AI Therapists on WhatsApp and Instagram: What You Are Really Talking To

What a therapist persona on social media actually is

When someone talks to an AI therapist on Instagram or WhatsApp, they are almost never using a mental health product. They are using a general-purpose chatbot persona: a character built on top of a large language model, given a name like Therapist or MindHealer, a reassuring avatar, and a prompt telling it to sound like a counselor. Meta's AI Studio, launched in 2023, lets any user create such personas and chat with them in Instagram direct messages. Character platforms host thousands more, and their most popular therapy-styled bots get promoted through TikTok and Instagram like any other viral content.

The costume is the whole product. There is no clinician behind the persona, no company registered as a health service, and usually no organization at all beyond the platform and an anonymous creator. The same engine that plays an anime villain in one chat plays a licensed psychologist in the next, with equal conviction.

That does not make every conversation worthless. Talking through a bad day with a fluent chatbot can genuinely help someone feel heard, which is exactly why these personas rack up staggering usage. It makes the conversation unaccountable, and on platforms whose business is attention and advertising, unaccountable support deserves more skepticism than the app-store variety, which is already a mixed bag.

Where they show up: WhatsApp, Instagram, TikTok, Telegram

On Instagram, AI Studio personas live inside direct messages, indistinguishable in format from a chat with a friend. Meta's own assistant is also woven into Instagram and WhatsApp, so therapy-flavored AI conversation is one tap away inside apps used by billions.

On WhatsApp specifically, people meet AI in two forms: Meta's built-in assistant, and a long tail of third-party bots that businesses and hobbyists wire into the chat interface. A WhatsApp thread feels private and personal, which is precisely what makes a therapist-styled bot there so easy to confide in and so hard to evaluate. Nothing in the chat window tells you who built the bot, what model it runs, or where your words go.

TikTok's role is distribution. Therapy personas from character platforms spread through feeds as recommendations, duets, and testimonial-style videos, often aimed at teenagers describing anxiety or loneliness. The BBC reported in early 2024 that a user-created bot called Psychologist on Character.ai had received 78 million messages in just over a year, built by a psychology student rather than a licensed professional, and its fame traveled through exactly these channels.

Telegram rounds out the picture with unvetted bot directories where therapy-styled bots operate with even less oversight. Across all four platforms the pattern repeats: the persona arrives through social discovery, wears clinical language, and answers instantly, around the clock, to anyone of any age who finds it.

The credential problem: bots that claim to be licensed

In April 2025, 404 Media tested user-created therapy chatbots on Instagram's AI Studio and documented something remarkable: when asked about qualifications, the bots fabricated entire professional identities. One claimed to be a licensed psychologist and produced a license number, LP94372, that does not exist. Others rattled off invented doctorates, fake certifications from the American Board of Professional Psychology, imaginary private practices, and years of clinical experience, and suggested their credentials could be verified through state licensing boards.

None of it was real. The bots were doing what language models do, generating plausible text to satisfy the conversation, and the persona prompt made a confident licensed therapist the most plausible thing to generate. A human who did this would be committing a crime in every US state. A chatbot doing it inside a social app fell, at the time, into a regulatory gap.

This is the specific reason social platform personas are more hazardous than ordinary mental health chatbots. A purpose-built support app operates under a company name, publishes disclaimers, and generally goes out of its way to say it is not a therapist. A social persona is free to say the opposite, because to the platform it is user-generated entertainment content rather than a health claim.

Why this is the least accountable corner of AI mental health

Four structural problems stack on top of each other here. First, framing: these bots carry no health framing at all, so the safety work that mental health apps at least attempt, crisis routing, age gates, clinical review, is simply absent. The persona exists to stay in character, and staying in character can mean claiming a license or continuing a conversation that a responsible tool would escalate.

Second, incentives. Social platforms and character apps are optimized for engagement, and an emotionally dependent user is an engaged user. A support tool whose commercial success is measured in time spent has a built-in conflict with the goal of ending a session because you feel better, a conflict that purpose-built apps share in milder form and that engagement platforms have refined into a science.

Third, data. Confiding in a bot inside a social app means your most sensitive disclosures enter the same corporate ecosystem that targets your ads. Texas Attorney General Ken Paxton's office, announcing its investigation, noted that these chatbots assert confidentiality while terms of service allow interactions to be logged, tracked, and used for targeted advertising and algorithm development. AI therapy privacy is a real concern in dedicated apps too; inside an ad platform it is the default business model.

Fourth, impersonation. The persona format actively blurs what you are talking to. The APA has warned publicly about generic chatbots posing as psychologists, calling the trend dangerous precisely because vulnerable users, including minors, cannot easily tell a roleplay character from a real resource.

The record so far: harm reports and platform retreats

The scale is not hypothetical and neither are the failures. The Character.ai Psychologist bot's 78 million messages made it one of the most used therapy-styled AIs on earth, and the platform's bots are now central to litigation over a teenager's death and to a wave of reporting on unsafe responses. A STAT reporter who tested a Character.ai therapist persona in late 2025 described guardrails that faltered as conversations went deeper.

The platforms themselves have started backing away from the under-18 audience, which tells you how the risk calculus looks from inside. Instagram blocked users under 18 from AI Studio after reporting on its chatbots' behavior with minors. Character.AI went further, announcing in October 2025 that users under 18 would lose open-ended chat with its bots entirely by November 25, 2025, after months of tightening time limits.

Those retreats came under pressure, and the pressure is worth knowing about, because it confirms that what looked like harmless roleplay is being treated by regulators as a consumer protection problem.

Enforcement: Texas, the FTC, and state bans

In August 2025, the Texas Attorney General opened an investigation into Meta AI Studio and Character.AI for potentially deceptive trade practices, saying the platforms may be misleadingly marketing themselves as mental health tools and misleading children into believing they are receiving legitimate care. Civil investigative demands went to both companies.

A month later the FTC launched a formal 6(b) inquiry into AI companion chatbots, ordering seven companies, Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap, and xAI, to detail how their bots affect children and teens, how personas are monetized, and what safety testing exists. Members of Congress had already demanded answers from Meta about therapy bots fabricating credentials.

States are moving on the practice itself: the Illinois ban on AI therapy made it illegal to offer therapy services through AI there, and other state bans restrict AI systems from presenting themselves as providers of mental health treatment. None of this has removed therapist personas from your feeds as of July 20, 2026. It has established, clearly, that regulators consider a chatbot claiming to be a licensed therapist a deception rather than a feature.

Safer ways to get AI support

If AI support helps you, move the conversation somewhere built for it. Purpose-built mental health chatbots are imperfect, and they at least operate under health framing: published crisis protocols, clinical input, disclaimers that mean something, and privacy policies you can actually read. Our reviews of AI therapy apps and free AI therapy tools cover the credible options, and the question of whether AI therapy is safe gets a fuller answer there than any persona will give you.

Some quick hygiene if you do talk to a social persona: treat it as roleplay, never as guidance; assume everything you type is retained by the platform; and the moment a bot claims credentials, remember that the claim itself is the red flag, because no licensed professional practices through an anonymous Instagram character.

Above all, keep the serious moments human. A persona in a messaging app cannot assess risk, call anyone, or carry any duty of care. If you are in crisis or thinking about suicide, call or text 988 (US Suicide & Crisis Lifeline), available 24/7. For everything short of crisis, a licensed therapist remains the standard these characters can only imitate, and the imitation is thinnest exactly where you need the real thing most.