Therapy Speak: Why everyone calls their ex a narcissist now.

A friend texts you. The text says her ex is a narcissist. Three days later a coworker tells you his mother has BPD. By the weekend a podcast has explained that the way your boss assigned the new project was, technically, gaslighting. Everyone in the conversation is using the words confidently. Almost none of them have read the DSM-5-TR.

This is not new. People have always reached for the closest big word to describe a small problem. What is new is which words. A generation ago the closest big word was probably jerk, or crazy, or some unprintable variant of either. Today the closest big word is narcissist, or borderline, or gaslighter, or trauma response, or attachment wound. Words built inside clinical settings, by clinicians, to describe specific patterns of suffering, are now ambient internet vocabulary. They get thrown around like nickels pretending to be manhole covers — small, light, easy to flick at someone, dressed up as if they were heavy and load-bearing.¹

I see this every week in my consulting room. Someone walks in convinced their entire family system meets criteria for Cluster B personality disorders because TikTok said so. Someone else has decided their partner is gaslighting them because the partner forgot a conversation. A third person is certain they have ADHD, complex PTSD, autism, and BPD all at once because a checklist video matched, in their words, "every single one." None of these people are stupid. Most of them are unusually self-aware. They have just absorbed, without noticing, a vocabulary system whose precision has been hollowed out by repetition. This essay is about how that happened, what it costs, and what to do instead.

— Section 01What "therapy speak" actually is

The phrase therapy speak describes the migration of clinical vocabulary into ordinary conversation, and the changes the words undergo on the trip. The most useful academic frame for this comes from Nick Haslam, an Australian psychologist who in 2016 named the phenomenon concept creep.² Haslam was looking specifically at psychology vocabulary and noticed two distinct kinds of expansion happening at once.

The horizontal version is when trauma, originally reserved for events that overwhelm coping capacity in lasting ways — combat, sexual assault, severe accidents — extends to cover hearing a rude comment at brunch. The vertical version is when abuse, originally reserved for sustained patterns of physical, sexual, or coercive harm, contracts to cover a partner being short on the phone. Both directions sound, from inside the moment, like progress. We are taking emotional life seriously. We are validating people who were dismissed before. The trouble is that a word that means everything means nothing, and the people who actually meet the original threshold lose the only word that used to point at them.

Concept creep is not a moral failing. It is a structural property of language under emotional pressure, and it tends to accelerate when three things converge: a vocabulary that grants moral authority to the speaker, a medium that rewards quick emotional pattern-matching, and a culture stressed enough that more people want a name for their pain. We have all three right now.³

— Section 02The vocabulary in question

Five words have been doing most of the heavy lifting in the migration. Each one started as a precise clinical signifier. Each one has now been stretched — sometimes thinned, sometimes weaponized — into something that does not survive contact with its own definition.

Narcissist. Originally a personality-disorder diagnosis with strict DSM-5-TR criteria. Now: any selfish ex, any boss who took credit, any parent with rules.

Borderline. Originally a personality-disorder diagnosis with nine criteria of which five must be present and producing impairment. Now: any woman who is angry on the internet.

Gaslighting. Originally a sustained pattern of psychological manipulation in which one person tries to make another doubt their own perception of reality, named after the 1944 Ingrid Bergman film of the same name.⁴ Now: forgetting a thing, disagreeing with a thing, or simply not capitulating to someone's recall.

Trauma. Originally an event so overwhelming that it disrupts normal coping and lays down lasting changes in arousal, memory, and meaning-making.⁵ Now: anything mildly bad.

Triggered. Originally a clinical term in PTSD literature for a sensory or contextual cue that activates a trauma memory and the autonomic cascade attached to it. Now: irritated, offended, or simply contradicted.

Two of these — narcissist and borderline — have been so badly degraded by the migration that they deserve their own autopsies. They are also the two that show up most often in my office as misapplied weapons. Worth slowing down on.

— Section 03The autopsy of "narcissist"

Start with what the word actually points at. Narcissistic Personality Disorder, DSM-5-TR code 301.81, is one of ten personality disorders in the current diagnostic manual. It describes a pervasive pattern of grandiosity, need for admiration, and lack of empathy that begins by early adulthood and shows up across contexts. Five of nine criteria must be met, and the pattern must produce significant distress or impairment.⁶

How common is this clinically? The largest U.S. epidemiological study, the National Epidemiologic Survey on Alcohol and Related Conditions, estimated lifetime prevalence of NPD at roughly 6.2 percent, with men diagnosed at higher rates than women.⁷ Other prevalence estimates run lower, in the 0.5 percent to 1 percent range, depending on whether community or clinical samples are used and which version of the criteria is applied.⁸ Take the upper bound and you still get roughly one in sixteen adults. Take the lower and you get one in two hundred. Either way, the word is supposed to point at a meaningful minority, not a majority of the people you have ever dated.

Now drop in TikTok. The hashtag #NarcTok had crossed 5.4 billion cumulative views by early 2026, with creator content overwhelmingly framing narcissism as something to spot in an ex, parent, or boss rather than as a clinical condition requiring assessment.⁹ A 2025 study published in JMIR Infodemiology coded high-engagement TikTok videos tagged with mental health diagnoses and found that the majority contained clinically inaccurate or misleading information, with personality disorders especially affected.¹⁰ Investigative reporting in The Guardian the same year showed that "narcissist" was the single most common diagnostic label applied by short-form mental health creators to people not in the room.¹¹

Then layer in the public-figure problem. Every time a politician, executive, or celebrity does something cruel in the news cycle, the word narcissist shows up in the next viral thread. As a clinician I cannot diagnose someone I have not personally evaluated — that is the Goldwater Rule, codified by the American Psychiatric Association in 1973 and reaffirmed in 2017.¹² What I can do is name the observable behavior. This person dismissed their critics with ad hominem attacks. This person took credit for someone else's work and retaliated when called out. Those statements describe behavior. They are not a diagnosis. The distinction matters because diagnoses describe people — patterns of self that endure across contexts — while behavioral descriptions describe acts that anyone is capable of and may or may not repeat.

The word "narcissist" is not load-bearing in most of the conversations where it gets used. It is performing the work of "I am angry at this person and I want my anger to sound clinical." That is a different sentence. It deserves to be said in its own words.

Why does this matter outside of pedantry? Three reasons. First, when the word means everything, it stops meaning anything, and clinicians lose vocabulary that used to be useful for describing a real and treatable pattern. Second, the people who are actually navigating relationships with someone who meets criteria — people for whom the diagnosis describes a real and persistent pattern of harm — find their experience drowned out by everyone whose ex was just a bad partner. Third, the misuse encourages a kind of finality of judgment that closes the door on any future conversation. Narcissist sounds like a verdict. It is not. It is a hypothesis a clinician confirms over time using structured assessment.¹³

— Section 04The autopsy of "borderline"

The clinical term Borderline Personality Disorder, DSM-5-TR code 301.83, describes a pervasive pattern of instability in interpersonal relationships, self-image, and emotion, with marked impulsivity, beginning by early adulthood. Diagnosis requires five of nine criteria.¹⁴

Lifetime prevalence in U.S. adults is approximately 5.9 percent, with no significant gender difference in community samples — a finding that surprises most people because the clinical stereotype is overwhelmingly female.¹⁵ The clinical stereotype is wrong. Men diagnosed with BPD are more likely to be funneled into substance-use treatment or jail; women are more likely to be referred to mental health services and labeled. Same disorder, different system response, different cultural visibility.¹⁶

The internet has run with the female-coded version. The word borderline appears most often online as a near-synonym for angry woman, clingy woman, too-much woman, or woman who texted me twice. The misogyny is loud and old. Hannah Arendt didn't have BPD. Britney Spears didn't have BPD. Most of the women who have been called borderline in tweets and group chats over the last five years do not meet criteria for BPD. They are people experiencing entirely human reactions to entirely human situations, who have had a clinical word weaponized against them as a way of dismissing the reaction without examining the situation.¹⁷

— Section 05What the studies actually show

This is where the data gets hard to dismiss. The 2025 JMIR Infodemiology study analyzed the top one hundred trending TikTok videos under #mentalhealth, #adhd, #bpd, #autism, #ocd, and related tags. Independent clinical raters coded the videos against established diagnostic criteria. The majority — somewhere between 52 and 80 percent depending on the diagnosis — were rated as containing clinically inaccurate, misleading, or unsupported claims. Personality disorders and ADHD were the worst offenders.¹⁸

That study followed a 2022 finding from researchers at the University of British Columbia who analyzed mental health misinformation on TikTok and found similar rates of inaccuracy, plus a powerful self-recognition effect: viewers who watched diagnosis-themed content were significantly more likely to self-identify with the condition afterward, regardless of whether they met any clinical criteria.¹⁹ The recognition was not driven by accuracy. It was driven by the universality of the symptoms listed — symptoms vague enough that almost anyone could match a few. This is the Barnum effect, named after the showman P.T. Barnum, applied to diagnostic vocabulary.²⁰

Then there is the dictionary itself. Merriam-Webster named gaslighting the 2022 Word of the Year, citing a 1740 percent increase in lookups that year. The dictionary's editors specifically noted that the word's usage had drifted significantly from its original meaning and that the spike correlated with political and relationship discourse online, not with any new clinical literature.²¹ When Merriam-Webster has to announce that a word's usage has decoupled from its meaning, the migration is no longer subtle.

— Section 06Why right now

Concept creep is not new. Therapy speak as a phenomenon predates TikTok by decades — the first journalistic coverage of "psychobabble" appeared in the late 1970s.²² But several recent forces have compressed the migration into a much faster cycle.

Pandemic isolation drove a 25 percent global increase in anxiety and depression in the first year of COVID-19, per WHO data. More people went looking for language to describe what they were experiencing, and the language they found was overwhelmingly online.²³

Algorithmic short-form video rewards the exact features that degrade clinical specificity: emotional resonance, fast pattern-matching, universal applicability, and a clear villain. A nuanced video about how five of nine criteria must be met across multiple contexts and produce significant impairment performs poorly. A video titled 10 SIGNS YOUR EX WAS A NARCISSIST performs well. The algorithm picks. The vocabulary follows the algorithm.

The mental health professional shortage means that millions of Americans seeking care cannot access it. HRSA designated more than 169 million people as living in mental health professional shortage areas as of 2024.²⁴ When the consulting room is unavailable, people use the next-closest authoritative-sounding source, which is whoever has a microphone and a ring light.

Public-figure incidents — high-profile divorces, political crises, celebrity meltdowns — give the discourse weekly grist. Each cycle pulls a clinical term into public commentary, and each time the term comes back to the consulting room a little blunter than it was. By 2026 the word narcissist has been applied at one time or another to almost every major American political figure, almost every breakup featured in the tabloid press, and large fractions of the relationship advice column genre. The word is exhausted. It still has clinical work to do, and it is no longer up to the task.

— Section 07What this costs

Five distinct costs, in roughly ascending order of severity.

Diagnostic dilution. When everyone has trauma, no one does. When everyone is gaslit, no one is. The clinical specificity that made the word useful gets lost, which makes the work of identifying the small number of cases that genuinely meet criteria much harder. This is a direct cost to clinicians and to patients.

Confirmation bias in conflict. The diagnostic frame closes the loop on any conflict before the conflict has been examined. Once you have decided your partner is a narcissist, every subsequent piece of evidence reinforces the diagnosis. Disconfirming evidence — kindness, repair, accountability — is reinterpreted as manipulation, which is also a narcissistic trait, which confirms the diagnosis. The label becomes a closed system.

Self-misdiagnosis. A 2024 systematic review found that adolescents and young adults exposed to mental health content on social media were significantly more likely to self-identify with conditions they did not meet criteria for, with measurable downstream effects on health-seeking behavior, identity formation, and parental help-seeking.²⁵ The cost here is real and clinical, not theoretical.

Erasure of the people who actually have the condition. When BPD becomes shorthand for any difficult woman, the women with actual BPD — for whom the diagnosis points at a real and treatable pattern of suffering — find themselves in conversations where their condition is the punchline. People diagnosed with NPD, who already face severe stigma in clinical settings, find that the public conversation has decided they are villains rather than people with a treatable disorder.²⁶

Erosion of repair. The deepest cost. Therapy speak, when used as a weapon in a relationship, replaces the slow work of describing what hurt with the fast work of pronouncing a verdict. Verdicts foreclose. They do not invite the other person into a conversation. They tell them what they are. Most relationships do not survive being told what they are. Most of them might survive being told what hurt.

— Section 08What to do instead

This is not a plea for people to stop caring about mental health vocabulary. The opposite. It is a plea to use it more carefully so that it remains worth caring about.

Describe the behavior, not the diagnosis. Instead of he is gaslighting me, try he insisted I was misremembering things I had documented. Instead of she is a narcissist, try she dismisses what I want and centers herself in conflict. The behavioral description holds up to challenge. The label invites a fight about whether the label is correct, which is the wrong fight.

Reserve diagnostic terms for diagnostic contexts. If a clinician who knows the person has rendered a diagnosis, the term is appropriate. If you are using the term to describe someone the clinician has never met, you are making an inference that even the clinician would not make. The Goldwater Rule applies to clinicians for a reason. The reason applies to non-clinicians too.

Be suspicious of any piece of content that makes you feel diagnosed. If a checklist makes you go oh my god that's me, ask whether the checklist is specific enough to point at one condition or generic enough to point at being a person. Most online checklists are the second.

If you suspect a real pattern, get assessed. Not by a hashtag. By a clinician. Personality disorders, ADHD, autism, complex PTSD — all of these are diagnosable in trained clinical settings, and all of them deserve the rigor of structured assessment. A real diagnosis is the beginning of treatment, not a label you wear on the internet.

— Section 09Frequently asked