AI in Journalism: Ally, Amplifier, and the Digital Divide

Prof. M Zahidul Haque

Prof. M Zahidul Haque

Generative AI is rapidly getting popular in journalism. It is now become a standard assistant in newsrooms—speeding up routine work, expanding language access, and aiding data-heavy investigations—while simultaneously testing journalism’s core duties of verification, transparency, and public trust.

Its societal impact, however, is double-edged: without deliberate policy and investment, AI can widen the digital divide; with the right safeguards and support, it can help narrow it What generative AI can (and cannot) do for journalism Generative AI refers to systems that create new content—text, images, audio, video, or code—by learning patterns from large datasets. Crucially, AI lacks independent judgment or a built-in sense of truth. It can “hallucinate” facts, reproduce bias from training data, and blur lines between genuine and fabricated media. That is why leading outlets treat AI output as unverified source material requiring full editorial verification.

How newsrooms are managing accuracy and ethics Major organizations have moved from experimentation to codified policy. The Associated Press (AP), Reuters, BBC, and others now require:

--Independent verification of every AI-generated claim against primary sources (documents, records, interviews Human editorial control: AI may assist with research, transcription, translation, or headline suggestions—but not replace reporting or verification

--Clear disclosure when AI plays a material role in published content and strict rules against AI-created or altered news imagery. It requires visual verification workflows (reverse-image search, metadata checks) and correction procedures when errors occur.

In practice, this means a story that begins with an AI summary still ends with a journalist confirming names, dates, quotes, and context—and an editor signing off.

AI’s uneven footprint: the Global South and the digital divide

AI’s benefits are real but unevenly distributed. Research in 2025–2026 shows that journalists in the Global South frequently use AI for newsgathering and production, yet face structural barriers that can reinforce inequality.

How AI can widen the divide

--Infrastructure and cost: Advanced tools (real-time translation, large-scale data analysis) often require stable internet, capable devices, and paid subscriptions—advantages held by well-funded urban outlets.

--Language and context gaps: Models trained predominantly on English and major languages can mistranslate nuances or miss local context, disadvantaging communities speaking less-represented languages

--Concentration of power: A small set of companies shape model “logic,” influencing what counts as newsworthy and how verification is performed—potentially marginalizing local editorial norms.

How AI can narrow the divide

--Lowering skill and cost barriers: Automation of routine tasks (data briefs, transcription) lets small teams cover more ground and compete on certain stories

--Expanding language access: When it works well, AI translation can bring a single investigative story to readers in Bengali, and other languages without large translator teams

--Fighting misinformation: AI-assisted fact-checking can help communities where rumors spread fastest and media literacy resources are scarcest

--Open-source momentum: Cheaper, capable open models are beginning to reduce dependency on expensive enterprise tools.

Conclusion

Artificial intelligence is no longer a peripheral tool for newsrooms — it is becoming embedded in how stories are researched, drafted, and even visualized.

Yet its value depends entirely on how responsibly it is integrated. A practical newsroom policy built on four pillars — verification first, human accountability, transparency, and equity by design — offers a workable path forward: AI outputs should be treated as leads requiring verification against primary sources, not as facts in themselves; editorial judgment and final responsibility must remain firmly with journalists and editors. As provenance and detection tools such as AP Verify and C2PA signing become part of standard workflows, their effectiveness will depend on two parallel investments:

training newsroom staff to critically evaluate AI-generated content, and educating audiences to recognize disclosures and understand the limits of verification. Neither technology alone nor policy alone can safeguard credibility

— it is this combination of institutional discipline and public literacy that will determine whether newsrooms can harness AI's efficiency without sacrificing trust.

(Prof. M Zahidul Haque is an Adjunct Faculty of Agricultural Journalism Program, Sher-e-Bangla Agricultural University, Dhaka.)