What is the new imposter syndrome?

You finished something good this week and you cannot say which part of it was yours. The argument arrived whole, the room accepted it, your name is on it, and when you go looking for the moment you first thought it you find a conversation with a model in March, a podcast on a walk, a line you cut from your own draft, and no seam between them. That doubt has a shape of its own, and answering it with evidence of your competence misses what it is asking.

Clance and Imes described a different condition#

The phrase everyone reaches for comes from a 1978 article by Pauline Rose Clance and Suzanne Ament Imes in Psychotherapy: Theory, Research & Practice. Their subject was an internal experience of intellectual phoniness in women whose records plainly contradicted it: the professor who believed her appointment was an administrative error, the woman with two master's degrees, a doctorate and a list of publications who thought herself unqualified to teach a remedial class. Graded entry.

Read at source, the paper corrects something that is repeated constantly about it. The popular account says the old syndrome could be cured with evidence, that you gathered the testimony and the outcomes and confronted the distortion with what was true. Clance and Imes say the opposite in their own abstract: achievements "do not appear to affect the impostor belief". In the body they are blunter still, reporting that they "have not found repeated successes alone sufficient to break the cycle". The therapy they propose is group disclosure, Gestalt work and behavioural experiments, with evidence-gathering appearing only as one homework task whose point is noticing the refusal to accept a compliment.

So the difference between the old condition and the new one has nothing to do with whether evidence fixes it. It has to do with what the doubt is about. Clance and Imes describe a person who has done the work and cannot believe she did. The condition on this page describes a person who cannot establish what she did. The first is a misreading of a record. The second is a record with a hole in it, and the hole is real.

One more thing belongs on the record about the 1978 paper, because it is cited far more often than it is opened. It reports no study. There is no control group, no instrument, no sampling frame and no measurement: it is five years of the authors' own caseload at two universities, about 150 women, a third of them already in therapy for something else. It is the authoritative source for what the term means and a source for nothing at all about how common it is.

The phrase was in use before this research reached it#

Rahim Hirji wrote The New Imposter Syndrome in Box of Amazing on 17 May 2026. John Nosta had published "AI and the New Impostor Syndrome" in Psychology Today on 13 March 2025, fourteen months earlier, under the subtitle "The uneasy feeling of outsourcing your own intelligence". Others have used the wording too, and a phrase of this construction is likely to have been reached independently many times.

No claim of first use is made here, and the estate would not make one on a phrase this available. What Nosta describes is also a genuinely different condition from the one defined above, which is the part a reader is best served by. His subject is effort: intelligence has moved from something you do to something you get, success arrives without the struggle that used to justify it, and the doubt is about deserving an achievement that came too easily. That is recognisably the 1978 question about desert, transplanted. The condition on this page leaves desert alone and asks a prior question, about who produced the thing at all.

What does belong to Hirji, and carries a dated first publication, is the separation of the experience into three forms. Everything in the next section is his, from 17 May 2026.

Three experiences that the one phrase runs together#

Phantom Authorship. The output is real, colleagues credit you, the work is genuinely good, and in private you cannot find yourself in it. Hirji's image for it is the phantom limb: the authorship is present without being there. He puts this as the form senior professionals meet most often, where the work continues at standard and the doubt is invisible to everybody except the person holding it.

Velocity Vertigo. The tool is moving you faster than judgement can be built to match, so you are producing at a level you have not reached. This is the felt, first-person version of what this research calls synthetic seniority, the one form of the three with a real body of evidence under it. Sankaranarayanan's 2026 experiment gives it a number: unrestricted AI users matched a scaffolded group on the work itself and then failed a later maintenance task, with the tool withdrawn, 77 per cent of the time against 39. He calls them fragile experts. Graded entry. Bastani and colleagues found the same shape in a school: students with unrestricted access scored 17 per cent below students who never had it once the tool was removed. Graded entry.

The Hollowing. Doubt at the level of identity instead of the project, in people whose sense of self was their craft. The nearest thing to a field observation of it is Ehsan and colleagues at CHI, who spent twelve months inside a hospital group adopting an AI treatment-planning system and recorded clinicians describing themselves as babysitters and bystanders in their own practice. They name it identity commoditisation. Graded entry. That study is qualitative and measures no skill. The evidence base holds no other account of its kind.

The one thing anybody has measured#

Draxler, Werner, Lehmann, Hoppe, Schmidt, Buschek and Welsch ran two experiments, 126 participants in total, published in ACM Transactions on Computer-Human Interaction in February 2024. Participants used a model to produce a personalised text, then published it and said who should be credited. They did not regard themselves as the authors or the owners of what the model had written. They mostly did not name the model either: about one in five mentioned AI in the free-text credit field in the first study, and fewer than one in five declared it as an author in the second. Sense of ownership rose with how much of the text a participant had shaped. Graded entry.

The reasons participants gave make the result sharper. The same people who withheld the credit said, when asked, that AI contributions ought to be credited for transparency and ethics. The gap between private authorship and public credit opened without anybody setting out to conceal anything.

That is a measurement of the attribution gap. It is not a measurement of the thing this page is named for, and the distance matters. Draxler's participants watched a machine produce a text and then decided what to say about it; the uncertainty is about disclosure, and they know perfectly well who wrote it. The condition described here is what happens weeks later, to an argument that formed across a dozen exchanges, when the knowing itself has gone.

Nobody has tested whether people really cannot reconstruct where a thought came from#

That sentence is the honest limit of this page. The claim at the centre of it, that fast private collaboration with a model leaves a person unable to say which part of an argument was theirs, has not been put to a test. The psychology of remembering where you learned something is an old field and it was built before any of this, on laboratory tasks with defined sources; nothing in it was designed for a collaborator that writes in your own register and leaves no artefact.

Two further limits belong to this page. The three forms are a clinician's kind of taxonomy, proposed from observation and conversation, and none of the three has been shown to be a distinct thing instead of one experience described from three angles. And the closest measured evidence, Jakesch and colleagues on 1,506 people whose own opinions moved after writing alongside an opinionated model, establishes that influence is real and invisible to the person influenced. It does not establish that they then lose track of the authorship, which is a different thing again.

So the state of it is this. The mechanism is plausible, one adjacent piece of it is measured, and the central claim is an argument from experience. Treating it as an established finding would be the mistake this estate spends most of its time catching in other people.

If you cannot find yourself in your own work#

Keep something that survives the loop. The reason a written position before the first prompt keeps coming up in this research is not discipline for its own sake; it is that a recorded prior view is the only artefact that lets you answer the question afterwards. Human at the start sets out why.

Then tell the three forms apart before acting on any of them, because they want opposite things. Phantom Authorship is uncomfortable and mostly harmless: the work is good, the capability is there, and the cure is a record rather than a change of practice. Velocity Vertigo is the one to take seriously, because the gap it describes between output and judgement is real, measured and closes only through work done unaided. The Hollowing is a question about what the job is for, and no working practice answers it.

If you manage people, the useful move is to stop reading the second form as the first. A junior who says they feel like a fraud may be describing a feeling with nothing behind it, or reporting accurately that they are producing above their level. Reassurance is the right answer to one and the wrong answer to the other, and the way to tell them apart is to look at what the person can still do with the tool switched off. That is the test behind a capability audit, and it works on one person as well as on a team.

Key sources

• Clance, P. R. and Imes, S. A. (1978). The imposter phenomenon in high achieving women: Dynamics and therapeutic intervention. Psychotherapy: Theory, Research & Practice, 15(3), 241-247. Graded entry.
• Draxler, F., Werner, A., Lehmann, F., Hoppe, M., Schmidt, A., Buschek, D. and Welsch, R. (2024). The AI Ghostwriter Effect: When Users do not Perceive Ownership of AI-Generated Text but Self-Declare as Authors. ACM Transactions on Computer-Human Interaction, 31(2), 1-40. DOI 10.1145/3637875. Graded entry.
• Sankaranarayanan, S. (2026). Mitigating "Epistemic Debt" in Generative AI-Scaffolded Novice Programming using Metacognitive Scripts. arXiv 2602.20206. Graded entry.
• Ehsan, U., Passi, S., Saha, K., McNutt, T., Riedl, M. O. and Alcorn, S. (2026). From Future of Work to Future of Workers. CHI '26. Graded entry.
• Hirji, R. (2026). The New Imposter Syndrome. Box of Amazing, 17 May 2026. The source of the three forms.

Related SuperSkills research#

On producing above your level, synthetic seniority and the missed reps. On whether the idea was yours, is it still my idea if AI helped me write it and keeping your own voice. On the gap a self-report cannot see, the illusion of competence. On forming a view before the first prompt, human at the start. On testing what you can still do unaided, am I becoming dependent on AI.

About this research#

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. The phrase "the new imposter syndrome" is not claimed by this research and was in published use before it was used here. The impostor phenomenon belongs to Clance and Imes, 1978. The separation into Phantom Authorship, Velocity Vertigo and the Hollowing is Hirji's, first published 17 May 2026. The 1978 article and the 2024 experiment were both read in full at source, and the interpretation is kept separate from the findings.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.