Everyone Is Diagnosing Themselves But Whose Getting Better?

I have watched a friend describe her ex as a narcissist, her boss as a gaslighter, and her own procrastination as executive dysfunction, all in the same conversation, all without ever having sat across from a licensed clinician. She is not unusual. She is, if anything, fluent in a language that has become so common it barely registers as jargon anymore. Somewhere in the last five years, the vocabulary of the therapy room walked out the door, got on a phone, and moved in with the rest of us. I do not think this happened by accident, and I do not think it happened for free.

There is a version of this story that sounds like progress, and for a while I believed it. Mental health stopped being something people whispered about and started being something people posted about. Depression, anxiety, ADHD, trauma, none of these words carried the same shame they carried when I was younger. That shift is real and it matters. But somewhere along the way, having the words for a feeling started to feel like the same thing as understanding it, and understanding it started to feel like the same thing as fixing it. Those are three different things, and the gap between them is where I think a lot of us are quietly stuck.

Also Read: Do We All Know What We Want?

Start with the numbers, because they are not subtle. A systematic review published in early 2026 pulled together twenty seven studies and more than five thousand posts and videos across social platforms, and found that mental health and neurodivergence content online carries an average misinformation rate of roughly 26 percent. On TikTok specifically, that number climbs to nearly 35 percent. A separate analysis from PlushCare, which had licensed clinicians review five hundred TikTok videos under mental health hashtags, found that 83.7 percent of the advice was misleading, and that content on bipolar disorder was inaccurate 94.1 percent of the time, with depression content close behind at 90.3 percent. ADHD content, the single most searched mental health topic on the platform, fared no better. A study in PLOS ONE examining ADHD videos specifically found that clinical psychologists rated the accuracy of the content as low across the board, and that people who had self-diagnosed with ADHD after watching this content were, in a striking number of cases, told by an actual psychologist’s explanation video that they likely did not have the condition at all.

I want to sit with that last finding for a moment, because I think it is the whole essay in miniature. People are watching content that makes them believe they have a disorder, and then watching different content, sometimes from an actual professional, that tells them they probably do not. The information environment is not just inconsistent. It is actively producing false positives at scale, and it is doing so inside a format optimized for engagement rather than accuracy, where a fifteen second video claiming that forgetting your keys means you have ADHD will always outperform a nuanced explanation of differential diagnosis. One study I read described exactly that example, the keys-forgetting claim, as a textbook case of disinformation dressed up as relatability.

None of this stopped one in four American adults from suspecting they have ADHD, according to research cited by Harvard Law’s Petrie-Flom Center, even though only around six percent of the population actually meets the clinical threshold. The same research found that people attempting to self-diagnose get it wrong somewhere between five and eleven times more often than they get it right. I do not point this out to mock anyone reaching for an explanation of why their brain feels the way it feels. I point it out because the confidence with which people now diagnose themselves has completely outpaced the accuracy of the information they are diagnosing themselves with, and almost nobody in that pipeline is incentivized to slow it down.

The language problem runs even deeper than diagnosis, though, because it is not just clinical labels that have escaped into the wild. It is the entire vocabulary that used to live inside a therapist’s office. Gaslighting. Narcissist. Trauma bonding. Love bombing. Boundaries. Triggered. Merriam-Webster made gaslighting its word of the year back in 2022, which tells you how far this had already traveled before most of us noticed it happening. Psychologists interviewed by Time described watching these words get stretched until they meant almost nothing. A clinical psychologist quoted in that piece said gaslighting is meant to describe a sustained, deliberate pattern of manipulation designed to make someone doubt their own reality, not a single disagreement about what someone said last Tuesday. But that is not how the word gets used in a group chat. It gets used the moment someone remembers an event differently than you do.

The same drift has happened with narcissist, a word one Boston based mental health counselor described as being thrown around so carelessly that it now functions mostly as a label for anyone whose personality you simply do not enjoy. This matters for more than semantic tidiness. When a word that describes a specific, often devastating pattern of psychological abuse gets applied to an ex who forgot your birthday, it does two things at once. It trivializes the experience of people who lived through actual narcissistic abuse, making it harder for them to be taken seriously when they finally do try to describe what happened to them. And it gives the person using the word a kind of unearned authority in whatever conflict they are having, because nothing ends a conversation faster than accusing the other person of a diagnosable condition.

There is a name for this now too, which feels almost too on the nose. Therapists call it weaponized therapy speak, the strategic use of clinical language not to build understanding but to shut it down. A psychotherapist who wrote a book on the subject, released this year under the blunt title Psychobabble, argues that we have started pathologizing ordinary life, using words like trauma for things that would more honestly be described as embarrassing, or upsetting, or just plain annoying. He is not wrong, and I think most people who have spent any time in a modern group chat already know he is not wrong, even if they would not put it in those words. We have built a vocabulary sophisticated enough to sound like insight while doing none of the work insight actually requires.

Also Read: A New Cold War Is Coming

Here is the part that I keep circling back to, the part that gives this essay its title. None of this would bother me nearly as much if the outcomes were improving. If self-diagnosis culture and therapy speak were genuinely making people psychologically healthier, I would call it a reasonable trade, messy vocabulary in exchange for real relief. But the data does not support that trade. Diagnosed anxiety among adolescents rose 61 percent in a recent multi year period, moving from 10 percent to over 16 percent, according to Health Resources and Services Administration figures, while diagnosed depression rose 45 percent over the same window. Roughly 40 percent of high schoolers report persistent feelings of sadness or hopelessness. Teen anxiety disorders have tripled over the past two decades. And despite all of this awareness, all of this language, all of this content, a majority of teens with depression, by some estimates around 60 percent, still receive no professional treatment at all.

Awareness went up. Suffering also went up. If naming the problem were the same thing as solving it, these two lines on the graph should have started moving in opposite directions by now, and they have not. I do not think that is a coincidence, and I do not think it is simply because access to real care remains limited, though it does. I think a meaningful part of what is happening is that the act of naming a feeling has started to substitute for the much slower, much less shareable work of actually sitting with it, understanding where it came from, and changing the conditions that produce it.

This is not a new idea, exactly. Every generation finds a language for its own distress, and every generation eventually discovers that the language is not the cure. What is new is the speed and the audience. A teenager in 2010 who suspected something was wrong with them might have mentioned it to a friend, or looked it up quietly in a library book, and either way the process stayed mostly private, mostly slow, and mostly free of an algorithm deciding what to show them next. A teenager today can post a single video describing a vague feeling of restlessness and receive, within minutes, a wave of comments confidently informing them that they have ADHD, or are a highly sensitive person, or are experiencing complex trauma from an emotionally immature parent. The diagnosis arrives faster than the reflection that should have preceded it, and it arrives with the validating warmth of a community rather than the sober, sometimes uncomfortable process of a clinical evaluation.

I think this appeals to something very real in people, and I do not want to pretend the appeal is shallow. A diagnosis, even a self-applied one, offers something genuinely valuable: an explanation that is not your fault. If you are anxious because you have generalized anxiety disorder, that is a condition, something that happened to you, not a character flaw. If your relationship failed because your partner was a narcissist, that is their pathology, not a referendum on your judgment. Researchers studying the online self-diagnosis phenomenon have described exactly this dynamic, noting that for many people a diagnostic label represents validation, identity, and belonging, not attention seeking or naivety. I believe that. I think most people reaching for these words are reaching for relief from a very old and very human fear, the fear that whatever is wrong with them is simply who they are, with no explanation and no path out.

But relief and resolution are not the same thing, and I worry that the culture around self-diagnosis has quietly convinced a generation that they are. Naming a pattern can be the first step toward healing it, genuinely, if it leads somewhere. If it leads to a conversation with an actual clinician, to a period of honest self-examination, to a change in behavior that gets tested against reality over months rather than confirmed by a comment section within the hour, then yes, the label did its job. But if the label is the destination rather than the starting point, if identifying as anxious, or traumatized, or a highly sensitive person becomes the entire intervention, then nothing about the underlying condition has actually been addressed. You have simply given your suffering a shareable name.

I think about this every time I see the phrase self-care used to justify avoiding a hard conversation, or the phrase setting boundaries used to explain cutting off a friend after a single disagreement, or the phrase trauma response used to describe getting annoyed. The Cleveland Clinic put together a pair of sample exchanges that captured this almost too well, showing how a simple missed errand between partners can escalate into an accusation of gaslighting within two lines of dialogue. That is not a couple working through conflict. That is a couple using clinical vocabulary to avoid working through conflict, because clinical vocabulary sounds like it has already done the work of figuring out who is right.

Also Read: AI Will Create More Billionaires Than Any Technology Before It

There is also a quieter, more structural problem underneath all of this, and it is one I think about often given how much of my own work involves following where incentives actually point rather than where people say they point. Platforms do not make money when someone spends six months in therapy slowly untangling why they struggle with intimacy. Platforms make money when someone posts a fifteen second video explaining why they struggle with intimacy, gets validated in the comments, and comes back tomorrow for more. The entire economic logic of the attention economy rewards the naming of a problem far more reliably than it rewards the solving of one, because naming is fast, shareable, and endlessly repeatable, while solving is slow, private, and usually boring to watch. I do not think the platforms set out to build a machine that mistakes insight for content, but I think that is functionally what they built, and I think most of us are living inside it without fully realizing the machine has its own preferences about what kind of self-understanding gets amplified.

I also think the format itself is doing more damage than we tend to credit it for. A therapist working with a real patient has time. They can sit with a vague, contradictory description of a feeling for weeks before offering anything close to a label, and even then the label usually comes wrapped in qualifications, in a plan for follow up, in an acknowledgment that the picture might change as more sessions unfold. None of that survives the transition to a fifteen second video. The format demands a thesis in the first three seconds and a payoff by the end, which means nuance is not just unwelcome, it is structurally impossible. A creator cannot say something might be a sign of ADHD, or could be one of several overlapping possibilities, or really depends on context we do not have, and expect the algorithm to reward that honesty with reach. The incentive is to say you have this, full stop, because certainty performs better than accuracy, and the platform has no mechanism for telling the difference between the two.

This is part of why the autism and ADHD content specifically has become such a well documented mess. A content analysis of TikTok videos tagged with autism found that 41 percent presented inaccurate information and another 32 percent presented information so overgeneralized it was functionally misleading, and those videos alone had accumulated more than 144 million views. The behaviors being described as symptomatic in these videos were often things like disliking loud noises or preferring a messy bedroom, traits that exist on a spectrum in the general population and are not, on their own, diagnostic of anything. But framed as a checklist in a confident voiceover, they read like a personality quiz with clinical stakes, and personality quizzes with clinical stakes are exactly the kind of content that travels fastest.

What strikes me most, though, is not that misinformation exists. Misinformation has always existed, on every subject, on every platform, since long before TikTok. What strikes me is how well suited mental health topics specifically are to being misunderstood in this particular way, because mental health symptoms are, almost by design, things everyone experiences to some degree. Everyone has forgotten their keys. Everyone has felt overwhelmed in a social situation. Everyone has had a period where getting out of bed felt harder than it should. A list of symptoms for almost any psychiatric condition will describe universal human experiences at some intensity, which is exactly why differential diagnosis, the actual clinical skill being replaced here, exists in the first place. It is not enough to feel restless. The question a real diagnostic process asks is how restless, for how long, compared to what baseline, causing what specific impairment, and ruling out what other explanations first. That process is boring to watch and impossible to compress into a viral format, so it simply does not get made, and its absence leaves a vacuum that confident, symptom-checklist content fills instead.

None of this is an argument against therapy language existing in public at all. I do not want to go back to a world where anxiety was a secret and depression was a character defect and asking for help meant something was fundamentally broken in you. That world was worse, not better, and the destigmatization that got us here is one of the genuinely good things to happen culturally in the last two decades. The problem is not that we started talking about mental health. The problem is that talking about it started to feel like enough. A generation raised on trigger warnings and diagnostic vocabulary has become extraordinarily good at identifying what is wrong, and comparatively untrained in the much harder, much less glamorous work of sitting with discomfort long enough to actually move through it.

I do not have a tidy solution to offer here, and I am suspicious of anyone who claims they do, because a tidy solution to this problem would itself be exactly the kind of shareable, name-it-and-move-on answer I am arguing against. What I have instead is a smaller, less satisfying suggestion, which is that the next time you find the right word for what you are feeling, treat it as a beginning rather than a conclusion. Ask what comes after the label. Ask whether the label is leading you toward a harder conversation or away from one. Ask whether you are using the word because it is accurate, or because it is available, and because everyone around you already speaks the same language fluently enough that reaching for it costs nothing.

We built ourselves an extraordinary vocabulary for pain in a remarkably short amount of time. What we have not built, not nearly as quickly, is the patience required to use that vocabulary well. Until we do, I suspect the numbers will keep telling the same uncomfortable story they are telling right now: more people than ever can name exactly what is wrong with them, and fewer of them than we would like are actually getting better.

Also Read: How Anthropic Quietly Proved Its Own AI Would Kill to Avoid Being Shut Down