Camyster

Bots, stolen accounts and fake couples: spotting rigged profiles on an 18+ chatroulette

· L'équipe Camyster

Bots, stolen accounts and fake couples: spotting rigged profiles on an 18+ chatroulette

The person on the other side smiles, fixes their hair, takes a sip. The picture is sharp, the framing careful, the lighting flattering. Then, forty seconds later, they fix their hair in exactly the same way, take the same sip from the same glass, and smile at the same moment. This isn't a person: it's an eight- or ten-second loop playing on repeat, while a script types messages into the text chat.

It's one of the most ordinary mechanics of 18+ random video chat, and one of the least documented. There's plenty of talk about age checks, moderation and personal data. But the question adults actually ask themselves the moment the screen lights up is usually more down-to-earth: is there really a living person in front of me, and is that person really what they appear to be? This article offers observable markers — no tools, no technical skill, and never any description of explicit content.

Older man with a white beard in an orange t-shirt using a smartphone held in both hands outdoors

Why these streams exist: three different economies

Not every fake presence is chasing the same goal, and that's worth knowing: the goal determines the signals.

1. Commercial redirection. The most widespread. An appealing stream holds your attention for thirty seconds, then a message arrives: "I can't talk here, find me on…", followed by a link, a username to look up, or a QR code held up to the camera. The aim is a paid sign-up on another site, paid out on commission. The video stream is nothing but bait, usually lifted from content published elsewhere.

2. Harvesting. The goal: obtaining a lasting identifier — phone number, messaging account, email address — in order to leave the platform, where there is no longer any moderation or reporting trail. This is the classic entry point for romance fraud, which the Direction générale de la concurrence, de la consommation et de la répression des fraudes (DGCCRF) and Cybermalveillance.gouv.fr have documented for years under the heading of "arnaque aux sentiments".

3. Volume. Some streams want nothing more than to artificially inflate the number of connected users, so a service looks busy. Those bots ask for nothing, barely reply, and vanish after a fixed interval.

Understanding this typology avoids a common mistake: believing that only "fake dating profiles" exist. On a chatroulette, the vast majority of fake streams are simply videos, not people telling lies.

The signs of a pre-recorded video

A video loop, however well chosen, is still a loop. Six clues are enough, and none of them requires more than thirty seconds of attention.

  • Exact repetition. A trivial gesture — scratching the back of the neck, adjusting a strap, glancing to the right — that comes back at regular intervals, identically. A human never reproduces the same micro-gesture twice at the same tempo.
  • A gaze that isn't looking for you. Live, people look at the screen, so slightly below the lens. In a video shot to be replayed, the gaze is often fixed on the lens, or sweeping across nothing.
  • The cut. At the loop point there's almost always a micro-jump: a shift in posture from one frame to the next, a change in the light, a hand that teleports.
  • Sound that's absent or unrelated. Many of these streams are silent, or play music. When there is a voice, it never answers a specific question.
  • Quality that's too good. An ordinary webcam in a real bedroom produces digital noise, imperfect white balance, a slight softness of focus. A perfectly stabilised, colour-graded image with a bokeh background usually comes from a shoot, not from a random connection.
  • No reaction to the unexpected. Live is unpredictable: a noise at your end, a barking dog, a change in light. A video reacts to nothing.

The two tests that settle it

No expertise needed. Two simple, politely phrased requests almost always decide the matter.

The number test. Ask the other person to hold up a specific number of fingers — three, then two. A human does it in a second. A loop never will. A script will reply "sure" in the text chat while the picture stays unchanged.

The object test. Ask them to move an object visible on screen, or to make an unlikely gesture: put a hand on the wall to the right, turn a book towards the camera. It's a variation on the liveness check that professional identity verification systems use themselves — age estimation and identity checks rest on the same principle: proving there is a living body in front of the lens.

If the other party refuses both requests as "weird" while continuing to push an external link, you already have your answer.

Real-time deepfakes: where things actually stand

Live face-swapping tools exist, are easy to get hold of, and run on consumer hardware. The Agence nationale de la sécurité des systèmes d'information (ANSSI) and Europol have both published analyses of the criminal use of synthetic content, and CEO fraud via faked video calls is now a documented case.

In random video chat, however, these fakes remain fragile, because they require frame-by-frame computation. The visible weak points are fairly consistent:

What you ask forWhat gives the fake away
Turning the head to a full profileThe mask peels away at the ear, the jawline distorts
Passing a hand in front of the faceThe face "smears" or rebuilds itself with visible lag
Bringing the face very close to the cameraThe outlines of hair and glasses shimmer
Speaking fast, laughing out loudThe mouth falls out of sync, the teeth go blurry
Suddenly changing the lightingThe skin doesn't follow the shift in colour temperature

One non-technical detail is often even more telling: the consistency of the setting. A face that's too smooth on a body whose hands, neck and shoulders have neither the same texture nor the same apparent age is a strong signal.

The special case of fake "couples"

This is the most profitable staging of all, because the demand is genuine: plenty of adults are looking to chat as a pair with another duo. Three variants keep coming up.

The couple that isn't one. Two people who don't know each other, or a single person with a video inset next to them. A recurring sign: the two never look at each other, never speak to each other, never touch each other's arm or shoulder. A real pair talks over each other, laughs at an inside joke, turns towards one another. A staged performance looks at the lens.

The couple that's selling. The script arrives within a minute: a mention of a "private room", a payment, a platform to join. The real etiquette of a meeting between two couples, set out in this article on the codes and everyone's consent between couples, is the opposite: you start by talking, you check that everyone agrees, and nobody is in a hurry.

The couple that's recording. The most troubling of all. One keeps the conversation going while the other films from off-camera. Recording a person without their consent and distributing that recording fall, in France, under articles 226-1 and 226-2 of the penal code, a subject covered in detail in "18+ doesn't mean no rules". The warning sign is physical: a second screen lighting up the face, a phone propped at an angle, repeated glances towards a fixed point off-frame.

The phrases that should make you hang up

The scenarios are scripted, so they repeat. When one of these sequences shows up, you already know the rest of the conversation.

  • "My mic isn't working, write to me on [another app]" — an immediate move away from any moderation.
  • "Just verify your age on this site, it's free, you only need a bank card" — a fake age check used to capture a payment method. The actual legal obligations, and how platforms apply them, are described in this feature on Arcom, London, the EU and Moscow.
  • "I'm stuck in another country, I need a top-up" — classic romance fraud, often accelerated to test the target within minutes.
  • "Turn your screen brightness up, show your whole face, I don't believe you" — groundwork for a usable screen capture.
  • "Can you give me your Telegram?" within the first thirty seconds, with no prior conversation at all.

One principle covers it all: the sooner the request to leave the platform arrives, the less the other person is interested in talking.

Reducing your exposure

You can't stop fake streams existing, but you can make their work pointless.

First, how identifiable your surroundings are. A medicine label, a letter, a framed diploma, a street sign visible through the window: all of these allow someone to cross-reference an identity. A simple foldable green screen placed behind the chair, or an indoor screen divider, settles the question in a second — and comes in handy for work video calls too. The more intimate details of a shared home are explored in the article on protecting your household when the camera comes on.

Next, the hardware. A webcam with a built-in privacy cover solves the problem of a stream that stays active after a tab is closed — a more common occurrence than people think on browsers loaded with extensions. And for long sessions, a closed-back headset with a mic avoids broadcasting the voices in the room, children's and neighbours' included.

Finally, lighting. This isn't just about looks: decent frontal light, such as a ring light for webcams, reduces the shadowy areas where you can't tell whether the other person is holding a phone — and, symmetrically, makes your own image legible, which discourages people who count on blur to make you doubt yourself.

A notebook, rather than your memory

The scripts repeat from one session to the next. Jotting down on the fly, in a small spiral notebook kept beside the keyboard, the exact phrase and the time of a suspicious behaviour serves two purposes: it lets you recognise a script next time round, and it gives you precise details if you file a report. A dated report quoting the wording you received is handled far more effectively than "there was a weird one tonight" — how teams actually work is described in this article on moderation, reporting and bans.

What to do once you've worked it out

No confrontation. Arguing with a script achieves nothing, and arguing with a human operator who is lying gives them information about you.

  1. Cut the stream. Move to the next one, without explanation.
  2. Click nothing. No link, no QR code, no "verification site". The redirects frequently lead to phishing pages or hidden-subscription sign-ups.
  3. Report before moving on. Thirty seconds, and it's the only action that reduces the volume of these streams for everyone else.
  4. If money or pressure is involved, keep the evidence and turn to Cybermalveillance.gouv.fr, or the PHAROS platform to report illegal content. In Russia, Ukraine, Kazakhstan and Belarus, the available remedies differ, and the backdrop of communication restrictions makes things harder — a point addressed in this article on throttled messaging apps and VPN rules.

The question isn't "is my eye sharp enough?", but "did I ask for one simple thing before investing myself?". Three raised fingers cost a second. A phone number given to the wrong person costs far longer than that.

The essentials in five lines

  • A large share of fake streams are looping videos, not liars: look for repetition, not psychological inconsistency.
  • Two tests are enough: a number of fingers, an object to move.
  • A real-time deepfake breaks down on a full profile, a hand in front of the face and a hearty laugh.
  • A "couple" who never look at each other and mention money within a minute is not a couple.
  • Any request to leave the platform within the first thirty seconds is, statistically, the opening of a script.

Recognising these set-ups doesn't make you suspicious: it frees up attention for the conversations that are real. And they're more common than people say, provided you haven't handed your time to the ten manufactured streams that come before them. To dig deeper into how trustworthy a service itself is, the four-point method in recognising a reputable 18+ chatroulette is a useful complement to these individual habits.

Ready to meet people by webcam?

🎥 Start now
🎥 Start now