This year, a duel for popularity took place at the Festival of Humanity. between the French Communist Party (PCF) and France Insoumise (LFI). During this large gathering of the French left held this weekend, the two parties both sought to mobilize around their candidate for the presidential election, Fabien Roussel (PCF) and Jean-Luc Mélenchon (LFI), who each held a campaign meeting.
Online, both camps have also strongly communicated about their respective rallies, presented as successes in terms of mobilization – the PCF for example declared that 40,000 people were present during Fabien Roussel’s speech.
But the day after the speech given by the communist candidate, pro-LFI accounts on X claimed that a photo of the PCF meeting had been modified by IA.
“Long live AI eh”
The photo in question, published by a communist internet usershows Fabien Roussel from behind facing an audience of supporters gathered in front of him. But on the left in the photo, we can see an orange arrow located just above a spectator, which may seem, at first glance, to be digitally added.
Immediately, many pro-LFI accounts mocked a supposed montage which would have consisted of artificially adding people to the photo. “Ptdrrrr (fart laughing, Editor’s note) the arrow that appears above the guy on the left, long live AI eh,” says one of these accounts, whose publication broadcast on September 13 has been viewed more than a million times on X.
Several other accounts also denounced manipulation, supposed to hide the supposed lack of people at Fabien Roussel’s meeting.
Examples of publications on X which claim that the photo published by a communist Internet user was modified by AI. © Montage Les Observateurs
An arrow-shaped headband
In reality, this orange arrow does not demonstrate manipulation: it is indeed found on several other photos published by the PCF. But above all we can observe this orange arrow on numerous occasions, and always above the same person, during the meeting broadcast live, accessible on the PCF Youtube page (like at 1m40 of the video for example).
At 1’40 of the live broadcast on the PCF website, we can see this same orange arrow in the audience (framed in yellow). We can also see this arrow on other occasions, such as at 1’59 in the video. © Montage les Observateurs
During the video, a closer shot of the crowd allows us to understand that the person in question is wearing a headband with an arrow motif.
A shot filmed in the crowd shows that the orange arrow is a headband worn by one of the people in the audience (at 3’27 in the video, then at 3’33), on the same model as headbands available on the internet (photo top left). © Montage les Observateurs
Why such a headband? This model, available on many e-commerce sites, is often worn by festival-goers wanting to be easily recognizable from afar during such gatherings.
Screenshot of the crowd in front of the stage on which Fabien Roussel held his meeting on September 12, 2026 at the Fête de l’Humanité. © PCF / Youtube.
No trace of modification by AI
Our editorial team was, however, unable to precisely verify the announced figure of 40,000 participants in the meeting. The images of the meeting filmed by the PCF show a compact crowd arranged on the esplanade facing the stage. Some videos and photos posted online (like this one taken about twenty minutes after the start of Fabien Roussel’s speech) still show a more dispersed crowd a few dozen meters further on.
However, no trace of general modification of the photo by AI was detected by the Observers editorial staff. Our editorial team relied on the new detectors implemented by several major AI platforms (such as Google, ChatGPT, TikTok Or Meta): thanks to digital marking technology, these companies are implementing tools capable of detecting whether the photo has been modified or generated by their platform (our explanations to read here).
Also readTagging content generated by AI: towards more transparency?
Result: none of these detectors detected a trace. Note that this method only concerns platforms that have implemented this detection method – where a large part of the content generated by AI to date comes from these different platforms.
In May 2025, Jean-Luc Mélenchon himself was wrongly accused of having manipulated by AI the photo of a demonstration against Islamophobia. The editorial staff of the Observers had already demonstrated at the time that the French flags visible in the photo were real, contrary to accusations made by several online accounts.
Also readWhy Jean-Luc Mélenchon is wrongly accused of having published an AI-generated image
