Let’s talk about a bold use of consumer-facing generative AI: A TV anchor that is allowed to interview India’s prime minister and its biggest movie star. 

India Today — which owns TV channels, digital channels, and a radio station — built a GenAI-powered anchor in three months and has even given it a spot on its prime time 9 p.m. news broadcast. The AI anchor, called Sana, speaks multiple languages in a consistent tone and style, has no learning curve (“Ask me anything under the sun — I am always ready,” it says), and can easily switch between topics and formats.

Within a few months, the media company had also rolled out AI avatars for its human anchors, ensuring undisrupted coverage even when the actual anchors are out in the field reporting, such as during the recent election.

“This is something that has been validated by sustained advertiser interest as well,” said Vibhor Gandotra, head of strategy and special projects at India Today.

The company first built the AI anchor to handle repetitive content, “mundane in nature, for which it became difficult for us to get a human element to deliver that particular piece of content,” Gandotra said, citing as an example India Today’s Superfast News, which pulls together 100 different news headlines into a 10- to 15-minute video package that can be viewed on YouTube. 

The project managed to overcome some early challenges, including immature technology that was expensive, as well as questions around legal and ethical implications.

“We had a very, very strong validation from our leadership team — a very ambitious vision and an aggressive timeline was set,” Gandotra told the INMA Generative AI master class in May.

They decided to keep the team small, to five or six people, so they could move quickly. They created small proofs of concept and socialised them with various internal newsroom teams, “and that is how the adoption really gained scale,” he said. As Sana was launched, “the various functions were really excited about the possibilities of what it could enable for their businesses.”

Building trust with India Today’s audience was a multi-step process, Gandotra said. The news brand featured Sana on its prime time slot to build familiarity and also created a social media avatar. Sana is also a radio host that reads the latest sports scores, traffic updates, and provides personalised content recommendations across Web sites, all to make them more visually appealing than simply text. 

Sana, which won two INMA Global Media awards in April, now provides content in several languages across the sub-continent. The India Today group reaches 750 million people a month.

Certain parts of programming are now AI-automated, such as weather and stock updates and horoscopes, while “other components that have a longer shelf life and are more research-driven are still human-collaborative,” Gandotra said. Human anchors focus on mainstream news and on-the-ground reporting, fostering content differentiation.

He exhorted his peers in the global media industry to dream big and be bold: “Our biggest key learning has been to adopt first, think later.”

Interested in hearing about other fascinating AI innovations coming out of India? I hope you’ll join me at the INMA South Asia News Media festival.

Interested in learning more about AI anchors? Please join us on the INMA Los Angeles Tech Innovation study tour, where we will be visiting Channel1.

The flight to quality in the generative AI-content era 

I worked as a business news journalist through a couple of major financial crises. One term that often pops up at times like that is “flight to quality” — that at times of uncertainty, investors prefer to put their money in assets that are of high quality, with lower risk profiles, so they can sleep better at night.

That’s a theme I am hearing echoed in the media world as we talk about generative AI, the Internet, and the uncertain times we face. There are two main concerns these days:

• AI-powered search will subsume traffic to news publishers’ sites by providing the summarised answers to users’ questions instead of sending them to the source Web sites (and this could lead to US$2 billion in ad revenue loss annually across the U.S. publishing industry).

• Growing user mistrust of news on the Internet because there is so much AI-generated misinformation out there.

When I have spoken to news industry executives about these risks in recent days, half a dozen of them have confidently spoken about flight to quality: They believe that in times of uncertainty, audiences will turn away from commodity content on the Internet generated by AI.

“By definition, content that is produced by generative AI is average content,” one executive told me. When it comes to writing, LLMs are basically auto-complete functions trained on the entire Internet. The next word generated in each instance is the most likely word that will follow — in other words, the average result.

He’s not the first one to point out that this does not sound like fine writing. LLMs write for a median audience using median words and do not write well in a way that sounds human: “Terribly uniform and uniformly terrible,” as one academic said, adding that the writing is clear but “the reader has the impression of reading an instruction manual.”

And this was before Google’s AI Overviews started making terrible mistakes. (I do think that they will improve, though.) Google’s CEO has gone on the record as saying that AI Overviews actually have increased clicks and referral traffic to news companies. A Google representative told the INMA Generative AI master class that certain queries, such as those about breaking news, will be excluded from AI Overviews. 

It’s not just about the writing when it comes to quality — it is also about the user experience, Tim O’Rourke, vice president of content strategy at Hearst Newspapers, told me. 

“We are creating products that make us the place to come for the best local, unique, high-quality journalism in the region,” he said. “The more unique our journalism is, the better off we’re going to be. The less we try to make content to game SEO, the better.” 

You can learn more about Hearst’s efforts in the GenAI internet era here. O’Rourke was presenting along with Edward Hyatt, director of SEO at The Wall Street Journal, who also talked about how we can future-proof the news business at a time when the Internet is overrun with AI-generated content and many are concerned about how our audiences will find our content.

Worthwhile links

• GenAI and the Financial Times: How the AskFT chat product was built and what’s up next.
• GenAI and The Washington Post: Looks like the new CEO is counting on AI for a turnaround.
• GenAI and subscriptions: Surprise: A large percentage of people say they are as willing to pay for mostly AI-generated content as they are for human-generated content.
• GenAI licensing deals: News Corp signs a licensing deal with Open AI worth US$250 million over five years.
• GenAI licensing deals II: The Atlantic and Vox Media sign licensing agreements with OpenAI, which now has deals with news publishers representing about 70 news brands.
• GenAI licensing deals III: They raise a raft of familiar questions in the fraught world of platforms and publishers.
• GenAI inspiration from retail: Save US$10 million annually, be “brutally efficient.” 
• GenAI and coding: Mistral released its first AI model focused on generating computer code.
• GenAI and sarcasm: Researchers have created a programme able to discern sarcasm. So appropriate for the Internet.

A non-AI diversion

As someone from India who lives in Canada, I have always been aware of the influence of the British across centuries and continents. The fascinating Empire podcast explores not only the British empire but also the Turkish and Russian empires and slavery. I have very little interest in history, but this podcast has me hooked. 

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.

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