Why do people expect colleagues to see them as less capable just for using AI at work, and who gets the credit?

This explores why using AI at work seems to cost people socially in a way other tools don't. One caveat first: the collection has strong evidence that the penalty exists, but no study that directly compares AI with spreadsheets, calculators or search engines. What it does offer is a convincing explanation of what makes AI different. The core evidence comes from four experiments with more than 4,400 participants. People who used AI expected colleagues and managers to rate them as less competent and less diligent, and they became less willing to say they had used it Do people fear judgment when they use AI at work?. The penalty is aimed at the person, not at the quality of the work. That points to a question of who gets credit, not whether the output is good.

The likely difference is that most workplace tools leave a clear line between your contribution and the tool's. A spreadsheet does the arithmetic, but everyone knows you built the model. AI blurs that line. Research on what the corpus calls the 'LLM Fallacy' finds that when AI output is smooth and fluent, users themselves lose track of where their work ends and the machine's begins Do AI-assisted outputs fool users about their own skills?. Four overlapping causes make this worse: it's unclear who did what, polished output feels like skill, the thinking gets handed off, and the steps are hidden from view How do AI tools trick users into overestimating their own skills?. Here is the twist. That same blur cuts both ways. Users tend to resolve the uncertainty in their own favor and overrate their skills. Observers tend to resolve it against them and assume the AI did the real work. The penalty and the self-flattery are two sides of one problem.

The 'diligence' part of the penalty has a specific explanation. An analysis of 1,250 worker interviews found that people carefully protect identity cues like their own voice and where the work came from. Meanwhile, signs of effort, attention and uncertainty quietly disappear into the finished deliverable Which workplace cues survive AI mediation and which disappear?. If the only visible evidence of work is the polished result, and AI could have produced that result, observers have nothing left to prove you tried. Hiding AI use, which the Reif study found people do, makes this worse because it removes even more context. The penalty is also concentrated in information-heavy jobs, where AI delegation is most common Where have workers actually delegated tasks to AI?. In those jobs the written output is the main evidence of competence.

There's also a reasonable basis for the suspicion. Competent-looking AI can quietly weaken people's skepticism and spread accountability across several actors until nobody clearly owns a decision How do competent systems quietly undermine safety oversight?. A colleague who discounts AI-assisted work may be responding, imperfectly, to a real loss of clarity about who stands behind it.

The design research hints at a way out. Systems that offer guidance, such as pointing out what to look at, instead of handing over decisions keep responsibility visibly with the human Can AI guidance reduce anchoring bias better than AI decisions?. Assistants that ask reflection questions alongside their advice also produced better decisions than ones that only answered Do reflection questions help people make better decisions with AI?. Neither study measured how colleagues perceive the user. But both suggest that the penalty may depend less on whether you use AI and more on whether your own judgment stays visible in the work.