choice at Clarin. If you haven’t yet checked out Clarin.com’s reading assistant, UalterAI, take a look now. It appears in all articles as a widget that the reader can click on. Then, a summary appears, covering 10% to 20% of the original text’s length.  

It also presents five types of analysis modules:

• The first is bullet points — a list of up to 10 concise sentences that delve into the summary information.

• A second analysis module extracts quotes and presents them alongside the names of the individuals who made those statements. All data, numbers, and figures in the text are extracted, classified, and presented in two-column tables for ease of comprehension. 

• Information can also be read as a FAQ, a questionnaire of up to 10 questions and answers that condense and simplify the news, or a glossary module that extracts and classifies the keywords of the text, accompanied by a brief definition.

Most news publishers, on average, produce content that would take the average reader about 15 hours to read, while most readers do not spend more than 20 minutes a day reading the news, Clarin points out. “The content is overwhelming, and readers miss a significant amount of relevant information.”

Does summarisation cannibalise normal readership? Norway’s Verdens Gang said last year it actually increases reader engagement, and now, the Norwegian public broadcaster found the same outcome. The Frankfurter Allgemeine Zeitung found visits per visitor were almost twice as high for those subscribers who read summaries, and 81% of its users deemed the summarisation feature useful.

Worthwhile links

• GenAI and money: Perplexity considers revenue-sharing deals with publishers.
• GenAI in news production: What are audiences most comfortable with? Behind-the-scenes applications; but within audience-facing products, text scores higher than video.
• GenAI in journalism: A free course to help reporters, editors, and others understand how AI works.
• GenAI for journalism: GPT-4 has a surprisingly good understanding of what it entails.
• GenAI for coding: Maybe it’s not as good at this as we thought it was.
• GenAI for project management: Asana introduces a “teammate” to triage incoming requests from co-workers, determine if all necessary information has been included, and ask for more details or move the request to the next step in the process.
• GenAI models: When will they run out of training data? Within the next seven years, perhaps.
• GenAI bot capability: Which one is good at what? A look at health, personal finance, creative writing and other topics. Spoiler: Perplexity is surprisingly capable.
• GenAI and elections: Three chatbots are put to the test.
• AI and economic crisis: Widespread use of AI could turn an ordinary downturn into a deep and prolonged economic crisis, says a top IMF official.

An AI diversion

To mark Father’s Day, a look at some awful “dad jokes” — some human-written and others machine-generated.

About this newsletter

Today’s newsletter is written by Sonali Verma, based in Toronto, and lead for the INMA Generative AI Initiative. Sonali will share research, case studies, and thought leadership on the topic of generative AI and how it relates to all areas of news media.

This newsletter is a public face of the Generative AI Initiative by INMA, outlined here. E-mail Sonali at sonali.verma@inma.org or connect with her on INMA’s Slack channel with thoughts, suggestions, and questions.

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