I posed the question as to the number of nonviolent demonistrators to an AI program (LUMO) and the response was:
No — and I want to be precise about that, because the question keeps getting asked and the absence of an answer is itself the finding.
The specific number you're asking for does not exist in any credible source. Nobody has measured the share of French demonstrators who are non-violent, because:
- The denominator is contested. For 17 October 2026 the Interior Ministry counted 550,000 nationally; the CGT claimed 800,000 (franceinfo, AFP). That's a ~45% spread before you even get to behaviour.
- Nobody classifies crowds by behaviour. Demonstrations are counted as totals, not sorted into peaceful/violent. "Non-violent participants" is only ever inferred as a remainder — and the subtracted quantity is unmeasured.
- Arrests are not a usable proxy. French practice includes preventive gardes à vue, so arrests overcount; most people who clash with police are never detained, so they undercount. Both errors run simultaneously.
What does exist, and what it actually shows
The closest thing to systematic data is franceinfo's
Le vrai ou faux comparison of French and European protest policing (14 Oct 2026). It reports that between 2020 and September 2026, French police intervened in
598 demonstrations that were proceeding calmly, and that officers intervene in
12% of demonstrations in France even when they are calm — with France ranking fourth in relative terms. That statistic is about
policing, not protester behaviour, but it's a useful corrective: a demonstration generating police intervention is not the same as a demonstration containing violence.
Ouest-France published a comparative tally for 17 October against other mobilisations since 2017 — 550,000 participants, 1,800+ arrests, 600+ police injured — again without any breakdown of participants by behaviour.
Journalistically, the characterisation is consistent across outlets: Al Jazeera noted "widespread peaceful protests, with unions, teachers and regular citizens taking to the streets" alongside the clashes; AFP described violence as occurring "on the sidelines"; CBC said "the majority of protests have been peaceful."
Academically, the strongest formulation comes from Olivier Fillieule (Lausanne) and Fabien Jobard (CESDIP) in
Politiques du désordre (new edition Feb 2026). Their framing is that violence "until then had been the attribute of ultra-minority groups" — a qualitative judgement from twenty-plus years of fieldwork, deliberately not quantified.
One trap worth naming
IFOP has polled French opinion
about violence in protests (including during the pension-reform movement). That measures what the public thinks, not who is in the street. Those two things get conflated constantly in this debate, and the conflation is usually deliberate.
Bottom line
The defensible claim is directional only: violent participants are a small minority, plausibly low single digits at most. Any specific percentage of "non-violent demonstrators" circulating in French media or politics is derived from a contested denominator, recycled from a minister's on-the-day assertion (Castaner's "1,500 ultra-violents" in Paris in March 2019 is the classic case), or invented. The metric doesn't exist because no institution has an incentive to produce it — the government benefits from emphasising disorder, organisers from emphasising peace, and neither funds a participant-level survey.
If you want, I can dig into whether anyone has attempted a proper participant survey (
enquête par questionnaire in a cortège) — that's the only method that could generate the figure, and I'd want to check whether French social-movement researchers have run one.
www.zerohedge.com
www.foxnews.com
