Second-generation impostor syndrome

by Miguel Lucas

You review the document before sending it. It’s impeccable. Clear structure, solid argument, precise tone. And then the question you don’t dare say out loud appears: how much of this is mine and how much is the AI’s? Welcome to second-generation impostor syndrome.

An American Psychological Association study of 1,923 professionals in North America 1 quantifies it without ambiguity: 58% acknowledged that AI “did most of the thinking” in planning and reasoning tasks. Even more revealing: the data show a significant negative correlation between prompt dependence and confidence in independent reasoning. The more you delegate, the less you trust yourself.

Performance data confirm the paradox. A controlled Harvard study with 758 BCG consultants 2 showed that AI improved output quality by 40% and speed by 25% on tasks within the model’s capabilities. But when consultants applied AI outside that frontier, performance dropped dramatically — producing worse results than the group using no technology at all. The tool works. What erodes is not the output but the perception of authorship: the professional stops seeing themselves as the architect of the solution and begins to feel like an operator of machinery they don’t fully understand.

This is not the classic impostor syndrome. That one was rooted in subjective distortion: the competent professional who felt like a fraud. This new variant feeds on an objective technical reality: the delegation of logic and creativity to artificial intelligence systems. And speed amplifies everything — enterprise adoption of generative AI jumped from 55% to 78% in a single year 3. There has been no time to build psychological defenses or evaluation frameworks.

And yet, one finding inverts the equation. Participants who maintained active oversight — modifying or rejecting AI suggestions — reported significantly higher confidence levels and a stronger sense of authorship 1. The difference is not in whether you use the tool, but in how you relate to it.

Professional value no longer lies in what you produce — that gets commoditized — but in the judgment you apply to what has been produced. In the ability to say “not this” when the machine says “yes this.” The insecurity that millions of professionals feel today is not a character flaw: it is the signal that the center of gravity of intellectual work has shifted. And it is in judgment — not production speed — where genuine differential value resides. The one thing you should never delegate.

Related theses

References

  1. American Psychological Association — Overreliance on AI programs may undermine confidence at work
  2. BCG — How People Can Create—and Destroy—Value with Generative AI
  3. Stanford HAI — The 2025 AI Index Report