How a San Diego Nonprofit Newsroom Uses AI to Build Brand Affinity

Giovanni Moujaes has a clear line he won't cross. As assistant editor for audience and innovations at inewsource, a nonprofit newsroom in San Diego, he's spent the last two years experimenting with AI tools. But there's one thing he refuses to do - let AI write the stories.

"We don't believe at this time that AI should be writing articles that our reporters could be writing," Moujaes points out. "When I can have AI write an investigation based on some data, we're going to have our reporters do that."

Instead, inewsource uses AI strategically to solve a different problem: making their journalism more accessible to readers who have personal limitations or just don’t have enough time and might otherwise bounce from their site within 30 seconds. 

On a recent Zoom call, Moujaes and I talked about inewsource’s AI strategy and also about what the News Product Alliance AI Collaboration Lab, is developing for the benefit of the whole news industry.

One example is Contextualizer, a tool that highlights specialized terms in stories. Click on "San Diego Association of Governments" and you get a pop-up explaining what it is, related articles that mention it, and a donation prompt tied directly to Contextualizer.

"Sometimes people give based on a particular utility they really appreciate," Moujaes explained. Rather than generic donation campaigns, inewsource tests whether readers will support specific features they find valuable.

The tool's earlier incarnation came from a company that went under. So Moujaes used ChatGPT to help rebuild it. inewsource now hosts the tool themselves, a small example of how AI coding capabilities let small newsrooms maintain products they couldn't afford to develop traditionally.

Sponsored content

Staying behind on tech trends can be a career killer.

But let’s face it, no one has hours to spare every week trying to stay updated.

That’s why over 100,000 engineers at companies like Google, Meta, and Apple read The Code twice a week.

Here’s why it works:

Join 100,000+ engineers who read The Code to stay ahead of the curve.

inewsource also uses AI-generated story summaries at the top of longer pieces, working with a company called Abridged Media. The summaries include bullet points, key facts, notable quotes, and links to more stories from the author.

Critically, every summary carries the following disclosure: "Answers are generated by AI and edited by humans."

"Nothing front-facing that you see that has had AI involved in that process is unfettered AI," Moujaes said. "It's all been reviewed and edited by humans.” The core principle: AI generates first drafts, but humans clean up language, fix contextual inaccuracies, and ensure tone matches the newsroom's voice.

As a result, stories with additional AI-generated summaries and/or explanatory features show increased time on page, even if some users only read the summary. That might seem paradox but it actually makes sense because these are users that would not read the full length article. So the choice is either offering them a more snackable version or losing them.

Moujaes acknowledges that this conclusion isn’t the result of a rigorous A/B test, but inewsource uses heat mapping data from Microsoft Clarity that shows readers clicking on these features are staying engaged longer.

For audio, inewsource uses Everlit, which employs ElevenLabs’ voices to create text-to-speech versions of stories. Every audio file starts with a disclosure: "This is an AI-generated narration of the story, which may include mispronunciations of certain words."

Moujaes frames these tools as accessibility features rather than engagement gimmicks. Text-to-speech serves readers with dyslexia, ADHD, or those commuting. Story summaries help people who lack time for deep reads but still want to understand local news.

"If we're catering to a broader audience and we want to bring in everyone from the person who's working 14 hours in a day to a single mother or father, our job is to make sure people know what's going on," he said.

Survey data from a Trusting News cohort showed readers were largely neutral about AI use. Moujaes interprets this as: AI didn't detract from the news experience but didn't vastly enhance it either. The real value shows up in engagement metrics, not sentiment surveys.

Beyond inewsource, Moujaes co-founded the News Product Alliance AI Collaboration Lab, funded by the McGovern Foundation. The project tackles a problem that many newsrooms are facing - making sense of scattered first-party data.

"We're trying to think about what can AI do to help us better understand first-party data," Moujaes explained. Most newsrooms have data spread across Mailchimp, survey platforms, Salesforce, and other tools. The lab is building a schema that lets AI analyze all these sources together.

The goal? Help newsrooms understand their audiences at a granular level. Which readers engage most with education coverage? Who responds to surveys? Can you create a targeted newsletter for education superfans and convert them to donors?

The schema will be open source and available to all newsrooms, not just News Product Alliance members. inewsource plans to prototype the tools themselves once development reaches that stage.

"Small organizations don't always get to make or have the benefit of taking that great research and great product thinking and bring it into their organizations," Moujaes said. "So we're trying to democratize that, especially around first-party data."

On product development with limited budgets: inewsource joins cohorts like the American Journalism Project's Product and AI Studio (partnered with OpenAI) to access enterprise-level tools and connect with other smart people thinking about journalism's future. The applications take time, but Moujaes sees the cost-benefit as worthwhile for both resources and community.

On measuring success: When surveys prove inconclusive, inewsource relies on heat mapping tools like Microsoft Clarity and analytics dashboards from their vendors. They track clicks, time on page, and engagement patterns rather than waiting for perfect A/B tests.

On AI principles: Moujaes is adamant about two things. First, AI should augment journalism, not replace it. Second, humans must review everything AI touches before it reaches readers. This "humans in the loop" approach maintains quality while gaining efficiency.

On the future: While many newsrooms discuss building databases that sell structured content to AI systems, inewsource takes a different path. They're focused on creating unique storytelling experiences that can't be replicated by AI scrapers.

"People follow news outlets because of the great people behind the work," Moujaes said. "To say that we should just be looking at us as a database and not as a room full of journalists first, I think is the wrong way to approach it."

It's a philosophy that puts reporters and readers first, using AI as a tool to strengthen those relationships rather than replace them.