Exploring AI in Journalism – Babel’s experience
The digital media outlet Babel is experimenting with AI avatars to tackle newsroom challenges in Ukraine. Discover what they have learned so far
Babel’s ‘That’s the History’ project video thumbnail.
At Babel, we’re strong AI enthusiasts — from news editors to designers, our team actively uses a range of tools. So when we saw the call for applications for the JournalismAI Innovation Challenge, supported by the Google News Initiative, we immediately decided to apply. It felt like the perfect opportunity to expand our space for experimentation.
The challenge we’re addressing is very real: Ukraine is currently facing a serious staffing crisis due to migration and mobilization. This problem can affect any newsroom — unexpected emergencies can suddenly remove key people from the editorial process, putting entire projects at risk. That’s why finding ways to keep journalists and authors engaged — even remotely — is critically important. Our initial idea was to create an AI avatar of our historian Serhiy Pyvovarov, whose expertise is central to our YouTube series ‘That's the History’.
In the first few months, we built a solid foundation — assembling the team, researching AI tools, testing options like ElevenLabs and Synthesia, and planning out the content. But not everything went as planned. One of the main lessons we’ve learned is that not all AI tools are equally suited for journalistic content. Synthesia, the platform we initially chose to generate AI avatars, turned out to restrict historical and political themes — which are at the core of our work. Since we couldn’t test this without a paid subscription, we ended up shifting our strategy. We are now focusing on generating a realistic voice model of Serhiy using ElevenLabs, and the results have been largely satisfying.
Babel NGO staff members.
Another insight from our experimentation: realism in avatars remains a major challenge. Despite their technical sophistication, AI presenters often appear stiff and unnatural — something our editorial team felt would make it difficult for audiences to stay engaged over time.
While the full impact of this project will only become clear by the end of the implementation period, the process has already reshaped how we think about AI in the newsroom. We realised we were a bit too optimistic when we first began content production. Video creation with AI avatars is still too expensive for small newsrooms, and the results don’t always meet expectations — the on-screen substitute can look artificial and fail to inspire trust or empathy.
As we moved forward with implementation, we noticed that even major players like Netflix have encountered similar issues. In the Pop Culture miniseries ‘Dirty Pop’, they “resurrected” a deceased producer using similar technology — and the result was equally uncanny. You can see it for yourself at the 2:35 mark in this video.
While we’re still in the middle of our journey, this project has already offered invaluable insights into both the potential and the limits of AI in journalism. Thoughtful experimentation remains the key to discovering what truly works. In the months ahead, we’ll be releasing new episodes, continuing to test and adapt, and moving toward a clearer understanding of how AI can sustainably support journalism.
—
This article is part of a series providing updates from 35 grantees on the JournalismAI Innovation Challenge, supported by the Google News Initiative. Click here to read other articles from our grantees.