AI Postal Service — Multilingual Chatbot for POSTAL

The service does not produce new reporting. Instead, it helps readers find and understand journalism already published by POSTAL, providing a more accessible route into local news for people who may struggle with Portuguese-language searches or conventional digital news formats.

Project Activities

POSTAL do Algarve implemented the project with its technology partner Dengun through three main phases.

The first phase established the technical and editorial architecture. The partners selected a retrieval-augmented generation model, commonly known as RAG, so that the chatbot’s answers would be based on POSTAL’s own reporting rather than on generic information gathered from the open web. The newspaper’s archive was indexed, and multilingual prompts were tested in Portuguese, English, Ukrainian, Hindi, Bengali and Cape Verdean Creole.

During the second phase, a staging environment was used to compare different implementation options in terms of answer quality, operating costs and token efficiency. Prompt design and multilingual handling were refined through successive tests, with particular attention to accuracy, clarity, editorial control and the long-term financial sustainability of the service.

The final phase combined controlled testing with the public launch. Feedback was collected from users, including non-Portuguese speakers and people with different levels of familiarity with digital tools. This led to clearer user guidance, interface improvements and adjustments to the chatbot’s tone and prompts.

The AI Postal Service was launched publicly in May 2026 and remains available on the POSTAL website. It retrieves relevant articles through POSTAL’s content infrastructure and operates under newsroom editorial oversight.

Since the launch, POSTAL has monitored conversation volumes, message activity, language distribution and recurring types of questions. This information is being used to improve the service, identify gaps in readers’ information needs and support editorial planning. The technical infrastructure and editorial oversight model developed through the grant continue to be used beyond the funded period.

What is the impact?

In the Algarve, a news desert can result not only from the absence of local media, but also from the inability of some communities to access the journalism that already exists. Many migrant and non-native residents live and work in the region but face language or format barriers when trying to understand local decisions, public services, employment issues, healthcare information or events affecting their daily lives.

The AI Postal Service helps narrow this access gap by allowing readers to ask direct questions and receive contextualised answers based on POSTAL’s reporting. It extends the reach of public-interest journalism without replacing the newsroom’s editorial judgement or relying on unsourced information from the open web.

The project should be regarded as an important initial intervention rather than a complete solution. English accounted for 267 of the 374 conversations recorded by 19 August 2026, indicating demand for a non-Portuguese route into local news. However, the available analytics do not establish users’ demographic backgrounds or demonstrate longer-term civic effects, and public use has so far been concentrated in English and Portuguese despite wider multilingual testing. Continued outreach and longer-term monitoring will be needed to assess its reach across the Algarve’s different language communities.

The project gave readers a simpler and more direct way to engage with local news. Instead of searching through individual article pages or depending on social media feeds, users can ask a question in natural language and explore a subject at their own pace.

This makes POSTAL’s reporting easier to discover for non-Portuguese speakers, newer residents who may be unfamiliar with local institutions and readers who find traditional digital news formats difficult to navigate. The conversational format also allows users to follow up on an answer and obtain additional context without having to formulate several separate searches.

The project also created a new source of editorial feedback for POSTAL. Readers’ questions indicate which subjects generate uncertainty, which information is difficult to find and where further reporting may be needed. These signals complement conventional page-view analytics and can inform both editorial decisions and product development.

POSTAL also gained a technical foundation that can continue to be developed, including an indexed knowledge base, content integration, multilingual prompt structures and an editorial oversight process.

Project Links

• AI Postal Service — live chatbot
• POSTAL do Algarve Facebook post

Project Pictures

Photos by Freepik