How do you grow subscription conversions by more than 280%? 

For Taiwan’s United Daily News Group (UDN), the answer lies in data-driven curation and an AI “co-pilot” for journalists.

At the INMA Asia/Pacific News Media Summit, Amelie Lin, director of data development, detailed UDN's playbook for embedding AI into newsroom decisions without sidelining human judgment. 

She began by establishing UDN’s legacy and scale: “United Daily News Group is one of Taiwan's legacy media organizations, founded in 1952.” 

Its flagship platform, UDN.com, now attracts 60 million monthly unique visitors, over 380 million visits, and more than 8 million registered members, supported by an archive of 27 million articles and photos.

Despite this strong digital presence, Lin noted the organisation faced operational slowdowns in 2017. In response, they launched the data development department, which has since grown into a team of 20 analysts and engineers managing over 400 terabytes of data with a clear mission: to transform UDN into a data-first, AI-powered media business.

Lin framed this transformation around three core strategies:

1. Reimagine the value of story: Ensuring everyone speaks the same language, backed by clear, impactful metrics.
2. Reconnect with audience: Using continuous feedback loops to align coverage with reader needs.
3. Rebuild with AI: Cultivating a “co-pilot culture” where AI optimises workflows and supports transparency.

This strategic framework laid the foundation for Curate X, UDN’s AI-driven content curation system. Lin described it as more than a tool. It was a fundamental workflow transformation. 

She explained, “It's a new way of working. It brings together smart insights, design thinking, data-driven curation, and constant testing. And most importantly, it connects data with editorial decisions, supporting both newsroom workflows and business goals.”

At the heart of Curate X is the Consensus Map, a structured guide for story planning that aligns editorial judgment with subscription goals. 

“The consensus map is our story writing guide,” Lin said. “It takes editorial judgment and turns it into a full-tier structure: topics, categories, subcategories, and angles. And we decided to build subscription conversion directly into story strategy. So, editorial choices naturally aligned with business goals.” 

This system is powered by 36 data models — 24 built specifically for the newsroom — that classify topics, suggest angles, and provide real-time audience insights.

The impact on UDN’s subscription strategy has been profound. 

“Subscription conversion grew by more than 280%,” Lin said. “Over 250 journalists have already used Curate X in their work. And through this system, our newsroom has produced more than 7,000 articles. Most importantly, at a group level, membership penetration grew by more than 180%.”

Data has also been built directly into UDN’s newsroom workflows. Journalists use the Curate X Dashboard to track metrics in real time, a significant shift from the past. 

“In the past, data was always backward-looking,” Lin said. “We only saw the results weeks later. Now, we can check subscription conversion and audience engagement in real time. The newsroom can adjust strategy right away.”

The system's effectiveness is clear across UDN’s products.

For its general news service, stories surfaced by Curate X converted readers at three times the rate of those without it.

On Economic Daily News, its financial news platform, articles using Curate X achieved 3.2 to 4 times higher membership penetration over three years.

These results demonstrate Curate X’s scalability across different content areas and newsroom cultures.

To further support journalists, UDN developed aiAssist, a newsroom co-pilot. It handles routine tasks, freeing journalists for deep reporting while ensuring human oversight. 

As Lin explained, “This is our AI co-pilot for the newsroom. But every editorial decision still goes through humans.” 

She reiterated the guiding principle: “AI assists, but never replaces.”

These results demonstrate Curate X’s scalability across different content areas and newsroom cultures.

To further support journalists, UDN developed aiAssist, a newsroom co-pilot. It handles routine tasks, freeing journalists for deep reporting while ensuring human oversight. 

As Lin explained, “This is our AI co-pilot for the newsroom. But every editorial decision still goes through humans.” 

She reiterated the guiding principle: “AI assists, but never replaces.”

UDN’s transformation fundamentally reshaped decision-making. Before 2022, story choices relied on senior editors’ judgment. The introduction of Curate X brought data models and dashboards into newsroom meetings for everyone. 

Lin explained, “Data models and dashboards were used in newsroom meetings — not only by journalists and editors but also by content leads and the editor-in-chief.”

Lin identified six critical success factors:

1. Leadership endorsement: Strong top-level support to drive adoption.
2. Cross-functional collaboration: Editorial, data, and product teams working as one.
3. Common language: The consensus map creating a shared newsroom vocabulary.
4. Trust from journalists: AI supported, rather than replaced, editorial judgment.
5. Agile experimentation: Encouraging trial and error without fear.
6. Transparent results: Clear metrics on story performance to build confidence.

Lin emphasised that technology alone doesn’t change newsrooms; culture does. 

“The real turning point was not the dashboard or the tools. It was joining newsroom meetings, sharing the same goals as journalists, and slowly building their trust.”

Building on this success, UDN is now piloting its next-generation system: Publish X.

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