These predictions about AI journalism are grim reading

Fundamentally, journalism is a game of relationships.

The articles are the output.

It's about building and maintaining a network of contacts who are willing and able to disclose information that's not on the public record.

Breaking a story isn't about rewriting a press release and getting it up on the website fastest.

It's about building a friendship with that senior public servant over years. So that when they're pissed off about the current government's backflip to lobbyists, you're the person they gripe and leak documents to.

It's also about being good at conducting interviews. That means having a deep knowledge of your beat, so you know which questions to ask.

And it's about verifying what you hear, rather than being a stenographer.

The article is the outcome of that process, and those relationships. Ideally, it adds important new information into the public record.

Meanwhile, large language models work is they predict the most statistically likely word in a sentence, based on their training data.

An LLM article is an output, based on existing training data.

Based on this, let's look at the Reuters Institute's top five predictions for how AI will reshape the news: https://reutersinstitute.politics.ox.ac.uk/news/how-will-ai-reshape-news-2026-forecasts-17-experts-around-world

Warning: It's grim.

1. Audiences will increasingly access news through AI

"As audiences increasingly use generative AI-powered chatbots, search or other AI-informed tools when searching for information, traditional ways of accessing the news will give way to news stories being discovered and accessed via AI tools. This was our most prevalent forecast overall: a plurality of both our expert forecasts and our audience contributions fell under this broad category."

So how do news organisations make enough money to pay journalists to build a network of contacts and break news?

2. There will be increased demand for verification​

"Credibility will differentiate news outlets in a world where information is easy to access, but trust is low. Audiences will want to see evidence and sources to back up what they are told online, and news publishers have an opportunity to meet this need. This idea was more popular among our audience contributions than among our expert forecasts."

This is an important point. In an age of AI slop, conspiracies, and disinformation, properly sourced original news becomes more important than ever.

3. Automation and agents will reshape newsrooms

"This theme encompasses deeper and more comprehensive integration of AI by newsrooms: AI will become embedded in newsrooms’ CMS and workflows. Publishing of some information will be automated and carried out by agents, and stories will be automatically updated. The much-discussed ‘human in the loop’ might be quietly retired."

This directly undercuts point two. If the value of news organisations is in providing verified information, why would you put unverified AI slop and hallucinations on your news website?

But there's a deeper issue.

Large language model agents might be able to emulate outputs by churning existing information into articles.

But how do they replicate building a network of trusted contacts?

They can't.

4. Newsrooms will upskill and build new AI infrastructures

"2026 could be the year in which newsrooms invest in the infrastructure and training necessary to make the most of what AI already has to offer. This is particularly relevant for small newsrooms, which may not have dedicated roles or investment yet."

In other words, the work subeditors have traditionally done (writing summaries and headlines) will be outsourced to LLMs to save news organisations cash.

The issue is, in many news organisations, subbies also double as proofreaders and fact-checkers.

Getting rid of them might save cash in the short term. Until the defamation lawsuits hit.

5. AI will further empower data journalists​

"Data journalists have long utilised AI, and as the technology improves and becomes more accessible, it could enable newsrooms to gather and sort through an unprecedented number of documents. This was mentioned twice in our expert forecasts, but was only a minor theme in our crowdsourced responses."

We already have a technology that can gather and search through documents.

It's called a search engine.

Final thoughts

Unfortunately, many of the people who will ultimately make the decisions about the points above have never worked as journalists.

They don't recognise the difference between the outcome of relationship building and the output of an algorithm trained in existing content.

They will trash their mastheads by confusing the latter with the former.

And the public sphere will be further degraded as a result.

#auspol #news #media #ausmedia #journalism #LLM #LLMs #ChatGPT #Grok #Gemini