How Generative AI Is Powering Facebook’s Reactionary Content Economy

A Le Monde Pixels investigation published on September 9, 2026, reports that hundreds of Facebook pages are using generative AI to produce reactionary content for French audiences. The pages are linked to activity in countries including Sri Lanka and Algeria, according to the report’s summary, and appear to benefit from the attention generated by highly provocative posts.

The central finding is not simply that political material is being produced with AI. It is that generative tools can make the creation of targeted, emotionally charged content faster and cheaper, supporting a commercial ecosystem built around reach and engagement.

The available reporting identifies the activity as lucrative, but does not establish a single operator, organisation or centrally controlled campaign. That distinction matters: a collection of pages can share production methods or business incentives without being part of one formally coordinated network.

What the investigation reports

Le Monde describes hundreds of pages carrying reactionary or far-right material and targeting French users. The pages reportedly publish content adapted for online audiences rather than relying on a single political message or a single national context.

The use of generative AI is significant because it lowers the production cost of images and text. Operators can create large volumes of material without commissioning conventional graphic design or writing each post manually. The result is a steady supply of memes, visual posts and short-form political messages designed to attract reactions and shares.

The report’s public summary does not provide enough information to independently verify the total number of pages, the identities of the page administrators or the exact revenue generated. Those details should not be treated as established beyond the reporting unless they are documented in the full investigation or supporting evidence.

Fact and source trail

Why the AI angle matters

Political propaganda and reactionary campaigning existed before generative AI. The technology changes the economics of production. A small group can now create more images and text, test different messages and maintain multiple pages with fewer people and fewer resources.

That does not mean every AI-generated political post is deceptive, illegal or part of an influence operation. It does mean that volume becomes harder to use as a signal. A page publishing dozens of polished posts each day may be difficult to distinguish from a legitimate media, activist or meme account without examining its provenance, administrators and publishing patterns.

For platform operators, the case also raises a moderation problem. Detecting synthetic content is not the same as determining whether that content violates a policy. An AI-generated image may be clearly artificial but still fall outside rules covering hate speech, impersonation or political misinformation. Conversely, a realistic-looking post may be harmful because of its message and targeting, even when the image itself is not technically deceptive.

The commercial incentive behind engagement

The investigation frames the pages as a business rather than only an ideological project. Reactionary material can attract strong responses, and strong responses can increase distribution on social platforms. The available summary does not specify the exact payment mechanism, so claims about advertising revenue, in-stream ads or other monetisation tools require confirmation from the full article.

What can be said from the reported findings is narrower: operators appear to be using AI-assisted production to pursue audiences at scale, and the pages’ commercial value is connected to their ability to generate attention.

What remains unclear

Several important questions are unresolved in the available material. It is not clear how many people operate the pages, whether the pages share administrators or technical infrastructure, or whether they are linked by a common organisation. The report summary also does not establish how much of the content is fully generated by AI, how much is edited by people, or which specific tools are involved.

The status of the pages is another open question. Readers should not assume that a page has been removed, demonetised or formally identified by Meta unless the investigation provides evidence of that action. Nor should the presence of AI-generated imagery alone be treated as proof of a coordinated disinformation campaign.

For developers, trust-and-safety teams and researchers, the practical takeaway is to examine more than image authenticity. Page creation dates, administrator locations, posting frequency, repeated templates, audience targeting and monetisation status may provide stronger evidence of coordinated commercial behaviour than an AI detector used in isolation.

Source: Le Monde Pixels, “Sur Facebook, le lucratif business des contenus réactionnaires générés par IA,” https://www.lemonde.fr/pixels/article/2026/09/09/sur-facebook-le-lucratif-business-des-contenus-reactionnaires-generes-par-ia_6768826_4408996.html