Five Ways To Lose Yourself With AI (and How to Stay Grounded)

Our research involves tracking how AI reshapes the person using it—how they think, how they decide, and how they hold their own value. We are focused on changes that go beyond the work they do and the outputs they deliver. The psychological changes we are interested in rarely appear in a single moment. They creep in through small, everyday acts with the tools, and by the time they’re visible, they’ve often become routine.

Across hundreds of stories, five recurring patterns stand out. They are less about prompt technique than about a person’s orientation to the tool—how permeable their thinking becomes, how tightly their identity couples with the system, and how fluidly they let meaning, value, and standards move.

Voice loss often begins subtly. Familiar turns of phrase or stylistic choices start coming from the system rather than the individual. Over time, older work reads as thought-in-progress, while newer work feels polished but thinner. One writer, looking back at her past reports, noticed she could immediately tell which sections carried her own tone and which had been shaped by the model.

Erosion of confidence can masquerade as thoroughness. Decisions and judgments are rerouted through the model for confirmation. What once reinforced professional assurance now depends on machine alignment, shifting confidence from internal feedback to external validation. A planner described how she now runs scenarios she’s already decided on through the system “just to be sure,” even when nothing new emerges.

Role confusion appears when authorship becomes difficult to trace. People struggle to identify which parts of a hybrid product are their own, or describe their work in terms of “we” — themselves and the system. The symbolic line between human contribution and machine output loosens, leaving value harder to locate. One product manager admitted he could no longer answer the question of which paragraphs in a strategy document were his.

Avoidance takes the form of opting out entirely. Some professionals hold back from using AI to preserve their own style or wait for the noise to quiet. Their expertise remains intact, but it is never tested against the conditions where AI is already embedded. A curator says she deliberately kept her entire workflow AI-free, not because she doubted the tools, but because she didn’t want their influence creeping into her voice.

Symbolic overload shows up in restless experimentation. Rapid cycling between tools, keeping multiple competing outputs, or producing streams of inconsistent drafts reflects the pull of novelty without a stabilizing frame for coherence. A researcher confessed she had four versions of the same paper from different systems and no clear sense of how to reconcile them.

In the stories where adaptation remains healthy, the tone is different. AI is used as input, but judgment stays with the person. A policy analyst runs scenarios through the system, then pauses before drawing her own conclusions. A research lead considers alternative framings but decides herself what the real problem is. A journalist takes a machine outline and reworks it until the cadence is unmistakably her own. The frame holds steady: permeability is active but bounded, and identity stays coupled to the work rather than the tool.

AI changes you in ways that can be seen and named. Here are the key tips from our research on how to stay consciously grounded when using AI:

Maintain clear boundaries

• Step back before making final judgments, even when AI provides input
• Regularly check if you can still separate your thinking from the AI's output
• Set specific roles for yourself vs the tool in any process

Preserve your decision-making authority

• Don't make AI agreement a requirement for your own expertise
• Trust your judgment first, then use AI as one input among others
• Avoid running every decision through AI before committing

Stay aware of gradual changes

• Regularly review your recent work to spot shifts in your voice or thinking
• Notice when you start referring to "we" (you + AI) instead of "I"
• Check if you can identify your specific contributions in AI-assisted work

Choose your level of engagement deliberately

• Decide consciously how much AI input you want rather than drifting into patterns
• Avoid both total avoidance and uncritical adoption
• Pick specific use cases instead of experimenting with every new tool

Anchor your identity to your work, not your tools

• Focus on the problem you're solving, not the AI system you're using
• Rebuild AI-generated content until it reflects your actual thinking
• Make the final product genuinely yours, regardless of what helped create it

Our research suggests people who adapt well stay actively aware of how AI is changing them and make conscious choices about those changes rather than letting them happen automatically.