How a War-Zone Newsroom in Ukraine Uses AI To Listen

Hi, I’m media innovation journalist Ulrike Langer and you’re reading my weekly AI in media newsletter. This issue is inspired by the European Publishing Congress in Vienna on June 18 where Elsa Court of the Kyiv Independent and I both spoke. She gave one of the most interesting talks of the day. I followed up afterwards with questions about the newsroom's use of AI. This piece draws on her talk and her written answers.

Paying subscribers get the slide decks from my three EPC talks — descriptions and link below.

Most newsrooms that talk about AI and audience mean more or less the same thing: a recommendation engine, a paywall model, a chatbot bolted to the archive. Elsa Court, audience development manager at The Kyiv Independent, does something more specific. She takes the questions her paying members ask about the war in Ukraine — hundreds of them, far more than any article could hold — strips out the personal data, and feeds the whole pile into Google’s NotebookLM to find out what her most committed readers still don’t understand.

The Kyiv Independent is an English-language newsroom in Kyiv, founded in November 2021 by journalists who left another Ukrainian paper after the owner interfered with editorial policy. Three months later Russia's full-scale invasion began. It isn't a household name, but for the audience that follows the war closely it has become one of the most prominent English-language sources on it: a staff of about 90 reporters and editors, more than 2 million monthly readers, over 30,000 members paying upwards of $5 a month, and roughly 70 percent of last year's revenue coming from those members.

Just under 97 percent of that audience is based outside Ukraine (US 35%, UK 14%, Canada 10%, Australia 6%, Germany 2%, Sweden 2%), and in a reader survey two-thirds said they had never been to the country or had no personal connection to it. That is the audience Court is mining for questions: foreign, committed, and following a war almost none of them is experiencing first-hand.

The questions come from a recurring series, “Our readers’ questions about the war, answered,” now in its eleventh volume. Every three to four months the newsroom asks members what they want to know, and journalists answer individually. There are always far more questions than one article can hold, and Court keeps the leftovers  — not for any single answer, but for the pattern across them: which topics return, which fade, which are new. She sorts the answers into dated documents, loads them into NotebookLM, and queries the model for how the themes shift over time.

One topic kept resurfacing among US readers: why Ukraine does not draft its 18-year-olds. Inside Ukraine the answer is settled and rarely asked — mobilization rules are lived fact — but Americans found the exemption baffling and kept returning to it. The newsroom answered the gap directly, with explainers and added context in its coverage. The signal came from the members; the model surfaced it.

It is a research tool used for management. NotebookLM gets discussed in journalism circles as something reporters use for parsing large external documents. Court points it at her own audience — their knowledge gaps, their curiosity, where their attention is moving.

Listening to the members really matters because their attention is draining. Five years of Google Trends show worldwide interest in Ukraine spiking in February 2022 and never returning to anything near it. Court’s tactical answers to that decline — new formats, collaborations with other outlets, tying Ukraine to the wider geopolitical story — are the ordinary tasks of audience development. The two decisions that matter here are the ones where AI is the actual question: where she lets it in, and where she won’t.

What readers want most from the Kyiv Independent is its front-line reporting, and demand for it rises as Western coverage of the front thins out. The reason that coverage thins is not editorial boredom. First-person-view drones have turned the front into, in the newsroom’s own headline, a no-go zone for journalists. Most editors elsewhere will not authorize the risk. The Kyiv Independent still sends reporters, and their dispatches and video are among the most consistently popular things it publishes.

The Kyiv Independent has sent reporters to the front since 2022, before AI was the question. Only recently, Court says, did the newsroom start asking which of its formats a machine couldn't replace — and front-line reporting was the obvious answer. Audiences engage most with personality-driven reports, where the journalist is present in the story rather than narrating from a desk — and that pull grows, not shrinks, in an age of synthetic media. 

Court's sharper point is not that AI can't produce front-line video. It can, all too well, but the provenance is the problem. In an age of deepfakes, "having real people going to the frontline and actually filming what is happening on the ground is more important than ever,” Court says. “It's about recording this war as it was.”

AI assists the newsroom elsewhere: dubbing spoken Ukrainian into English on video, text-to-audio on articles, transcription — Court sees no remaining case for transcribing an interview by hand. What stays human is the core: reporting from the ground, countering disinformation, connecting Ukrainian voices to a global audience. The split is not sentimental. An AI summary can make complex information digestible, but it cannot reproduce what it is to lose a home as the front creeps closer, or to come under a drone attack — and as coverage tilts toward peace talks and diplomacy, someone still has to show what is at stake for the people who will live with the outcome.

There is an uncomfortable edge here. The human reporting Court defends is exclusive and valuable partly because it is becoming lethal. The same drone war that makes on-the-ground footage irreplaceable is what empties the front of journalists in the first place. An advantage that exists because going to the front is close to suicidal is not one any newsroom should want to hold for long, and Court does not pretend otherwise — she frames the mission as bigger than the outlet: recording the war as it is. Read plainly, the Kyiv Independent’s position is less a strategy to copy than a description of what only a newsroom inside the war can still do, and what it costs to do it.

At a reader-funded outlet the line is arguably less blurred than elsewhere, because the test is built into the model: does a given use of AI bring direct value to the reader, or improve the experience for the audience? Trust is the asset the whole operation runs on, and the wrong use of AI spends it faster than any efficiency buys back. 

That principle runs through both choices: AI on the audience, never on the reporting. The newsroom points AI at its members to understand why they read and pay in the first place, and keeps it away from the reporting that earned that trust. Zetland’s Jakob Moll has made a version of this argument from peacetime Copenhagen — AI in the backend, trust kept human. The Kyiv Independent is the same bet under fire, where the human side of it is a person standing somewhere a machine cannot reach.

Bonus content for paying subscribers: the slide decks of my three keynotes at the European Publishing Congress in Vienna:

Publishers who rely solely on content licensing agreements with OpenAI, Google, and others are acting defensively. The real opportunity lies in building their own infrastructure for AI distribution and setting their own rules: with structured data, API-enabled content, and archive monetization. Some media companies are already doing this – they are transforming their content into a platform, selling data points to B2B customers, enabling paid scraping, and creating new products from archive material.

US newsrooms are building AI tools that don't just work within their own companies – some are being made available as open source for the entire industry, while others generate licensing revenue. European publishers are facing a strategic decision: develop their own tools, adopt what's coming out of the US, or risk falling behind. Includes a map of the landscape and a decision-making framework for build versus buy.

This talk draws on my continuously updated News Machines case study database and walks the audience through the most important developments of recent weeks and months in a condensed form.

Log in or upgrade for the the download links.