Are Your Patients Using AI Instead of Therapy?

A therapist we work with told us about a new patient who sat down, opened her notes app, and read off a full list of what was wrong with her. Diagnoses, patterns, the name of a therapy approach she wanted to try. The therapist asked where all of that came from. The answer: she had spent three weeks talking to an AI chatbot every night before she finally booked. The bot had been her waiting room. By the time she walked in, she had already been in a kind of treatment for a month, just not with a person.

That is not a rare story anymore. It is quickly becoming the normal path into your office, and the numbers behind it are worth every practice owner's attention.

The waiting room moved into a chat window

In March 2026, KFF found that 32 percent of adults had used an AI chatbot for health information in the past year, and 16 percent had used one specifically for mental health. That is now the same footprint as social media as a health resource. A separate look by West Health and Gallup put it at one in four adults, more than 66 million Americans, using AI for health advice, sometimes before a visit, sometimes instead of one.

For mental health the pull is even stronger with younger people. KFF found adults under 30 were about three times as likely as adults over 50 to use AI for mental health information, 28 percent versus 8 percent. The reasons are painfully logical: it is free, it is instant, and it never makes you explain yourself to a receptionist. When affordability and access are the walls that keep people out of care, a chatbot walks right through them.

None of this is surprising once you remember how people actually search now. We wrote about this shift in why patients search symptoms, not doctors. The chatbot is just the next step down that same road. The person types the feeling, not the specialty, and the machine answers before your name ever comes up.

The real problem is not that AI replaces you

Let us be clear about the threat, because the scary headline is the wrong one. AI is not going to do your clinical work. The American Psychological Association's 2026 report on chatbots and mental health is blunt about it. Among psychologists surveyed, 94 percent said chatbots lack the nuance to treat real conditions, 97 percent worried they can reinforce negative thoughts or delusions, and 89 percent were concerned a chatbot could encourage self harm. The APA's own line sums it up: AI generates responses, but it does not reason like a mental health professional.

So the danger is not that the bot is a good therapist. The danger is what the KFF data shows next. Among people who used AI to ask about mental health, 58 percent did not follow up with a real provider afterward. More than half. They asked, they got an answer that felt good enough, and they stopped there. That is the number that should keep you up, not some fear of robots taking over therapy.

Read the APA report closely and you also see the risk to those patients. Psychologists reported patients using chatbots for self diagnosis (39 percent had seen it), as a stand in therapist (33 percent), even for friendship and intimacy. They also saw the fallout: 36 percent noticed patients forming a dependency on the bot, and 15 percent saw distorted thinking tied to it. In other words, some of the people talking to a chatbot every night are exactly the people who most need a human, and the bot is slowly becoming the reason they never reach one.

Your job now: be the easy human next step

Here is the healthy way to think about it. The chatbot is not your competitor. It is the new front door, the place a nervous person goes first because it asks nothing of them. Your practice does not need to beat the bot at 11pm. It needs to be the obvious, painless next move the moment that person decides a real human might help. Most practices are losing patients not to AI, but to the gap between the chatbot and their own booking page. Close that gap and you win the patients the bot was never equipped to keep.

That comes down to three unglamorous things.

1. Show up when the search starts, including inside AI answers

More people now read an answer from Google's AI or a chatbot and never click a website. If your practice is invisible in that layer, you do not exist to them. The fix is not a trick, it is clarity. AI tools build their answers from clear, credible, well organized information: a complete Google Business Profile, real reviews, and plain factual pages that answer the questions patients actually ask, like what conditions you treat, what insurance you take, and how someone gets started. That same groundwork is the heart of healthcare SEO, and it is what makes a bot mention you instead of a hospital three towns over. We broke this down in how to get your practice recommended by ChatGPT.

2. Make the website answer like a person would

People are choosing chatbots partly because your website reads like a brochure and theirs reads like a friend. A mental health site that works today speaks plainly, names the hard feelings out loud, and answers real questions without jargon. It says what a first session is like, whether you take their insurance, how long the wait is, and that it is okay to feel unsure. A warm, honest, fast site is the closest thing to that late night chat, except it ends with a real appointment. That is the whole point of a website built for a mental health practice, not a generic template with a stock photo of a sunset.

3. Kill the phone call as the only way in

This one matters more in mental health than almost anywhere. For a lot of people, picking up the phone and explaining why they need help is the single hardest step, harder than the therapy itself. That fear is a big reason they stay in the chat window, where nobody judges them. So let them book without it. Private online scheduling, where a new patient picks a time and confirms in under a minute without saying a word to a stranger, removes the exact wall that sends them back to the app. If a bot can help them at midnight with zero friction, your booking cannot demand a 9am phone call and a game of voicemail tag. For the person who does want to talk or has an off hours question, our AI receptionist can answer, reassure and book them right then, so a late night moment of courage does not die in a call queue by morning.

What the chatbot cannot do, and you can

The good news buried in all this data is how weak the bot really is at the thing that matters. It cannot sit with someone. It cannot notice the pause before an answer, or carry any responsibility for whether that person is safe tonight. The APA numbers say clinicians already see the damage when people lean on it too hard. That is your entire value proposition, and it is bigger than the bot's, not smaller.

So the marketing job is not to out talk a machine. It is to make sure that the moment a person is finally ready for a human, you are the easiest human to reach. That means being findable, being clear, and being one tap from a booked appointment. When a practice gets those three right, the chatbot stops being a rival and starts being a feeder, warming people up and handing them off to you.

How EtherealMinds helps mental health practices with this

Most therapists and psychiatrists did not go into this work to fight an algorithm for attention, and they should not have to. When we build a patient acquisition system for a therapy practice, we make the practice easy to find in the exact places the search now begins, from Google's map results to the AI answers sitting above them. We build a website that talks like a caring human and lets a new patient book privately in under a minute, no phone call required. And we keep a steady, compliant flow of reviews and honest social content for mental health practices going, so the practice looks alive and trustworthy to a stranger who is scared and deciding. The chatbot got them thinking. We make sure the next step lands on you.

So, are your patients using AI instead of therapy? A growing share are using it first, and too many are stopping there. You will not win them back by pretending the tool does not exist. You win them by being the warm, findable, one tap human step the bot can never be. That is a race a good practice can absolutely win, and right now most are not even running it.