Media organizations are getting to grips with AI. Across the industry, teams are experimenting while leadership works to put strategies and guardrails in place. Much of this activity has focused inward, using AI to improve efficiency across areas like data analysis, marketing, and sales.

Customer-facing applications have been slower to emerge. That caution is intentional. When AI moves closer to audiences, the stakes change. Trust is central to publisher value, and audiences remain wary of AI’s role in content and news-making.

Still, some publishers are testing where AI can serve audiences without undermining credibility. These efforts are not about replacing journalism, but about improving access, context, and usability. From conversational tools to article summaries and content repackaging, these experiments point to how AI can deepen engagement when it is clearly constrained and grounded in trusted source material.

AI-powered Q&A as a subscriber benefit

One publisher seeing success with customer-facing AI products is Harvard Business Review. In late 2024, they launched Ask AI, a subscriber-only conversational tool which brings source-linked answers to leadership and management questions directly from HBR’s own archives. It also draws on video and podcast transcripts from the publisher.

When the HBR team first looked to develop this feature, it was in response to some very specific customer challenges. Subscribers sometimes come to the site ready to engage with full articles and in-depth analysis. But other times they are faced with time constraints and don’t have time to read deeply to find out what they need.

“We’re seeing a couple of different use cases for it,” explained Erika Heilman, VP and Deputy Publisher. “Some use it as a coach or mentor, with back-and-forth interaction. Others see it as a more trusted source than just going to a foundational LLM, because it’s got a pure corpus.”

“It has very specific guardrails against it… and we made sure to pressure test it against hallucinations.” Dave Lefort, Managing Director of Digital Product Strategy added.

HBR has been pleased with the take-up so far.  A quarter of HBR subscribers have used Ask AI, and of those, a third have interacted multiple times. 

Now, the team is looking to increase usage and improve onboarding. This includes multimodal search toolbars offering quick AI-generated summaries of articles and the Ask AI tool, as well as showing the option to ‘Ask AI’ within articles. “We’re looking for ways to integrate it more into the subscriber experience more holistically,” Lefort explained. “So while you’re reading, you can query the article about anything, ask it to help me apply this to my situation.”

Heilman noted that Ask AI shows potential as a powerful retention tool. “The key for us going forward is to make sure that we are not trying to compete against ‘good enough’ free answers,” she said, explaining why it’s a subscriber-only benefit. “We are trying to create something of value for senior leaders who are, frankly, making consequential decisions every day, and are unlikely to do that from a chat-based foundational model. A trusted source is really key.”

AI article summaries as a reading companion

Like HBR, the Financial Times is also experimenting with AI conversational search. Ask FT is being used by their B2B brand “FT Professional” to search and summarize FT journalism going back to 2004. But this is not the only place the publication is using AI to engage readers.

The FT has been experimenting with using AI-powered summaries in its articles. Similar to HBR, they acknowledged that sometimes readers come wanting the facts of a story quickly, but other times they settle down for a deeper reading experience.

Bullet point summaries of longer articles aren’t a new idea. Many publishers have tried variants over the years, and AI has enabled faster production of article summaries without involvement in the actual content creation.

On FT content, rather than overviews appearing at the top of an article, readers are offered the option in a separate toolbar to summarize or translate the piece.

At the Definitive AI Forum in London, FT CEO Jon Slade said that these were proving successful, and actually increasing engagement. AI summaries “have led to more people reading more depth of the article,” he explained. “They’ve used it as a means to understand, is this an article I want to spend more time with, rather than replacing their reading of the article.”

AI embedded to improve the product experience

FOX is also integrating AI into their customer-facing products, albeit less explicitly. John Fiedler, FOX’s EVP of Product & Engineering outlined that they look at where AI and generative AI experiences can be built into different levels of products.

A recent example of this is FOX One, a direct-to-consumer streaming service which offers all of FOX’s news, sports and entertainment content. It launched in August 2025, with a price point of $19.99 monthly or $199.99 annually.

FOX One is powered by AI at a number of different levels. Fiedler explained that one of the more obvious touchpoints consumers will be aware of is their generative AI-style customer support tool. “We started using GenAI for customer care under the premise that traditional customer care, whether it’s a phone or email address or old-school chatbot, is antiquated,” he said. “We can likely replicate an agent conversation with GenAI, and have it be a better and more successful user experience.”

Users of FOX One can do more than just access customer support. They can ask the AI what shows are coming up, upgrade or cancel subscriptions, or even handle issues like outages.

Fielder noted that this feeds into many other functions AI can then help with. “AI is really good at understanding what a video is about, what a live stream is about…certain person’s history and behaviors, and how that should affect things like recommendations within the product,” he outlined. “That’s giving us a much deeper understanding of content through AI than we’ve been able to do in the past.”

Within FOX One, there is also a space for short-form video, much like YouTube Shorts or TikTok. The team uses AI here in multiple ways to help production. Fielder explained that the technology can ‘watch’ the live stream of their channels, and identify interesting moments, or “spicy takes”. 

“Say in a sports studio, you have a heated debate about what happened in the game last night, that’s a hot moment,” Fielder outlined. “So you cut out the beginning and the end, and now you have a clip. But it’s still formatted from a 16×9 resolution.

“So we pass that through another layer of [AI] technology that converts it into a vertical video where it overlays the right people, and the right layouts, and makes sure that it’s formatted nicely for vertical consumption. Then it will put words on the screen, and pass it through to the product.”

This kind of streamlining makes the content teams much more efficient at creating and reformatting video. Fielder was keen to emphasize that it speeds up the workflows, but there are still humans approving every step of the way.

What early audience-facing AI experiments suggest

Taken together, these examples show publishers experimenting carefully with where AI fits into the audience experience. The common thread is not novelty, but restraint. In each case, AI is used to help people find, understand, or repurpose existing journalism and video, rather than to generate new content or make editorial decisions. Human oversight remains explicit, and the underlying value proposition stays rooted in trusted source material.

That approach reflects a broader reality for publishers navigating AI in public view. Audiences may be curious about new tools, but their tolerance is closely tied to clarity and purpose. Where AI is introduced to meet a specific need, such as saving time or reducing friction, it can support engagement. Where it feels unnecessary or opaque, it risks eroding trust. Right now, publishers appear to be setting the terms, introducing audience-facing AI on their own conditions rather than racing to deploy it everywhere.