Therapists, What Are You Trading Away for a Few More Cases?

It has not happened here yet. Not fully. That is exactly why this needs saying now and not in three years, when it will already be too late to ask for different terms.

A handful of therapy platforms in India have started building AI into the actual session, not just the booking page. Not the chat support widget that helps a client reschedule. The session itself, recorded, running inside an app the platform owns, with a note somewhere in the fine print about "quality and training purposes." It is still small. Most of the field has not been touched by it yet. But small is not the same as slow, and I would rather sound early than sound right after it has already stopped being useful to sound anything at all.

Here is the deal actually being offered, even if nobody says it this plainly. Build AI into the platform, and funding gets easier, because "AI-powered" is the sentence that gets a cheque signed faster than "therapist directory" ever will. Get your sessions running on that platform, and you get more referrals, more cases, a fuller calendar, faster than word of mouth could ever fill it. That part is real, I am not pretending the incentive is imaginary.

A newer therapist trying to build a practice in a crowded city, a full calendar this quarter is not nothing. And it is not just individual therapists feeling that pull. Globally, right now, "AI-powered" is the label that gets a health-tech company funded faster and valued higher than the same company without it, and Indian mental health startups are not sitting outside that pressure, they are inside the same funding conversations, chasing the same round sizes. A platform does not need to announce a master plan for this to happen. It just needs a founder in a pitch meeting who has noticed which sentence makes the investor lean forward.

But look at what is actually sitting on the other side of that trade. Client data leaving your hands the moment the session starts, not because you decided to share it but because the app is where the session happened. Consent reduced to a line the client scrolled past to get to the "book now" button. And the thing nobody wants to say out loud yet - your own future, quietly financing itself out of your own sessions, one recording at a time.

The Part You Are Telling Yourself Is Fine

"It's just for quality and training purposes." I know that sentence, you have probably already heard a version of it, maybe said it back to a client who asked why the session was being recorded. It sounds administrative. It sounds like something a compliance team wrote, and mostly it is. But training whose model, for what eventual product, is a question that sentence is built to make you stop asking.

"Everyone is moving this way anyway." Maybe. Watch what that sentence actually does though - it turns a choice into a weather pattern, something happening to you rather than something you are agreeing to. Nobody is making you sign this particular empanelment form. Somebody is making it easy, and comfortable, and slightly embarrassing to be the one person in the group asking uncomfortable questions about it. Those are different things.

"I need the referrals right now." This is the honest one, and I am not going to pretend it isn't real. I would just ask you to sit with what you are actually financing when you take that trade. Not just your own calendar. Possibly the very thing that ends up filling it with fewer clients five years from now.

And this rarely arrives labelled as a threat. It arrives labelled as a convenience. First it is a note-taking assistant, something that listens to the session and writes up your case notes so you don't have to stay late doing it yourself, and that part genuinely helps, I am not going to pretend admin work is sacred. Psychologists everywhere are adopting tools like this fast, nearly three times as many using AI at least monthly compared to just the year before. Nobody hands you the whole plan on day one. You get the part that saves you an hour, and only later do you find out what else that hour of audio was quietly useful for.

Why This Will Work Here Faster Than It Worked Anywhere Else

This is the part that should actually worry you, more than the recording itself. In India, the appetite for an AI substitute already exists, at a scale most of us have not fully sat with. A recent survey of Indian youth found that more than half already turn to AI chatbots for emotional support, and among school-aged respondents using AI during moments of anxiety, the number climbs close to nine in ten. More than half of young women surveyed said they tell an AI things they would not tell another person. A meaningful share said using AI made them less likely to reach out to a human afterward.

None of that happened because a chatbot is a better listener than a trained clinician. It happened because access here is broken in a way most other countries are not dealing with at the same scale. India has something like 0.75 psychiatrists for every 100,000 people, against a WHO recommendation of three. Somewhere between 70 and 92 percent of people with a diagnosable mental health condition in this country get no formal treatment at all. A chatbot that is available at midnight, costs nothing, and does not require you to explain yourself to a receptionist first is not competing against good therapy. It is competing against no therapy, which is the actual alternative for most people most of the time here.

This is not a story about a handful of anxious teenagers with nothing better to do. Eighty three percent of Indians who need mental health care get none, and that gap is exactly the size of opportunity a founder pitching a mental health app to investors is describing when they talk about "the market." Nobody building that pitch deck is lying about the size of the need. They are just deciding, quietly, whether the answer to that need looks like more trained humans, or looks like one well-trained model that can serve a hundred thousand people at once for the price of the compute.

Now put those two things next to each other. An enormous, already-proven appetite for AI emotional support on one side. A slowly growing pile of real, culturally specific, regional-language therapy session data on the other, sitting inside platforms that are only just starting to notice what they are holding. You do not need a dramatic villain-in-a-boardroom story for this to go badly. You just need both of those trends to keep moving at their current, ordinary pace. A well-trained model, trained on real sessions in the languages people actually cry in, offered for free or near-free to a population that already prefers texting a chatbot at 1am over waiting three weeks for an appointment, is not a hypothetical. Everything required to build it already exists except the training data. Guess what a session recorded "for quality purposes" quietly becomes.

And this is not a fear I am inventing for the sake of an argument. Earlier this year, mental health workers at Kaiser Permanente in Northern California went on strike, in part over their employer using an algorithm instead of a clinician to triage patients calling in. Not a hypothetical model somewhere down the line. A live one, already making the first call on how urgent someone's suffering was. That happened in a market with far more regulation than ours, far older unions, far more institutional memory about where these lines are supposed to sit. It still happened. We do not get to assume it happens to us slower just because we noticed it later.

What Recording Actually Captures

It was never really about the name on the transcript. Strip that off and the platform still has the parts that matter more. What a good clarifying question sounds like around minute ten. What silence should be met with and what it shouldn't. How a Hindi-speaking client in her fifties describes shame differently than a twenty-two-year-old talking in English about the same thing, and what response actually lands for each of them. That is not identity, in the narrow legal sense a form asks about. That is clinical skill, and it is exactly what a model needs to learn in order to start sounding like someone who knows what they are doing.

And notice the asymmetry sitting underneath all of it. When you take informed consent from a client, you actually mean it. You tell them what stays private, when you would have to break that, and you make sure they understood before they agreed to anything, because that is not paperwork to us, it is the ethical spine the entire profession stands on. Now ask whether the app they downloaded before they ever found you extended them anything close to that same courtesy about what happens to the recording afterward. It gave them a terms of service document nobody reads, written to be technically compliant and practically invisible. We hold the room to a standard that the platform hosting the room was never built to meet.

This will feel familiar to some of you, because we have already lived a smaller version of it. Any of us who has run a "free" wellness workshop for a school or a corporate, because they offered exposure and a captive audience, already knows exactly what that audience means to the people asking for it. It means come do the real work, we will keep the value. Data extraction is the same instinct, just wearing a more technical outfit and a nicer pitch deck. Same question every single time, whether it is a free workshop or a free platform. Whose interests come first when nobody involved is being paid for the actual thing being taken from them.

Ask This Before You Sign Anything Else

Here is the actual test, and you can run it on any platform sending you referrals, any panel asking for your empanelment, any app promising you a fuller calendar by next quarter. Ask what the company is actually building. Is it a directory that happens to have a booking system, where you are the product being placed in front of clients? Or is it, underneath the therapist-friendly language, an AI company that currently needs a network of human therapists to generate the material it is training on, and will need that network a lot less in a few years? Those are different businesses wearing the same interface, and the second one is not lying to you exactly. It is just not telling you the whole plan, because the whole plan is still being written in a room you are not in.

We have a data protection law now, DPDP, which is a start, and it is still a young one, still being written into enforceable rules while every one of these platforms moves faster than any regulator possibly can right now. Nobody outside this profession is going to slow this down on our behalf. That part is entirely on us. Doctors have professional bodies that took decades to build real teeth. We do not have to wait decades to at least start asking one question consistently, out loud, before we sign anything, the same way we would ask about supervision or liability cover.

I keep coming back to something simple underneath all of it. A client decides, somewhere in that first difficult hour, that this room is safe enough to finally say the thing out loud. That decision, made one person at a time, is the entire profession, the only raw material we actually have. Every platform we plug ourselves into is quietly deciding what happens to that decision once it leaves the room. Right now, in India, that decision is still mostly ours to make. It will not stay that way by accident.

So ask, before the next platform and the next empanelment form and the next promise of a fuller calendar. What is this company actually building, and what is it building it out of. Say the answer out loud to yourself before you sign anything. If it is what your client just finished trusting you with, you already know what you are trading it for, and whether it is worth it.

Open your eyes before the window closes, not after.