Alyriaz is an online media outlet. We are therefore directly concerned by what we describe here, and it would be dishonest to pretend we are observing from the outside. We try to approach the subject with the same rigour as others, flagging where our own interests lie.
The most widespread fear is that machines will write articles instead of journalists. It is the least relevant one. The real upheaval is economic, and it is further along than people think.
What AI does well, and what is useful
Let's start with the uses that work, because they exist.
Processing large volumes of documents. This is the clearest benefit. Analysing thousands of pages of administrative documents, cross-referencing databases, spotting anomalies in financial filings: this work was impossible for a small newsroom, and it is now doable. Several recent investigations would not have existed without it.
Transcription and translation. A considerable time-saver, with no editorial stakes.
Producing derivative formats. Adapting an article into a summary, a short version, a social media thread. The editorial work still happens upstream.
These uses have one thing in common: they concern processing, not judgement. That is exactly the division we described in our article on jobs.
The real problem: money
This is where the matter is actually decided. Online journalism has been funded for twenty-five years by a tacit pact: you publish, search engines send you visitors, and those visitors pay for your work.
That pact is now breaking, as we analysed in our article on the future of the web. When an AI answers directly, the reader no longer visits the source. The information is used, the producer is not paid.
The arrival of AI Overviews in France accelerates this shift, with an impossible choice for publishers: feed the answers with no guarantee of return, or disappear from surfaces used by billions of people.
The consequence is not that journalists will be replaced by machines. It is that there will be fewer journalists, because there will be less money to pay them. The difference matters: in the first case, the work is done differently; in the second, it is not done at all.
Part of the online press has responded by producing more content at lower cost, with automated assistance. This degrades quality, fuels distrust, and pollutes the corpus that future models will feed on, which brings us back to model collapse. It is a perfect vicious circle: less revenue, therefore more cheap content, therefore less trust, therefore less revenue. The first damage is not caused by AI to the media; it is caused by media outlets responding badly to AI.
What seems to hold up
Three things hold, and they point to where value is shifting.
What does not exist elsewhere. An original investigation, a testimony gathered in the field, a document obtained. No AI can summarise what has not yet been written. This is the only truly non-substitutable material.
The direct relationship. Newsletters, subscriptions, communities: anything that connects a reader to an author without going through an algorithmic intermediary. It is no coincidence that this format is booming.
Verification as a product. In an environment saturated with plausible content, knowing that a piece of information has been verified by someone whose job it is takes on a value it did not have when information was scarce.
What this asks of readers
There is a concrete consequence, and we own it as a stakeholder. A business model based on visitor traffic will not hold. What can hold is a model where those who find useful information contribute to its continued existence, through a subscription, a share, or simply reading the source rather than the summary.
This is not a moralising appeal; it is mechanics. In the web taking shape, attention becomes a vote on what will survive.
What to take away
The question is not "will AI write the articles". It is technically capable of doing so, and that produces mediocre content nobody is asking for.
The question is who will fund the work that machines cannot do: going to see, asking, verifying, insisting. That work is expensive, cannot be summarised, and it is precisely what a society needs. Its disappearance would make no noise, and that is what makes it likely.