Chequeado’s Asistente IA is a lightweight assistant designed to make AI practical for fact checkers and journalists in Argentina. It works as a Chrome extension and is easily integrated within the Google Docs workflow reporters already use. The tool packages a growing prompt library into an interface that is easy to adopt, adaptable to each newsroom, and oriented to collaboration. As part of the JournalismAI Innovation Challenge, the team prioritised accessibility, editorial usefulness, and regional relevance.

The problem: Bringing AI to where the user needs it

The project responded to two recurring pressures. First, the gap between the promise of AI and what newsrooms could adopt in practice. Second, the need to automate certain fact-checking tasks without sacrificing rigour. “The main focus was how to make AI accessible to journalists, especially fact checkers,” said Joaquín Saralegui, engineer and product manager.

The team also aimed to support peers across Spanish- and Portuguese-speaking Latin America, where resources are limited but needs are shared. 

Building the solution: Roadmap to prototyping

The team combined user research with technical iteration. Reporters were interviewed and regular check-ins identified concrete tasks worth solving. “We discovered that journalists really needed to chat with the solution,” noted data journalist Ignacio Ferreiro, shaping the interface.

They prototyped a plug-in that launched prompt-driven tasks directly in Google Docs, then iterated with journalists to refine output quality and editorial fit. A demo showed how users could generate a social thread from an article, draft a newsroom-ready summary, and scan a speech for checkable claims, all within a single session. Tasks can be tweaked, reordered, or shared so improvements propagate to colleagues.

Tools and technologies

The stack combined familiar web technologies with newsroom-ready integration:

• Backend and core services in Python and Django

• Google Chrome extension interface to bring functionality into existing workflows

• A configurable prompt engine, with context and per task knowledge bases

• LLM access via API keys controlled by each organisation

• An open repository with more than 15 prompts in three languages, covering fact checking, social publishing, writing aids, and productivity tasks

The local open version works out of the box with an API key. Organisational features that require shared backend services are available in Chequeado’s managed deployment.

The team and the skills involved

Asistente IA was built by a cross functional group: engineers and developers, a product manager and designer, and Chequeado’s newsroom. The wider organisation supported logistics and infrastructure. Crucially, developers and journalists worked in tight loops to judge model outputs by editorial standards. As Saralegui put it, without that “synergy between the two it’s really difficult to evaluate how good you’re doing,” which risks shipping a tool misaligned with newsroom quality.

Challenges they encountered

• Aligning tempos: The fundamental difference in working pace between development teams and journalists created initial friction. "The development team takes like six months to develop a product and the journalists are working by the minute," Saralegui observes.

• Regional collaboration: Whilst essential for impact, regional coordination proved to be consistently challenging. Coordinating across multiple organisations, from multiple countries in similar situations of limited resources requires extraordinary patience and flexibility.

• Finding the automation frontier: Perhaps most significantly, determining which journalistic tasks could genuinely be automated required extensive experimentation. This uncertainty led to unexpected discoveries about AI's capabilities in newsroom contexts.

• Moving targets in model capability: The rapidly evolving nature of AI technology added another layer of complexity and the need of constant reevaluation. "The thing you tried on month one and didn’t work might do on month eight," Saralegui explains.

The opportunities built from a collaborative approach

The project has revealed significant potential for reducing repetitive, non-creative tasks in newsrooms whilst preserving the essential human elements of journalism. "We discover a great potential for this tool and the kind of approach to reduce these repetitive non-creative tasks that happen in the newsroom," Saralegui confirms.

Regional collaboration has strengthened participating organisations' AI capabilities while creating knowledge-sharing networks. 

Regional collaboration raised AI capacities across participating organisations and fostered knowledge-sharing networks. The tool also proved relevant beyond fact-checking, since many solutions apply to broader journalistic contexts.

Lessons for newsrooms

• Integration trumps innovation: Rather than creating a standalone AI tool, Chequeado's Google Chrome extension succeeds, because it meets journalists where they already work in existing workflows, removing adoption barriers that plague many AI solutions.

• Collaboration between technical and editorial teams is essential: Without this synergy, AI tools risk producing technically impressive but journalistically irrelevant results.

• Embrace the experimental nature of AI development: The rapid evolution of AI capabilities means that failed experiments may become viable solutions months later. Maintaining a backlog of unsuccessful attempts allows teams to revisit solutions as technology advances.