Makedonia, a traditional regional newspaper in Greece with nearly a century of history, faced a crisis familiar to local media outlets worldwide: how to remain relevant in an era where social media delivers instant neighbourhood news and younger generations increasingly avoid traditional journalism. Their innovative response – an AI-powered personalised newsletter system – offers valuable lessons for regional media grappling with similar challenges.

The problem: When local news loses its local advantage

Regional and local media across Europe are grappling with an existential challenge. The challenge is particularly acute for outlets like Makedonia, which historically filled a clear gap by covering local neighbourhood stories that major media ignored. As Nikolaos Panagiotou, the project’s lead researcher, explains: "Two decades ago it was very obvious that the local or regional [news organisations] were actually covering a gap regarding what is happening in my neighbourhood. But what about now that everything can be easily reached through social media?"

The challenges are multifaceted. "AI is considered to challenge the viability of media, but most importantly it presents a major challenge in how to integrate AI into current newsroom day-to-day operations,” Panagiotou notes. "If this is a major issue for big media, you can imagine that for regional and local media, it is an even bigger problem, because they lack the resources, the knowledge, and the relevant staff."

Andreas Panagopoulos, the technical lead on the project, adds another dimension: "Journalistic staff at Makedonia is quite small, and most of them lack training not only in AI but also in digital tools."

Building the solution: Finding opportunity in a crowded market

The project emerged when Makedonia's management decided to relaunch their website and increase newsletter subscriptions as a pathway to print subscriptions. However, the team quickly realised this required more than traditional digital marketing approaches.

The breakthrough came from recognising that personalisation could address multiple challenges simultaneously. Research revealed a gap in the market: "Our research showed that very few outlets are using this system. Perhaps major international media, but certainly nothing in Greece, and nothing in the broader region – the Balkans, Turkey – for personalised newsletters."

The team identified AI's potential to automate routine tasks while enhancing rather than replacing journalistic work, leveraging AI to deliver "specialised content to targeted audiences".

At the heart of the project is the development of an advanced Personalised Newsletters and Push Notifications (PNN) system. By analysing user behaviour and preferences, the team expects this AI-powered tool to enable Makedonia to deliver content that is tailored to each reader’s interests. The goal is simple yet powerful: to offer a more meaningful, personalised news experience that builds loyalty and trust over time.

Building the technical foundation

The tool Makedonia built combines technical innovation with editorial oversight. The infrastructure integrates ChatGPT for content analysis, Mailchimp for distribution, and a custom WordPress-based content management system.

When users interact with newsletter content and indicate preferences, AI provides feedback to ensure the next newsletter is delivered "without human intervention”. However, the system includes a crucial editorial safeguard: "We're working on a special feature where every newsletter will include an 'editor's pick’ – two or three articles – to avoid the filter bubble effect and give users what journalists feel they need to know, not just what they want to know,” Panagopoulos explains.

The technical implementation required integrating university AI experts with web developers to create optimal user experiences for both readers and journalists. The team planned to test multiple LLMs, starting with ChatGPT, but expanding to include Gemini and possibly others.

Bridging the newsroom divide

The project's greatest challenge wasn't technical – it was cultural. "What was more challenging was integrating the AI experts with the journalists – getting AI experts to understand journalists' needs and, on the other side, getting journalists to understand what the AI experts were doing,” Panagopoulos notes.

The team organised four comprehensive seminars to address this gap. "Journalists have fears: 'We're going to lose our jobs and AI is going to replace us.' So we had to start from the beginning, explaining the goal of this specific project and how AI can be their assistant, not something that will take their jobs."

Success required creating common vocabulary through "a small glossary so everyone could understand what we mean when we talk about LLMs, what an LLM is, how LLMs work”, and using appropriate benchmarks. The team learned to avoid inappropriate comparisons: "If you present journalists at Makedonia with the New York Times example, they'll say 'this is the New York Times' and immediately reject it."

The opportunities: Scaling beyond Thessaloniki

The Makedonia project represents more than a single outlet's digital transformation – it offers a potential template for regional media across Southern and Eastern Europe. "We think that this tool will be an opportunity for expanding to other media outlets, probably to have a startup on that,” Panagopoulos explains.

The team sees expansion possibilities once "the algorithm is trained from everyday use”. Future enhancements include testing headlines to optimise newsletter open rates and extending personalisation to push notifications. The broader implications extend beyond technology: "Newsletter subscribers and users are the more loyal audience for every website all over the world."

Lessons for newsrooms

Makedonia's approach suggests three critical insights for regional media: 

• Prioritise easy adoption: Develop solutions that are designed to be "easily introduced and adopted" by newsroom staff, minimising disruption and training overhead.

• Offer implementation support: Provide comprehensive support for implementation and change management, not just the technical tools themselves, to ensure successful integration.

• Reject "one size fits all" models: Avoid generic solutions and instead reject approaches that ignore vital local contexts and unique regional needs.

• Balance AI with editorial control: Create a sustainable model by leveraging AI for efficiency while firmly maintaining human editorial control. This allows outlets to expand reach without compromising their local mission.

• Use AI for personalised local value: View AI-powered personalisation as an opportunity to create appealing content that helps preserve the local focus, particularly the distinctive value of regional journalism, as traditional subscription models face limitations.