In short: Online homophobia can be measured. For several years now, SOS homophobie's annual report has ranked the internet as the leading context for LGBTphobia reported to the association, far ahead of the neighbourhood or the workplace; the French Interior Ministry, through its statistical service for internal security (SSMSI), publishes its own annual tally of anti-LGBT+ offences recorded by the police and the gendarmerie; and since the EU Digital Services Act (DSA) came into force, large platforms have had to publish the volume and handling of hate speech reports. Placed side by side, these figures tell three stories: anti-LGBT+ hate has moved online, it remains massively under-reported, and automated moderation sees a great deal of text but very little live video. This article explains what each source measures, what it misses, and what that means for men who chat on video with other men.
This article describes public data and regulatory frameworks. It does not constitute personalised legal advice. If you face threats, blackmail or outing, contact a specialised organisation and file a complaint.
Three counters, three different realities
We often read that "homophobia is rising" or "homophobia is falling" without any clarity about who is counting what. Yet the three main French and European sources are not measuring the same phenomenon.
| Source | What it counts | What it cannot see |
|---|---|---|
| SOS homophobie (annual report on LGBTphobia) | Testimonies received through its helpline, chat service and forms, sorted by context (internet, family, neighbourhood, work, public places, etc.) | People who never contact an association; the real volume, since these are reports rather than a population survey |
| Interior Ministry / SSMSI | Anti-LGBT+ crimes and offences recorded by the police and the gendarmerie (insults, threats, violence, discrimination) | Everything that never led to a complaint or an official report — in other words, the vast majority of incidents |
| Platforms' DSA transparency reports | Content reports received, removal decisions, account suspensions, moderation resources declared per language | The actual quality of decisions; unreported private exchanges; almost all live video |
On top of these three counters come the European surveys: the European Union Agency for Fundamental Rights (FRA) periodically questions tens of thousands of LGBTI people about their experience of harassment and discrimination, and ILGA-Europe publishes its Rainbow Map and Annual Review every year, documenting how the legal framework is advancing or regressing country by country. These surveys do not count complaints: they count lived experience. That is why their orders of magnitude are always far higher than police statistics.
The under-reporting gap
This is the most stable figure of all, and the most important to remember: the overwhelming majority of LGBTphobic incidents are never reported. For more than ten years, FRA surveys have measured a non-reporting rate well above four-fifths of incidents experienced. The reasons cited are always the same: the belief that nothing will be done, fear of a homophobic reaction from the person taking the report, and — specific to our subject — the fear of having to explain why you were on a dating site or a gay chat, and therefore of outing yourself to a police officer, an employer or a family.
That fear is not abstract. It explains why Interior Ministry figures, although rising almost continuously since the statistic has existed, must be read as an indicator of willingness to file a complaint as much as of how the phenomenon itself is evolving. A rise may mean more hate, or more trust in institutions. Both readings coexist.
Why the internet became the leading reported context
In successive SOS homophobie reports, the internet ranks first among the contexts of LGBTphobia reported to the association. Three mechanisms explain this concentration, and all three show up on random video chat.
Asymmetric cost. Insulting a stranger in the street exposes you to a physical reaction, to witnesses, to being stopped. Insulting a stranger from behind a screen costs nothing: one click and you are gone. Random video chat pushes that asymmetry to the extreme, since an exchange sometimes lasts three seconds.
Actively hunting for targets. Part of online hate is not a reaction, it is an activity. Some users log on precisely to "run into" LGBT+ people and film, insult or humiliate them. It is the same mechanism found, on a far more serious scale, in the ambushes documented by the press and by associations: the first contact is online, the trap is physical or financial. We covered it in detail in our guide on homophobic ambushes and blackmail arising from a video chat.
Content persistence. A spoken insult evaporates; a screenshot circulates. That is the structural difference between offline and online homophobia, and the reason outing-based blackmail has exploded along with video. A recording lasting a few seconds is enough to build a threat that can be reused for years.
What moderation catches — and what it cannot catch
The transparency reports published under the DSA give, for the first time, orders of magnitude for moderation resources. Read carefully, they reveal an imbalance that directly concerns video chat.
Text is well covered, live video far less so
Automated systems effectively detect:
- keywords and variants of homophobic slurs in written messages, including the classic spelling workarounds;
- already-known images and videos, via digital fingerprints (perceptual hashing) compared against databases of banned content;
- behaviour: mass account creation, repetition of the same sentence across hundreds of exchanges, reports concentrated on a single user.
They detect poorly, or belatedly:
- speech in a minority language or dialect, for lack of trained models;
- irony, innuendo, implicit threats ("I know where you live"), which contain no banned word;
- live streams. Analysing a video feed in real time is expensive, and most of the work is done by sampling frames: moderation sees snapshots, not the scene.
In other words, on a video chat, real protection depends less on the algorithm than on two things: how quickly the user reports and the quality of the human handling behind it. That is what we analysed in our feature on AI moderation of video chat in 2026.
What the DSA has concretely changed
Since 2024, the Digital Services Act has imposed several obligations on platforms that matter for LGBT+ people:
- An accessible reporting mechanism and an acknowledgement of receipt for reports.
- A statement of reasons for decisions: when content is removed or an account suspended, the platform must explain why and indicate the avenues of appeal.
- Periodic, public transparency reports, including the number of reports, the median handling time and the share of automated decisions.
- For very large platforms, an annual systemic risk assessment, which must explicitly cover harms to fundamental rights and therefore discrimination.
In France, there is also the PHAROS platform for reporting illegal content, the 3018 service for cyberharassment, and the option of filing a complaint online. The Interior Ministry also restated in 2025, around 17 May, the rollout of LGBT+ liaison officers in police and gendarmerie services — a useful contact precisely when the fear of outing yourself at the front desk blocks the process.
Reading the European figures without getting them wrong
The regional landscape has hardened, and the data shows it. In May 2026, UN experts publicly warned of a global rollback of LGBTIQ+ rights. In September 2026, the arrest of dozens of LGBT+ activists in Turkey in the name of "protecting the family" was documented by several associations, including STOP homophobie. In Senegal, a vote to toughen legislation overwhelmed international helplines, according to a franceinfo report in March 2026.
For a French-speaking reader, these facts have a very concrete consequence: the nationality of the person you are talking to on a video chat says nothing about their intentions, but the country you connect from changes your own level of risk. A screenshot that would, in France, be a breach of privacy becomes, in a country that criminalises homosexuality, the beginning of evidence. We have covered these two settings separately: the situation in Turkey in 2026 and criminalisation and online entrapment in the Maghreb.
Two pointers for reading ILGA-Europe's Rainbow Map without misinterpreting it:
- it scores the legal framework, not lived experience. A well-ranked country can still see a sharp rise in assaults;
- it scores the existence of protections, not their enforcement. A hate speech law that is never applied still earns a good score.
From figures to practice: what the data suggests
Statistics protect no one. But they show where to put your effort, and the order of priorities they trace is clear.
1. Reporting is the only lever that shows up in the figures
As long as incidents go unreported, they do not exist for platforms or for public authorities — and moderation resources are calibrated on declared volumes. Reporting homophobic behaviour takes ten seconds and produces two effects: possible sanctions against the user, and training data for the systems. Our detailed method is here: report, block, preserve evidence.
2. Evidence is prepared before you need it
The cases that succeed are the ones where the victim kept something: timestamp, username, exact wording, screenshot. That requires a minimum of practical organisation. Many people documenting harassment end up backing up their screenshots outside their phone, on an encrypted USB drive kept separately — a separate medium prevents someone looking through the device, whether a relative or an authority, from exposing both the evidence and the rest of your private life. A hardback notebook in which you write down the date, time and context as events happen is also worth more than a reconstruction from memory three months later.
3. Reducing your exposed surface costs less than repairing the damage
Outing-blackmail cases share a common denominator: something identifying has circulated. First name, city, workplace, a framed diploma on the wall, a company badge, a parcel sitting behind you. A sliding webcam cover physically closes the lens when you are not chatting, and a neutral acoustic panel or a simple folding screen behind you removes, in one move, all the information your surroundings broadcast. We devoted an entire article to what your background says about you.
4. Technical discretion has known limits
A VPN hides your IP address from the platform and from the person you are talking to; it hides nothing of what your camera shows. A password manager stops a chat account from sharing its login with your work email. A wired headset with a noise-cancelling mic keeps your conversation from being overheard in the next room — a trivial detail, but that is where many involuntary outings begin. For the rest, our analysis of what your IP address really reveals sets out what works and what is just marketing.
5. Choose a space where homophobia is beside the point
This is the simplest lesson from SOS homophobie's reports: a significant share of reported insults occurs in general-purpose spaces, where an LGBT+ person is a statistical anomaly for the aggressor. In a dedicated space, the same behaviour becomes marginal and is immediately reported by the community. It is not a guarantee — it is a change in baseline probability. We explain it in our article on homophobic reactions on a mixed chatroulette, and our video chat statistics give the orders of magnitude of usage by country.
Sources to consult directly
Rather than trusting a figure picked up third-hand, here are the publications to open for yourself:
- SOS homophobie, Rapport annuel sur les LGBTphobies — published every spring, free, with a breakdown by context and by motive.
- SSMSI (Interior Ministry), annual tally of anti-LGBT+ offences recorded by the security forces.
- EU Agency for Fundamental Rights (FRA), LGBTIQ Survey — the reference on lived experience and under-reporting.
- ILGA-Europe, Rainbow Map and Annual Review, and ILGA World for legislation outside Europe.
- Platforms' DSA transparency reports, available through the European Commission's transparency database.
- Défenseur des droits (France) for decisions and recommendations on discrimination.
One methodological detail deserves attention: always check whether a figure refers to reports, recorded complaints, convictions or self-declared experiences. These four populations are nested within one another, and confusing the smallest with the largest is the most widespread error in press coverage.
Key takeaways
Online homophobia is both better documented than it was ten years ago and still largely invisible. Platforms know how to filter written insults; they are almost incapable of reading a live video feed. Institutions record complaints; they do not record the fear of outing yourself at the front desk, which holds most people back.
For men who chat on video with other men, that translates into a simple hierarchy: reduce what you expose before you need to defend yourself, report systematically, keep records off your main device, and favour spaces where homophobia is beside the point rather than merely banned by the terms of service. Our safety page gathers the essential reflexes, and our comparison of gay dating apps versus video chat shows what each model exposes about you.
