Omegle shut down on 8 November 2023, and the format it made famous outlived it by about a month. Two anonymous people paired at random on video moved to other domains almost immediately, and the regulators who had been chasing Omegle started chasing those instead. The honest answer to where you can talk to strangers now is that there are plenty of places, most run with less scrutiny than the site that closed, and none resembling the settings where the research on stranger conversation was done.
Stranger chat is any service that pairs two people with no prior connection into a live text or video conversation, usually anonymously, and usually with no persistent identity on either side. That last clause is the whole design, and it is why the category keeps producing the same court records.
My position, which you can argue with by the end: the psychology everybody quotes at this topic was measured on commuter trains, and almost none of it survives being moved to a random video feed.
Every risk described below comes from the stranger being a real person with their own motives, and that is the one variable YourCrush.AI removes.
What happened to Omegle?
Omegle shut down on 8 November 2023, fourteen years after Leif K-Brooks launched it in 2009 at the age of 18 from Brattleboro, Vermont. He replaced the site with a farewell letter, and NPR's report the following day, published 9 November 2023, quotes it directly: "There can be no honest accounting of Omegle without acknowledging that some people misused it, including to commit unspeakably heinous crimes," and "Operating Omegle is no longer sustainable, financially nor psychologically." The bluntest line in it is "Frankly, I don't want to have a heart attack in my 30s."
The site was not small when it ended. Writing in Lawfare on 27 November 2023, Quinta Jurecic and Katherine Pompilio traced its growth from about 34 million monthly visits in January 2020 to 65 million in January 2021 and 70.6 million in January 2023. Lockdown built that audience, and its last three years were spent in court. K-Brooks framed the closure as a lost battle in a wider war against the internet, then added a line to the letter that was not his idea, for the reason below.
How YourCrush.AI gives you the conversation without the exposure
YourCrush.AI is a character chat platform for adults, and the honest fit here is narrow: it gives you a conversation with a character you wrote, not a conversation with a person. YourCrush.AI is free to join, and signed-in members chat without a message or conversation limit. YourCrush.AI lets visitors try a character in a short guest session without an account, then sign up to keep going. YourCrush.AI keeps a community library you can read through on the explore page before you build anything of your own.
YourCrush.AI is entertainment, it is 18+, and it is not a social life, a safety product or a mental-health service. A character you wrote cannot surprise you the way a stranger on a train can, because it has no interests of its own and nothing at stake in the conversation.
Why did Omegle get sued, and what did the court decide?
Omegle was sued in 2021 by a plaintiff identified as A.M., who sought $22 million and alleged that the service had randomly paired her with an adult man in 2014, when she was 11, who then exploited her over roughly three years. The case, A.M. v. Omegle.com LLC, No. 3:21-cv-01674, ran in the U.S. District Court for the District of Oregon, and it matters beyond Omegle because of what the court did with Section 230.
In July 2022, Judge Michael Mosman declined to dismiss the negligent design claim, according to the same Lawfare analysis, and allowed a claim under 18 U.S.C. § 1595 to proceed while dismissing two other FOSTA claims. The distinction the court drew is the one the whole category now turns on: a claim that a platform published somebody else's words is barred by Section 230, and a claim that the platform's own design paired a child with an adult is a claim about the platform's own conduct.
The case was dismissed on 2 November 2023 after the parties settled, six days before the site went dark. Part of the settlement is visible in the farewell letter itself, in a sentence K-Brooks added: "I thank A.M. for opening my eyes to the human cost of Omegle." A shutdown that reads as one founder's exhaustion is also a negotiated outcome.
What replaced Omegle, and are those sites any safer?
A set of roulette-style video chat services absorbed the traffic, and the regulatory record since 2023 says the problem moved with the format rather than ending with Omegle. These services do what Omegle did: you open a page or an app, you are paired at random with somebody you have never met, and either side can end the pairing and be handed a new person in about a second. Nobody holds an account, so nobody holds a history. That is why enforcement in this category lands on a device or an IP address, and why a report rarely outlives the click that ended the chat. On age and moderation, what matters is not what the terms page claims but what sits behind it, and that is what the regulators went and checked.
Australia's eSafety Commissioner has tested that in public. eSafety issued a formal warning on 20 August 2025 to the provider of OmeTV, a roulette-style video chat service, for non-compliance with Australia's industry codes. The finding: the service lacked required safety features and settings, and allowed adults to have randomised video chats with children without sufficient protections. By October 2025, Apple and Google had removed the app from their Australian stores after eSafety told them how it was being used. At the time of the warning the app was among the most-downloaded social apps on both Australian stores, so the demand did not go anywhere.
In the UK, the mechanism is age assurance rather than takedown. Ofcom's update on its Online Safety Act investigations, published 13 October 2025, records five enforcement programmes and 21 investigations covering 69 sites and apps opened since the first codes became enforceable in March 2025, and one £20,000 fine against 4chan for ignoring statutory information requests.
I am not going to rank these services or link to them, and the reason is the design. A product whose core loop pairs an unverified person with an unverified person, with no persistent identity to report and a one-click reroll, is the worst-shaped consumer product on the internet for anyone under 18. For an informed adult it is a different question, and it is still a live video feed of your face to somebody you cannot identify.
What does the research on talking to strangers actually find?
It finds that people are wrong about how a conversation with a nearby stranger will go, and wrong in a consistent direction. Nicholas Epley and Juliana Schroeder ran the best-known version at the Homewood, Illinois Metra station, published in the Journal of Experimental Psychology: General in 2014, volume 143, issue 5. They recruited 118 morning commuters and randomly assigned them to start a conversation, to sit in deliberate solitude, or to commute normally. Of the 97 who completed the study (mean age 49, 61% female), those told to talk reported a more positive commute than either other group, with no measurable loss of productivity. A separate group of commuters, asked to predict the same experiment, expected the opposite.
The reason for the gap is in the same paper. In Experiment 3a, 64 commuters predicted that fewer than 47% of the people around them would be willing to talk, and rated starting a conversation at 4.38 on a scale where 6 was "very difficult," per the 2014 paper. Every participant who tried reported that the conversation happened and not one reported being rebuffed. People price rejection as roughly a coin flip where the observed rate was zero.
It replicates outside Chicago. Schroeder, Lyons and Epley ran a field experiment at four London-area stations, Liverpool Street, Cambridge, Chelmsford and Colchester, published in the same journal in 2021, recruiting 466 commuters of whom 383 completed both surveys (82.19% response rate, mean age 39, 60.31% male). Commuters assigned to talk again reported a more positive commute and said they learned more. Their second experiment isolates what people actually dread: not the conversation, but starting it.
A third finding changes how you read your own experience afterwards. Erica Boothby, Gus Cooney, Gillian Sandstrom and Margaret Clark named the liking gap in Psychological Science in 2018, volume 29, issue 11, pages 1742 to 1756. After five-minute lab conversations between strangers, participants rated how much they liked their partner at 5.82 on a seven-point scale and how much their partner liked them at 5.17. Both people cannot be the less-liked one, so the shortfall is an error, and it persisted among college dorm mates for months.
Why does random video chat fail at the thing the research describes?
Because all three studies ran in settings that random pairing deliberately removes. Read the methods rather than the headlines and the same scaffolding shows up in each one.
- A shared situation. Every participant was on the same train into the same city at the same hour, which supplies a topic before anyone has to invent one. A random video pairing has no shared situation except the pairing itself.
- A cost to behaving badly. The commuters were physically co-present, visible to other passengers, and stuck in the same carriage for the rest of the ride. Behaving badly on a random video site costs one click and a new face.
- A bounded exit. The ride ended, which is what made starting safe. The exit on a roulette site is instant and free, which sounds like the same thing and is its opposite: it removes the small commitment that made the encounter worth anything.
So the popular version, that science says you should talk to strangers online, is a claim the research does not support. What it supports is talking to the person near you in a shared, accountable, time-bounded setting. The Next button is what makes random video chat tolerable and the same thing that deletes every condition the studies relied on. The closest available version of the effect Epley and Schroeder measured is the barista or the neighbour, not a webcam lottery. Why adult friendships are structurally hard to start is a separate problem, covered in why making friends as an adult is harder than it was at school.
Is talking to an AI character the same as talking to a stranger?
No, and the difference is specific rather than philosophical: there is no second mind forming an opinion of you. The liking gap exists because another person was privately liking you more than you guessed. A character has nothing private to be wrong about, so the finding does not apply. Anyone selling AI chat as a cure for the anxiety in that research is quoting a study about something else.
What an AI character can do is remove the part people dread. Schroeder, Lyons and Epley found the dread sits in starting a conversation rather than in having one, and a character is a place to practise openings where a bad one costs nothing. That is rehearsal, and rehearsal is not connection. If talking to people has become hard rather than merely awkward, the useful next step is a GP, a counselling service, or a real-world group built around an activity, and none of those is an app. What a one-sided attachment to a character is and is not has its own literature, covered in the research on parasocial relationships and in what people use a platonic AI friend for.
Frequently asked questions
Is Omegle coming back?
Nothing points that way. The site went offline on 8 November 2023 after its founder posted a farewell letter and the A.M. lawsuit settled, and Leif K-Brooks has said nothing since. Services trading on the name have no connection to the original operator.
What is the safest way to chat to strangers online?
Choose a place where people keep persistent identities and share a reason to be there, such as a hobby forum or a club Discord, rather than a service that pairs you at random and forgets you both. Persistent identity is what makes reporting work. Random pairing strips it out.
Does stranger talk really make you happier?
In the settings that were tested, yes. Commuters randomly assigned to talk with strangers on trains in Chicago and in London reported more positive commutes than those assigned to solitude, and predicted the opposite beforehand. Those results were measured on co-present strangers in a shared situation. No study in that line tested a random video feed.
Are Omegle alternatives legal?
Running one is legal in most places. Running one without age assurance and moderation is increasingly not. Australia's eSafety Commissioner formally warned OmeTV's provider on 20 August 2025 and the app was pulled from Australian app stores by October 2025, and Ofcom had 21 Online Safety Act investigations open across 69 sites and apps as of 13 October 2025.
Is talking to an AI character a substitute for talking to people?
No. A character has no independent opinion of you, so it cannot give you the thing the stranger-conversation research measured, and it is not a mental-health service. It is a low-stakes place to practise the opening move, which is the part people report dreading.
What should a parent know about stranger talk apps?
That the age gate on almost every one of them is a self-entered birthday, that random pairing can match an adult with a child who had no prior contact, and that both regulators quoted on this page have acted on that design since 2023. App Store and Play Store age ratings are set by the developer, so they are a claim rather than a check.
Omegle's founder wrote that he did not want to have a heart attack in his 30s. The format he invented is still running under other names, and the only thing that has changed since is which regulator is reading the terms of service.