The future of audience intelligence starts with a question

For nearly two decades, publishers have relied on page views, engagement time, subscriber conversions and similar metrics to understand their audiences. Those signals have shaped editorial priorities, product roadmaps and business strategy because they show what readers consumed. What they rarely reveal is what readers were actually trying to find before they clicked.

 

Speaking at the WAN-IFRA World News Media Congress, Joey Marburger, VP of Content Intelligence at Arc XP, a Washington Post company, described what his team learned after spending two years developing Ask The Post, The Washington Post’s AI-powered answer experience. Rather than searching the web, readers can ask questions directly on The Post’s website and receive answers grounded entirely in the newspaper’s own reporting.

 

The biggest lesson wasn’t simply that AI could answer readers’ questions. It was that the questions themselves exposed something publishers have never been able to measure at scale.

 

“Your readers are literally telling you what they want to know about,” said Marburger.

 

That insight has implications far beyond AI. Every question represents more than a request for information. It reflects an expectation that a trusted news organisation can provide an accurate answer. In that sense, reader questions become signals of both intent and trust, offering publishers a clearer understanding of audience needs before traditional analytics ever can.

 

For Arc XP, those lessons are helping shape a broader approach to how publishers can use AI to deepen audience relationships. As the digital experience platform built by and for media organisations, Arc XP works with publishers to apply technology to the challenges newsrooms face every day, from delivering trusted information to understanding and serving audiences more effectively.

 

What analytics can’t tell you

 

Digital publishing has always depended on feedback loops. Editors publish stories, audiences respond, analytics identify what resonated and future coverage evolves accordingly. Publishers have become exceptionally good at understanding reader behaviour. They know which headlines attract attention, which stories convert subscribers and which formats encourage readers to return. What those metrics cannot explain is why a reader arrived in the first place.

 

A page view tells you someone clicked on a story about inflation. It doesn’t tell you whether they wanted practical advice for managing household expenses, historical context or an explanation of what caused prices to rise. Those distinctions matter because they reveal the journalism readers expected to find.

 

Marburger describes traditional analytics as “lag metrics”. They explain what happened after publication but provide little visibility into emerging audience needs. Reader questions add an entirely different layer of insight by allowing publishers to observe demand directly instead of inferring it through clicks and engagement.

 

Rather than replacing traditional analytics, those signals complement them. Traffic data remains essential for understanding performance. Reader questions help explain the motivations behind that performance, giving publishers a richer picture of what audiences are actually looking for.

 

When no answer is the right answer

 

One of the most revealing lessons from The Washington Post’s work wasn’t that Ask The Post could answer questions. It was that it sometimes refused to.

 

Many consumer AI tools attempt to answer nearly every question, even when confidence is low. Ask The Post follows a different philosophy. If The Washington Post hasn’t reported enough to answer confidently, the system simply declines to.

 

As Marburger explained: “We bias to non-answer.”

 

That decision reflects a longstanding editorial principle: accuracy matters more than completeness. Protecting trust sometimes means acknowledging what you don’t know rather than generating an answer that may be incomplete or misleading.

 

Those unanswered questions also proved surprisingly valuable. When readers repeatedly asked questions the newsroom couldn’t answer confidently, editors didn’t view those moments as failures of the technology. They viewed them as signals that readers expected The Washington Post to have authority on a topic where its reporting was incomplete.

 

Marburger described how those patterns are brought back into editorial discussions. “[Reporters] take that into the newsroom and start thinking about that coverage as they see the patterns of it.”

 

Instead of asking only which published stories generated the most engagement, editors can also ask which questions audiences expected them to answer but couldn’t. That creates an entirely new feedback loop for editorial planning.

 

Listening at scale

 

Throughout his presentation, Marburger returned to one simple idea: “Trust is downstream of listening.”

 

Publishers have spent years strengthening audience relationships through subscriptions, newsletters, personalisation and community initiatives. All of those efforts depend on understanding what readers actually need, which has become increasingly difficult. Comments sections require significant moderation. Conversations have fragmented across social platforms. Surveys provide only occasional snapshots of audience sentiment.

 

Questions offer something different. They represent active curiosity rather than passive consumption. Readers aren’t reacting to journalism that already exists. They’re revealing where they remain confused, what context they still need and what they expect a trusted newsroom to explain.

 

Marburger reflected on his early career in local journalism, when conversations with community members naturally shaped newsroom priorities. As publishing became increasingly digital, many of those direct interactions disappeared, replaced by dashboards tracking impressions, clicks and engagement. Conversational AI creates an opportunity to restore part of that dialogue.

 

Imagine an editorial meeting that reviews not only yesterday’s traffic report, but also the questions readers asked most often during the previous week. Editors might discover repeated requests for the historical context of an international conflict, practical explanations of new legislation or local reporting on an issue receiving national attention. They might also identify important topics where readers consistently expected answers that simply didn’t exist.

 

Those conversations don’t replace editorial judgment. They strengthen it by giving newsrooms another way to understand the communities they serve.

 

From experimentation to practical AI

 

The opportunity for publishers isn’t simply to add another AI interface to their websites. It’s to determine where AI can meaningfully improve the relationship between journalism and the people it serves.

 

That is the approach Arc XP is taking as it brings lessons from The Washington Post’s AI experimentation to other media organisations. Through its AI-powered experiences, including Ask the News, Arc XP enables publishers to create conversational experiences grounded in their own journalism, giving audiences another way to explore trusted content while helping publishers better understand the questions behind that engagement.

 

The goal isn’t to replace the newsroom’s editorial judgment or hand audience strategy over to an algorithm. It’s to give publishers new signals they can use alongside the analytics, expertise and audience knowledge they already have.

 

That distinction matters as publishers evaluate where to invest in AI. The most valuable applications may not be the ones that generate the most content or automate the most tasks. They may be the ones that help media organisations better understand their audiences and create stronger connections with them.

 

A better way to understand audience needs

 

Much of the industry’s discussion around AI has centred on distribution and referral traffic. Those concerns are understandable as more readers begin their information journeys inside conversational interfaces. The Washington Post’s experience suggests another opportunity deserves equal attention.

 

Reader questions represent a new source of audience intelligence. They complement traditional analytics by revealing unmet demand, emerging interests and opportunities for deeper reporting before those needs appear in traffic dashboards or subscriber reports.

 

For publishers, that insight extends beyond AI products themselves. It points toward a different way of understanding audiences, one rooted not only in what people read yesterday, but in what they hoped their news organisation could help them understand today.

 

The industry has become exceptionally good at measuring behaviour. The next competitive advantage may come from listening more closely to curiosity.

 

For organisations built on trust, that may prove to be AI’s most meaningful contribution. Beyond improving how journalism is delivered, it creates a new way for publishers to hear their audiences, strengthen editorial decision-making and reinforce the relationship that quality journalism depends on.

 

Arc XP, a Washington Post company, helps media organisations build and evolve digital experiences designed around their audiences. Drawing on technology and insights developed within one of the world’s leading news organisations, Arc XP gives publishers the tools to experiment with AI while keeping their journalism, brand, and audience relationships at the centre.

---

Learn more about Arc XP and its approach to AI for media organisations

---

by Julie Jochims, Media Industry Analyst, Arc XP by The Washington Post