Why Do Designers Turn Off Generative AI?

Why does a designer who actively uses generative AI suddenly decide to stop using it during part of the creative process?

Zhang Yuzhuo and colleagues do not explain this behavior simply as resistance to technology. They focus instead on AI disengagement intention: the tendency to deliberately reduce, suspend, or avoid AI intervention at particular stages of creative work in order to regain control.

The authors propose a three-stage model: Source of Anxiety → Anxiety Transmission → AI Disengagement Intention. External pressures generated by AI are translated into specific forms of anxiety, which in turn influence whether designers decide to distance themselves from AI.

Designers Are Not Anxious About Just One Thing

One of the paper’s strongest ideas is that “AI anxiety” should not be treated as a single psychological state.

The study separates it into three forms: job replacement anxiety, skill obsolescence anxiety, and creative autonomy anxiety.

It also distinguishes three sources of pressure: the speed of technological iteration, pressure to continuously upgrade skills, and creative path deviation, a situation in which AI pushes the design process away from the designer’s original direction.

Creative path deviation is particularly interesting. It does not simply mean that AI produces a bad or irrelevant result. An AI-generated alternative may depart from the original concept while still being professionally viable. The designer must then decide whether to reject it, incorporate it, or reorganize the project around it.

What is often celebrated as AI’s ability to produce unexpected ideas therefore has another side. Each unexpected possibility creates another decision about who is actually directing the creative process.

The Strongest Concern Was Not Job Loss, but Skill Obsolescence

The researchers surveyed 382 design professionals and analyzed the relationships using PLS-SEM.

All three forms of anxiety increased AI disengagement intention. But their effects were not equal.

Skill obsolescence anxiety had the strongest relationship with disengagement intention at β = .391, followed by job replacement anxiety at β = .324 and creative autonomy anxiety at β = .277.

This is one of the paper’s most revealing findings.

The strongest behavioral pressure was not simply:

“Will AI take my job?”

It was closer to:

“Will the skills I spent years developing still have value?”

The survey operationalizes this concern through concrete questions: whether basic design skills such as composition and color are losing market value, whether accumulated professional experience is being eroded by standardized AI solutions, and whether previous investments in knowledge will continue to generate professional returns.

AI disengagement therefore appears to be connected not only to the ability to use AI, but also to the designer’s assessment of what prolonged AI use is doing to the value of their expertise.

Fast AI Development Does Not Automatically Produce Skill Anxiety

Table 6 produces another useful distinction.

The speed of technological iteration strongly increased job replacement anxiety, with a path coefficient of β = .495. Yet technological iteration speed had virtually no significant direct effect on skill obsolescence anxiety, at β = .036.

By contrast, skill upgrading pressure had a powerful relationship with skill obsolescence anxiety, at β = .602.

This separates two experiences that are often collapsed into the same complaint.

Watching AI improve at high speed is associated primarily with the thought that jobs may disappear. Feeling that one’s own skills are becoming obsolete is more closely connected to the repeated requirement to learn new tools, revise workflows, and reorganize learning plans.

In other words, the anxiety is not generated only by watching the machine become better. It emerges from having to continually rebuild oneself around the machine.

The study also finds that creative path deviation increases both creative autonomy anxiety and job replacement anxiety. When AI begins to redirect the creative process, designers may experience not only a loss of control over the work but also a threat to their professional function.

Novices React to Speed; Experienced Users Draw Boundaries

The most intriguing part of the study appears in the multi-group analysis.

Participants were divided into designers with less than one year of AI experience and those with one year or more.

Among short-term users, the effect of technological iteration speed on job replacement anxiety was β = .587. Among longer-term users, it fell to β = .379. The difference between the groups was statistically significant.

Experience appears to dampen the psychological impact of constant model releases, new features, and rapid improvements.

But the next stage reverses the pattern.

The effect of job replacement anxiety on AI disengagement intention was β = .410 for long-term users, compared with β = .247 for short-term users.

Experienced users therefore appear less reactive to the speed of technological change itself. Yet once they interpret AI as genuinely threatening their occupational territory, they are more inclined to reduce or suspend its use.

This suggests a shift in what designers pay attention to.

Early in AI adoption, the central question may be:

“How fast is this technology improving?”

After prolonged use, the question becomes:

“How far should this technology be allowed into my work?”

The object of concern moves from technological performance toward the division of authority between human and machine.

The paper therefore offers a useful way to understand AI disengagement. Designers do not necessarily choose between “using AI” and “rejecting AI.” They may turn it on, turn it off, restrict it to certain stages, and renegotiate its role throughout the process.

For creative professionals, the central issue may increasingly be less about whether AI is used at all and more about which decisions, skills, and territories remain under human control.