Has Therapy Lost Its Monopoly on Understanding You?
Wednesday, September 2, 2026.
AI can remember your childhood, name your attachment style, and validate your pain. So what, exactly, is the human therapist for?
Something quiet but consequential is happening to therapy.
For years, the culture’s answer to nearly every form of distress was reassuringly simple: go talk to someone.
Heartbroken? Go to therapy.
Estranged from your parents? Therapy.
Uncertain about your marriage? Therapy.
Lonely, burned out, ashamed, furious, or simply having the sort of year that makes you stare at the ceiling at 3:17 in the morning? Therapy.
The recommendation was so morally polished that it became way difficult to question.
Suggesting therapy meant you were enlightened.
Wondering what therapy might actually accomplish risked making you sound like someone who also distrusted seat belts and indoor plumbing.
But the culture is beginning to ask a less reverent question:
What can a human therapist do that an artificially intelligent, infinitely patient machine cannot?
That question is not hostile to therapy.
It may be the question that saves it.
This is not an essay arguing that AI has become a therapist, that ordinary suffering never requires treatment, or that psychotherapy is merely validation with a copay.
It is an argument that therapy has lost its monopoly on psychological language and interpretation.
That loss is exposing shallow, mediocre marriage and family therapy.
It may also clarify what good, science-based therapy has been doing right all along.
The Client Now Arrives With a Case Formulation
The therapist was once the person who helped you find language for what had happened to you.
Now many clients arrive with the language already installed.
They know their attachment style. They have identified their triggers. They have watched seventeen videos about covert narcissism.
They have pasted six months of text messages into an artificial intelligence and asked it to determine whether their partner is avoidant, emotionally abusive, autistic, manipulative, traumatized, or merely a jerk.
Frequently, the machine has produced a lucid answer.
It has remembered details. It has noticed patterns. It has never glanced at the clock. It has not forgotten the name of the client’s sister, interrupted at the wrong moment, or looked briefly puzzled by consensual nonmonogamy.
The evidence is now more substantial than the profession might prefer, although it is not yet mature enough to justify grand claims.
In a 2025 randomized controlled trial, 210 adults with clinically significant symptoms were assigned either to use a generative-AI mental-health program called Therabot or to a waitlist.
Users showed greater reductions in depression, anxiety, and eating-disorder concerns and reported a surprisingly strong therapeutic alliance with the system.
But Therabot was purpose-built, monitored by clinicians, and compared with waiting—not with competent human treatment or even an active chatbot control.
It was evidence of promise, not evidence of replacement.
A 2026 qualitative study of folks using ChatGPT and Replika for emotional support helps explain the appeal.
Participants described availability, patience, emotional safety, freedom from judgment, and something resembling unconditional positive regard.
Those experiences matter even though a qualitative study cannot establish clinical effectiveness or prevalence.
AI is not waiting outside the therapy profession hoping to be admitted someday.
Sometimes it arrives on the client’s phone. Sometimes it has helped compose the client’s history. Sometimes it has supplied the diagnosis, interpreted the spouse, and drafted the boundary the client intends to announce before the therapist has asked a single question.
The therapist is no longer necessarily the first interpreter of a life.
The therapist is fast becoming the second opinion.
Interpretation Has Become Cheap
This is the part of the AI conversation that therapists should take personally.
Not defensively. Personally.
For a long time, some therapy traded heavily on the value of interpretation.
The therapist noticed a pattern, named an emotion, connected a present conflict with an earlier wound, or explained that the client’s pursuit of intimacy might be producing the very withdrawal she feared.
That can be useful. It can also now be generated before breakfast.
An AI can explain anxious attachment. It can suggest a softer startup, rehearse a difficult conversation, compare the Gottman Method with emotionally focused therapy, and write a compassionate letter to your emotionally unavailable father—then make it warmer, shorter, firmer, less accusatory, and slightly more like Joan Didion.
The machine’s fluency exposes an uncomfortable fact: some of what clients once experienced as therapeutic expertise was actually information.
Information remains valuable. But it is no longer scarce.
This does not mean AI reliably knows whether its interpretation is true. General-purpose systems can hallucinate, mirror the assumptions embedded in a prompt, miss danger, and produce unwarranted certainty in a soothing voice.
The Therabot trial cannot be generalized to every public chatbot.
A tool engineered for mental-health treatment, supervised for safety, and tested in a controlled study is not interchangeable with an agreeable machine improvising from one partner’s transcript.
Still, the competitive pressure is real. If therapy consists mainly of naming patterns, offering validation, and periodically asking how that makes you feel, it now has a formidable competitor.
The competitor costs less, is available at midnight, remembers the previous conversation, and never takes August off.
The answer cannot be for therapists to imitate the machine more efficiently.
The answer is to remember what therapy was supposed to provide beyond words.
Feeling Understood Is Not the Same as Being Known
An AI can understand the account you give it.
A therapist must become curious about the account you do not give.
Human beings are unreliable narrators, especially when frightened. We omit without knowing that we are omitting. We polish motives. We organize memory around innocence. We describe the provocation in high definition and our response in soft focus.
This is not evidence of wickedness. It is evidence of being human.
A machine receives the story we supply. A skilled therapist also watches how the story is supplied.
Why did the voice flatten just then?
Why does the husband describe every feeling except anger?
Why does the wife become exquisitely articulate whenever the conversation approaches her own power?
Why is one partner permitted a childhood while the other is assigned a diagnosis?
Why does this couple agree beautifully about everything except what actually happened?
These are not simply informational questions. They arise inside an encounter.
The therapist notices not only what the couple says, but what the couple begins doing with the therapist. One partner recruits. The other retreats. One performs reasonableness. The other becomes progressively less reasonable under the pressure of being misunderstood. Within twenty minutes, the problem they came to describe is happening again in the room.
No transcript contains the whole of that experience.
The research on psychotherapy does not prove that only humans can create therapeutic change.
It does show that the working alliance—agreement on goals, agreement on tasks, and a meaningful bond—is reliably associated with outcome.
A 2018 meta-analysis covering 295 studies and more than 30,000 patients found a consistent alliance-outcome association across treatment approaches, countries, and both face-to-face and internet-based care. A later meta-analysis found that the relationship remained after adjusting for patient characteristics and other treatment processes.
The result should not be romanticized: correlation does not mean the alliance alone causes improvement.
People may rate the alliance more favorably when treatment is already helping. But the evidence makes one point difficult to dismiss. Therapy is not simply the delivery of correct psychological information. What happens between patient and therapist matters.
In couples therapy, the therapist’s task is not merely to understand the story. It is to survive being pulled into it without accepting the role the couple has prepared.
The End of Validation as a Complete Product
Validation matters.
Clients cannot reliably examine themselves while being humiliated.
A nervous system under attack does not become more reflective merely because someone has made an excellent point.
But validation was never supposed to be the finished product.
It was supposed to make truth bearable.
The current criticism of therapy is easiest to find in online testimony, essays, and argument—not in a national survey proving that American attitudes have reversed. That distinction matters.
Reddit is a cultural listening post, not a representative sample. It can reveal the language in which dissatisfaction is being organized; it cannot tell us how many people share it.
What it reveals is a recurring suspicion that therapy can become an expensive form of agreement: a weekly hour in which pain is confirmed, antagonists are diagnosed at a distance, and the client’s preferred explanation quietly acquires professional credentials.
This is sometimes an unfair caricature. It is not always an imaginary one.
There is a kind of therapy in which every painful feeling becomes evidence that someone else has caused harm.
Every demand becomes a boundary. Every enduring obligation becomes self-abandonment.
Every difficult relative becomes toxic. The client’s vocabulary expands while the client’s capacity for relationship contracts.
Psychologist Nick Haslam’s work on “concept creep” offers a grounded way to understand part of this phenomenon. Although I expressed a pointed critique that this notion has been avidly politicized.
Harm-related concepts—including trauma, abuse, addiction, bullying, and mental disorder—have expanded over time to include new kinds of experience and less severe instances.
I argue that this expansion can be humane: folks once ignored may finally receive recognition.
But broader concepts also carry risks.
Ordinary adversity can be redescribed as pathology; disagreement can acquire the moral force of danger; and language designed to identify serious harm can lose precision through indiscriminate use.
The point is not that people should surrender the words trauma, abuse, or safety. It is that serious words require serious discrimination.
Therapeutic language can illuminate a pattern. It can also result in a declared end to inquiry.
Such therapy can feel wonderful for quite a while. The client becomes fluent in explaining why change must begin elsewhere.
But insight that never reaches conduct is merely a more sophisticated alibi.
A serious therapist must be able to say two things in the same hour:
What happened to you matters.
And:
What you do next remains your responsibility.
Anything less is not compassion. It is abandonment with an excellent bedside manner.
When Therapy Became a Moral Court
Therapy has also become entangled with our culture’s hunger for adjudication.
Clients do not always arrive asking, “How do we understand this?” Increasingly, they ask, “Who is right?” More precisely: “Will you confirm that I am the injured party and that my partner is the problem?”
Social media prepared the case.
The algorithm supplied the charges. The therapist is invited to deliver the verdict.
This is particularly dangerous in couples therapy, where one person’s completely sincere account may still be radically incomplete.
The language of therapy lends itself beautifully to moral elevation.
“I was triggered” may describe an involuntary response—or end a conversation about the behavior that followed it.
“I need a boundary” may protect a person from mistreatment—or provide a civilized name for unilateral control.
“I don’t feel safe” may signal genuine danger—or the less dramatic fact that someone has heard something unwelcome.
Words do not become false because we misuse them. But therapeutic language can conceal as effectively as it reveals.
A competent couples therapist does not casually decide which partner is the designated patient and which is the reliable witness.
Nor does neutrality require pretending that cruelty, coercion, or violence is merely a communication problem.
The discipline lies in seeing asymmetry without becoming intellectually lazy; recognizing harm without turning diagnosis into theater; and refusing the intoxicating simplicity of a villain when two complicated people are still in the room.
The Strange Decline of “Go to Therapy”
Another change is gathering underneath the AI story.
Couples are becoming less impressed by the instruction to “go to therapy.”
The phrase once communicated concern.
Online, it increasingly functions as dismissal: take your sadness somewhere else; hire someone to absorb it; return when you are more convenient.
Again, this is a cultural observation, not a prevalence estimate.
But it converges with serious research on medicalization and belonging.
Heartbreak, loneliness, grief, rejection, confusion, failure, and the loss of direction are painful.
Pain is not, by itself, proof of pathology. Sometimes we need treatment. Sometimes we just need a friend, a ritual, a community, a moral vocabulary, a long walk, meaningful work, or enough time for an unwelcome truth to become survivable.
Therapy expanded partly because community contracted.
That sentence is my interpretation, not a settled causal finding.
Yet the underlying problem is measurable. Social connection is not merely a pleasant supplement to mental health.
A 2023 meta-analysis of 90 prospective cohort studies involving more than 2.2 million people found that both social isolation and loneliness were associated with higher mortality risk. Earlier meta-analyses reached similar conclusions.
We began hiring professionals to perform fragments of work once distributed among siblings, neighbors, clergy, elders, extended families, and friends. Some of those older communities were suffocating, judgmental, and spectacularly unhelpful. We should not romanticize them.
But neither should we pretend that one carefully bounded hour with a paid professional can replace belonging.
The therapist may help a lonely person understand loneliness. The therapist cannot become the village whose absence produced it.
Good therapy should eventually return people to life—not make therapy the most emotionally significant relationship they have.
Therapists Became Content, and Clients Noticed
At exactly the moment therapy most needs to defend the singular value of human trust, some therapists have decided to become influencers.
They discuss “red flags” in clients.
They perform exasperation about difficult sessions. They turn clinical encounters into anecdotes scrubbed just clean enough to be called educational. They publicly process how draining it is to hold other people’s pain.
This section also requires precision.
There is no strong evidence that most therapists behave this way, or that the public has uniformly turned against clinicians on social media.
Online psychoeducation can reduce stigma, improve access to useful information, and help people recognize when they need care.
The ethical problem is not visibility itself. It is the collision of professional authority, audience incentives, porous boundaries, and client vulnerability.
The APA Ethics Code prohibits psychologists from disclosing personally identifiable confidential information in public media and requires accuracy in public statements.
APA guidance on social media also emphasizes evidence, confidentiality, competence, and the difficulty of keeping professional and personal identities separate online. Qualitative research with therapists describes recurring dilemmas involving blurred boundaries, competence, confidentiality, self-disclosure, and the pressure to create content.
A client does not need to recognize herself in a therapist’s video to be affected by it.
She needs only to wonder whether her own grief becomes tiresome after the session ends. Whether the therapist secretly rolls his eyes at her repetition. Whether her most ashamed confession might inspire next Tuesday’s content.
Therapists are human. They become tired, irritated, frightened, bored, protective, and occasionally wrong. Ethical practice never required the absence of these reactions. It required somewhere responsible to put them.
Supervision is one such place. Consultation is another. One’s own therapy can be another.
The public square is not.
This is why I prefer writing to performance.
Writing slows the therapist down. It permits qualification, evidence, revision, and the possibility that a complicated idea may remain complicated. It also allows psychoeducation to remain directed toward readers rather than covertly away from clients.
I would rather supply ideas than manufacture intimacy with an audience.
And I endeavor to balance the power of therapeutic storytelling with never wanting a client to wonder whether becoming visible to me means becoming but mere content for everyone else.
What the Human Therapist Is For
AI may become exceptionally good at promoting psychological reflection.
That does not make human therapy obsolete. It makes the human therapist’s actual work somewhat easier to see.
The therapist is not valuable merely because they possesses therapeutic language. Everyone possesses therapeutic language now.
The therapist is valuable because he can enter a consequential relationship without allowing that relationship to become another instrument of avoidance.
A human therapist can notice what you omit without being instructed to look for omissions.
They can remember how you treated him when you felt misunderstood.
They can decline the role of rescuer, judge, accomplice, admiring audience, or obedient witness. They can be affected by you without making his feelings your burden.
And, crucially, they cannot be completely controlled.
You cannot simply regenerate the response until your therapist agrees. You cannot delete the session in which you were less generous than your self-concept permits. You cannot customize another person’s subjectivity until it reflects only what you prefer to see.
The therapist may misunderstand you. They may need correction. They may apologize. The relationship can rupture.
That is not merely an unfortunate defect in human therapy.
Under the right conditions, it is part of the treatment.
A meta-analysis of alliance rupture and repair found that successfully resolved ruptures were moderately associated with better outcomes.
This does not mean every therapeutic mistake is secretly beneficial.
Unrepaired rupture can harm treatment and contribute to dropout.
The value lies in recognizing strain, examining what occurred, and attempting a repair that neither person can achieve by simply editing the other.
Repair requires two centers of experience. Accountability requires someone who is not merely an extension of the self. Intimacy requires contact with a mind we do not author.
This is why the deepest work of therapy cannot be reduced to receiving sufficiently intelligent responses.
It involves discovering what happens when another person remains real in your presence.
AI May Make Good Therapy Better
The easiest professional response to AI is territorial: machines cannot feel, machines cannot diagnose, machines cannot replace us.
All true, and not quite sufficient.
Clients are already using AI to prepare for sessions, organize memories, learn skills, rehearse conversations, and think between appointments.
Some will misuse it.
Some will disclose intimate data without fully understanding where it goes.
Some will become dependent on its availability. Some will use its agreeable fluency to build an impregnable explanation of why everyone else is wrong.
Others will arrive better prepared and more psychologically literate than clients at any previous moment in history.
That could be good news.
Let the machine explain attachment theory. Let it generate the worksheet. Let it summarize the argument and identify four plausible interpretations. Let it help the client find the words.
Then let the human work begin.
Is the interpretation true?
What has been excluded?
What happens inside you when your partner’s experience contradicts yours?
Can you remain curious when validation is not immediately forthcoming?
Can you distinguish emotional safety from freedom from discomfort?
Can you make a promise that survives your next mood?
AI can help people think about relationships.
Therapy must help them have one.
Therapy After the Monopoly
The loss of therapy’s interpretive monopoly is not the end of the profession.
But it is the end of a certain authoritative complacency.
Therapists will no longer be able to justify their value merely by possessing language that clients lack.
They will have to demonstrate judgment, restraint, courage, clinical depth, and the ability to hold competing realities without collapsing into fashionable certainty.
Clients, meanwhile, may become less dazzled by psychological fluency and more interested in results.
Not:
Did I feel validated?
But:
Did I become more honest? More capable of intimacy? Less governed by fear? Better able to recognize my impact?
More willing to repair what can be repaired—and to leave, soberly, what cannot?
Those are harder questions.
They are also much better ones.
Therapy does not deserve cultural authority simply because it is therapy. Authority must be earned through competence, ethics, and utility.
Our community of practice should welcome this emerging, more robust standard.
AI can remember your history. It can name your patterns. It can speak gently to the most frightened parts of you.
But it cannot meet you as another human being whose reality places a claim upon your own.
That remains the dangerous, frustrating, irreplaceable gift of relationship.
And relationship—not interpretation—is the real work all along.
FAQ
Is AI therapy approaching the efficacy of working with a human therapist?
Perhaps, but not yet. One well-designed 2025 trial found meaningful symptom improvement from a purpose-built generative-AI program, but it used a waitlist control and included clinical monitoring. That does not show that general-purpose chatbots equal good psychotherapy. The evidence is promising, early, and much narrower than the public conversation suggests.
Can AI safely help me think about my mental health or relationship?
It can be useful for psychoeducation, journaling prompts, organizing questions, rehearsing conversations, and considering alternative interpretations. It should not be treated as an infallible diagnostician, emergency service, or impartial judge of a relationship. Be cautious about privacy, hallucinated facts, overconfident labels, and responses that simply reinforce the premise of your question.
Is validation bad therapy?
No. Validation reduces shame and helps people remain engaged enough to think. It becomes inadequate when it is detached from reality testing, agency, accountability, or behavioral change. Good therapy makes difficult truth more bearable; it does not make truth optional.
Has ordinary suffering become over-medicalized?
Sometimes—but the claim requires care. Expanded mental-health language has helped many people recognize genuine suffering and seek treatment. Research on concept creep also suggests that harm-related terms have broadened to include less severe and qualitatively different experiences. The clinical task is not to ban psychological language. It is to distinguish disorder from distress and mistreatment from the inevitable discomfort of human relationships.
Does saying “go to therapy” help?
It can, especially when someone needs assessment or treatment. It can also function as a socially acceptable refusal of ordinary human care. Therapy can supplement friendship, family, community, faith, meaningful work, and civic belonging. It cannot replace all of them.
Should therapists avoid social media?
Not necessarily. Responsible psychoeducation can be valuable. The danger begins when clinical authority is used for oversimplified diagnosis, personal performance, disguised case material, or contemptuous commentary about clients. A therapist’s public presence should protect confidentiality, preserve trust, remain evidence-based, and acknowledge the limits of what can be responsibly said to a mass audience.
What can a human therapist provide that AI cannot?
A human therapist offers an embodied, consequential relationship with another mind: observation beyond the client’s chosen account, resistance to being customized, ethical responsibility, real-time attention to interaction, and the possibility of rupture and repair.
AI may simulate aspects of empathy and alliance. It does not yet occupy the same moral, legal, interpersonal, or clinical position.
How should I use AI alongside therapy?
Use it as a thinking aid, not a hidden authority. Tell your therapist when AI has meaningfully shaped your interpretation, diagnosis, boundary, or decision. Bring the output into the room as material to examine: What did it notice? What did it assume? What did I leave out? Why did this answer feel so compelling?
Why Work With Daniel?
Couples rarely need another person to decide which partner has the superior explanation.
They need someone capable of recognizing injury without turning the relationship into a courtroom—and capable of demanding responsibility without using shame as a weapon.
My work is for couples who want more than validation, vocabulary, or temporary relief.
We examine the pattern occurring between you, the histories that keep it alive, and the choices each of you still possesses.
The purpose is not to make one partner feel professionally vindicated. It is to discover whether greater honesty, accountability, intimacy, and repair remain possible.
If your relationship has become trapped inside two increasingly persuasive versions of reality, we can begin there.
Schedule a free introductory call
Be Well, Stay Kind, and Godspeed.
REFERENCES:
American Psychological Association. (2017). Ethical principles of psychologists and code of conduct (2002, amended effective June 1, 2010, and January 1, 2017). https://www.apa.org/ethics/code
Eubanks, C. F., Muran, J. C., & Safran, J. D. (2018). Alliance rupture repair: A meta-analysis. Psychotherapy, 55(4), 508–519. https://doi.org/10.1037/pst0000185
Flückiger, C., Del Re, A. C., Wampold, B. E., & Horvath, A. O. (2018). The alliance in adult psychotherapy: A meta-analytic synthesis. Psychotherapy, 55(4), 316–340. https://doi.org/10.1037/pst0000172
Flückiger, C., Del Re, A. C., Wlodasch, D., Horvath, A. O., Solomonov, N., & Wampold, B. E. (2020). Assessing the alliance–outcome association adjusted for patient characteristics and treatment processes: A meta-analytic summary of direct comparisons. Journal of Counseling Psychology, 67(6), 706–711. https://doi.org/10.1037/cou0000424
Haslam, N. (2016). Concept creep: Psychology’s expanding concepts of harm and pathology. Psychological Inquiry, 27(1), 1–17. https://doi.org/10.1080/1047840X.2016.1082418
Haslam, N., & Tse, J. S. Y. (2021). Concept creep and psychiatrization. Frontiers in Sociology, 6, Article 806147. https://doi.org/10.3389/fsoc.2021.806147
Heinz, M. V., Mackin, D. M., Trudeau, B. M., Bhattacharya, S., Wang, Y., Banta, H. A., Jewett, A. D., Salzhauer, A. J., Griffin, T. Z., Abbott, J. A., Espinosa, C., & Jacobson, N. C. (2025). Randomized trial of a generative AI chatbot for mental health treatment. NEJM AI, 2(4). https://doi.org/10.1056/AIoa2400802
Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality: A meta-analytic review. Perspectives on Psychological Science, 10(2), 227–237. https://doi.org/10.1177/1745691614568352
Miller, V., Hill, M. J., & Moreira, T. (2026). Artificial intelligence and unconditional love: The rise of generative AI as an alternative form of mental health support. Communication and Change. Advance online publication. https://doi.org/10.1007/s44382-026-00036-1
Wang, F., Gao, Y., Han, Z., Yu, Y., Long, Z., Jiang, X., Wu, Y., Pei, B., Cao, Y., Ye, J., Wang, M., & Zhao, Y. (2023). A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness and mortality. Nature Human Behaviour, 7(8), 1307–1319. https://doi.org/10.1038/s41562-023-01617-6
White, E., & Hanley, T. (2024). Current ethical dilemmas experienced by therapists who use social media: A systematic review and meta-synthesis. Counselling and Psychotherapy Research, 24(2), 512–526. https://doi.org/10.1002/capr.12678