In a speech to the 2026 Belt and Road Journalists Forum in Nanjing, journalists were urged to redesign verification for an era of deepfakes, synthetic evidence and increasingly confident AI errors.
Imagine waking up in Croatia tomorrow morning and seeing a video of the Croatian prime minister, one of the fiercest supporters of Ukraine in the European Council, announcing that military support for Ukraine has been withdrawn. The video looks authentic. The voice sounds real. Thousands of people have already shared it, and some news websites are reporting on the public reaction.
But there is one problem: the announcement never happened.
This is exactly why fact-checking has become more than a journalistic routine. In today’s Europe, where hybrid warfare combines cyberattacks, political manipulation and information operations, a convincing lie can become a security threat before the truth has even put its shoes on.
And this is not some distant, futuristic scenario. In October 2026, European security officials again warned about intensifying Russian and Iranian hybrid campaigns, while drone incidents and suspected sabotage continued to raise concerns across Europe. Information is now part of the battlefield, and journalists are among its first responders.
So, the question is not whether artificial intelligence will change journalism. It already has. The real question is: how do we use AI without allowing it to destroy the very thing journalism depends on – trust?
Traditional fact-checking was built around identifiable claims, traceable sources, photographs and recordings that could be examined. Today, AI can generate a convincing image of an event that never happened, clone a politician’s voice, or produce hundreds of slightly different versions of the same false story.
The challenge is no longer simply identifying a fake. It is dealing with an industrial-scale production of deception. And even when a deepfake is exposed, the damage may already be done. People may remember the allegation but forget the correction.
There is another danger: the so-called liar’s dividend. Once people know that deepfakes exist, politicians and public figures can dismiss genuine recordings as AI-generated. In other words, AI makes it easier both to manufacture false evidence and to cast doubt on real evidence.
For Croatia and Southeast Europe, this problem is particularly important. Our information space is connected across borders, languages and political communities. A misleading narrative can start in one country, be translated into Croatian or Serbian, and reach thousands of people before local journalists even notice it.
The regional SEE Check network has documented the growing use of AI-generated images, videos and voices in disinformation, including the misuse of public figures’ identities in fraudulent schemes. Its work reminds us that the threat is not only about geopolitics: it also affects elections, public health, climate debates and people’s personal finances.
And the technology itself is not the only problem. A 2025 European Broadcasting Union study, involving 22 public-service media organisations in 18 countries and 14 languages, found that AI assistants misrepresented news content in 45 per cent of the responses examined. The lesson is clear: a confident answer is not necessarily a correct answer.
What about journalists themselves? AI is already part of their working lives. A Thomson Reuters Foundation survey of 221 journalists across more than 76 countries and territories found that 81.7 per cent used AI at work. Yet 57.6 per cent of AI users were self-taught, and only 13 per cent of respondents said their newsroom had an AI policy.
These figures do not represent every journalist in the European Union, and we should not pretend they do. But they reveal a worrying pattern: adoption is moving faster than training, editorial guidance and accountability.
So, what should we do? I would suggest three practical solutions.
First, redesign verification for the AI era.
Journalists should verify the source, context and chain of custody of digital evidence – not just whether an image looks real. Reverse-image searches, geolocation, metadata analysis, frame-by-frame video examination and comparisons with original footage should become routine. AI-detection tools can help, but no detector should be treated as a final verdict.
The golden rule is simple: verify through independent evidence. Contact the original source. Find the earliest upload. Compare the footage with reliable records. Ask whether other credible sources independently confirm the event.
We should also practise prebunking: explaining common manipulation techniques before a crisis occurs. It is much easier to recognise a scam when someone has already shown you how it works.
Second, make AI a verification assistant, not an editorial authority.
AI can transcribe recordings, translate foreign-language material, identify inconsistencies, organise large document collections and help journalists search for earlier versions of a viral claim. It can also assist with monitoring suspicious narratives across several languages.
But journalists must independently verify every important factual claim against original documents, named sources or other reliable evidence. AI should never be asked to certify its own answer.
Here, journalists need a basic AI training package. Every newsroom should teach staff how to write precise prompts, recognise hallucinations and bias, check citations, protect confidential sources, understand deepfakes and use reverse-image and video-search tools. Training should include practical exercises with synthetic audio, manipulated images and real breaking-news scenarios.
The objective is not to turn every reporter into a computer scientist. It is to make every reporter a more informed sceptic.
Third, change newsroom management – not just newsroom technology.
Media organisations need clear rules stating which AI uses are acceptable, which require additional approval, and which are prohibited. They should establish human sign-off for high-risk material, maintain records of AI-assisted work, protect unpublished information and disclose significant AI-generated content to audiences.
Editors should also budget for verification time, specialist training and cooperation with independent fact-checkers. Speed cannot be the only measure of performance. A newsroom that publishes first but gets the facts wrong has not won the race; it has simply spread the problem.
And support already exists. The European Digital Media Observatory, or EDMO, connects fact-checkers, researchers and media-literacy experts across Europe. In April 2025, the European Commission announced nearly €5 million in funding calls to strengthen resilience against disinformation and increase the reach of fact-checked content.
Support must reach journalists’ employers, too. European funding and professional networks should help smaller newsrooms develop AI policies, share verification tools, organise joint training and build partnerships with universities and technology experts. Otherwise, advanced verification will become a privilege of wealthy media organisations, while smaller outlets remain exposed.
Ultimately, fact-checking in the AI era is not a competition between humans and machines. It is a question of whether we can combine technological speed with human judgement, editorial independence and public accountability.
AI can generate a thousand convincing falsehoods in minutes. Journalism cannot investigate all of them in the old way. But it can become more collaborative, more systematic and more transparent.
So, the next time a video makes you angry, frightened or absolutely certain that you know what happened, pause before sharing it. Ask three questions: where did it come from? What independent evidence supports it? And who benefits if I believe it?
Because in the AI era, the most important journalistic skill may not be knowing how to generate an answer. It may be knowing when an answer has not yet earned our trust.
Dragan Nikolić
Dragan Nikolić is a senior international affairs correspondent, media professional, and co-founder of the Foreign Policy Forum. With a career in journalism and international reporting spanning several decades, he has reported on global politics, conflict, European affairs, sustainability, and major political developments across five continents.