Coworkers Rate AI Users as Lazier, 5 Ways to Fix AI-Sounding Work Before Your Review

If your emails, reports or self-reviews sound like a chatbot wrote them, your colleagues may be noticing, and it can cost you credibility. Duke University researchers found that employees judge colleagues who use AI as less competent and lazier, and recruiting managers may act on that perception. At Metaintro we care less about whether you use AI and more about what your work signals when review season arrives. This is a practical guide to keeping AI in your toolkit without letting it show up as a weakness in your file.

What happened

Most workers are figuring out AI etiquette on their own. CNBC reported in August 2026 that roughly half of workers say they use AI anywhere from once a week up to multiple times per day, and the CNBC and SurveyMonkey Quarterly AI and Jobs Survey behind that report found that three in ten workers are using AI every day at work. The same survey found that 84% of people say they're self-taught with no formal training in AI, and most employers have not adopted any official AI usage policy (55%).

That vacuum shows up as suspicion. CNBC spoke to a 30-year-old HR worker in London whose colleague praised an email she had written and then asked whether she had used AI on it, even though she researched and crafted the message without the help of AI. Her reaction captures the stakes: you want the credit, you want your skill to be recognized. The penalty is measurable. In a June 2026 experiment by Atlassian's Teamwork Lab with nearly 1,000 U.S. knowledge workers, participants rated a hypothetical peer who disclosed their AI use as 10x lazier, and were 24 percentage points less likely to recommend them for a high-visibility project, even when the output was identical. That is striking because Atlassian also found that 94% of U.S. knowledge workers report using AI at work.

The strongest evidence comes from peer-reviewed research. The Duke team's paper in the Proceedings of the National Academy of Sciences, published on May 8, 2025, ran four preregistered experiments (N = 4,439) and found that people who use AI at work anticipate and receive negative evaluations regarding their competence and motivation.

People who use AI at work anticipate and receive negative evaluations regarding their competence and motivation, according to a 2025 PNAS paper covering four preregistered experiments with 4,439 participants.

The details make it hard to dismiss. In one experiment, Duke's summary of the study explains, participants read about a lawyer who sometimes asked generative AI to summarise information, or who asked a paralegal instead, and people who used AI were consistently rated as lazier, less competent, and less diligent than those who received help from human colleagues. Professor Richard Larrick said this social penalty for using AI held true regardless of the employee's age, gender, or occupation.

The part that hurts most is what happens when AI output is not very good. Researchers at BetterUp Labs and the Stanford Social Media Lab coined the term workslop, defined in TechCrunch's report on the study as AI generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task. In an online survey of 1,150 full-time U.S. desk workers conducted in September 2025, BetterUp found that 40% of U.S. desk workers received workslop last month.

The cost lands on the person receiving it. CNBC Make It, citing the BetterUp and Stanford research, reports that people who encountered workslop spend an average of one hour and 56 minutes dealing with the aftermath of it. BetterUp puts the price at $186 monthly cost per employee and a $9M annual cost for a 10,000-person company. Among recipients, 53% report being annoyed, 38% confused and 22% offended, according to CNBC Make It.

Roughly half of workers say they consider their co-workers less creative, capable and reliable after receiving workslop from them, according to CNBC Make It, citing 2025 BetterUp and Stanford research.

Managers are not exempt. A 2025 study in the International Journal of Business Communication surveyed 1,100 working professionals and found that respondents begin to question the authorship, confidence, caring, sincerity, and ability of senders who use medium to high levels of AI assistance. Euronews, covering the study, reported that only 40 per cent of employees viewed supervisors as sincere when they used high levels of AI, compared to 83 per cent for low-assistance messages.

Where the coverage agrees, and where it splits

The research agrees on the basic finding. There is a reputational cost to AI at work, and it reaches real decisions. Atlassian found colleagues were less likely to recommend them for a high-visibility project, and Duke found that recruiting managers may act based on this perception by penalizing job candidates who rely on AI.

The first split is over what triggers it. The Duke experiments and the Atlassian study show that simply being known to use AI can cost you, and Atlassian saw the penalty even when the output was identical. The workslop research points at quality instead. Stanford professor Jeff Hancock, the founding director of the Stanford Social Media Lab, told CNBC Make It that producing sloppy work used to take a fair bit of effort, and that now that the effort piece is gone, I can generate a lot of useless or unproductive content very easily. The London HR worker in CNBC's reporting was suspected without using AI at all. You cannot fully control whether people suspect you. You can control whether your work gives them evidence.

The second split is over who judges most harshly. In a third experiment, 1,718 participants were instructed to act as managers, and Duke found that managers who rarely used AI themselves were less likely to hire candidates who used AI daily, while managers who used it frequently favoured those candidates. The email study found a different asymmetry, with researcher Anthony Coman telling Euronews that professionals view their own AI use leniently, yet they are more sceptical of the same levels of assistance when used by supervisors. Put together, the person reviewing you matters as much as the tool.

The third split is the practical one, whether to disclose. Hancock argues for openness, saying that if you tell a colleague the work is AI-generated, they can have a better sense of what prompts you were working with and what your goal was. The Duke researchers are more cautious, suggesting that being transparent about AI use might be most effective when you can clearly demonstrate how it enhances rather than replaces your unique skills. In their first experiment, with 500 participants, participants in the AI tool condition believed they would be perceived as lazier, more replaceable, less competent, and less diligent, and Jessica Reif warned that employees might hide their AI, even when that AI use is beneficial for the organization. Atlassian landed in between. It found that workers who framed their AI use as team-focused were rated 11 percentage points higher on effort and 8 points more likely to be recommended than those who cited personal efficiency, but summed up its own result as framing matters, but silence is safer.

What the coverage mostly glosses over is that the penalty is not fixed. Duke's Jack Soll said that this penalty could disappear 10 years from now. Atlassian found it can already shrink, reporting that in companies that actively celebrate AI use, the laziness penalty almost entirely disappears, and AI users are rated as more efficient than peers who don't disclose. That is useful perspective, but you cannot choose your company's culture before your next review.

What this means for your job search

Our argument is that the AI penalty is really an effort penalty. Colleagues cannot see how you work, only what you hand them, so they use the texture of the work as a stand-in for how much you care. Writing that is long, generic and polished in a hollow way reads as low effort whether or not a model wrote it. That means the fix is in your control, and it comes down to five habits.

First, cut the length. Hancock told CNBC Make It that a tell-tale sign of workslop is purple prose, like using three paragraphs of text when one bullet point would suffice. Synthesia co-founder and chief executive Victor Riparbelli made the same point to CNBC, saying AI can make users look articulate but sometimes it's just a lot of filler. If a draft could be half as long, make it half as long before you send it.

Second, add what only you know. A model can produce fluent sentences but not the name of the client who pushed back, the number from last Tuesday's dashboard or the decision you made and why. Workslop, by definition, lacks the substance to meaningfully advance a given task, so specifics are the clearest evidence that a human did the thinking. The same principle is behind writing a cover letter with AI that still sounds like you.

Third, match your use of AI to the kind of message. The 2025 email study covered by Euronews found that when a message was seen as purely informative, employees tended to view AI use positively, but people were far less accepting for relationship or motivational messages. Duke found no significant difference in task fit ratings for digital tasks where AI was explicitly described as useful. Use it freely for summaries, meeting notes and first drafts of routine updates. Write thank-yous, feedback, apologies and anything about a person yourself.

Fourth, never push the work downstream. The BetterUp researchers warned, in TechCrunch's summary, that workslop shifts the burden of the work downstream, requiring the receiver to interpret, correct, or redo the work. Read your draft as the person receiving it. If they would need to ask a follow-up question to act on it, answer that question first. The same goes for redirecting people to a chatbot. The CNBC and SurveyMonkey survey found that one in four (23%) users of AI admitted to redirecting a colleague toward AI before providing their consultation, and Ask a Manager creator Alison Green told CNBC that it feels really lazy and sloppy to people.

Fifth, audit your review file before your manager does. Your self-assessment, goal updates and the project summaries you sent this year are the evidence your manager will skim. Rewrite anything that sounds generic, lead with outcomes and numbers you can stand behind, and make sure the voice matches how you talk in meetings. Companies are still working out how AI fits into reviews, as Meta's decision to stop grading engineers on AI token counts shows, so do not assume the rules your reviewer uses are the rules you expect.

One more point for anyone job hunting right now. The Duke study found the penalty reaches hiring, with managers who rarely use AI less likely to hire candidates who used AI daily. Your applications, outreach messages and take-home tasks face the same reader test. The candidates who do well will be the ones using AI to work faster while making sure every piece of writing still sounds like a specific person who did the work. If your team is already drowning in low-quality AI output, this guide to spotting and stopping workslop is a good place to start the conversation.

Related Articles

• When You Should Not Use AI at Work
• Drowning in AI Busywork? Why 'Workslop' Is the New Workplace Crisis
• How to Use AI at Work Without Crossing Ethical Lines
• The Hidden Risks of Bring-Your-Own-AI at Work
• Gen AI Performance Reviews, Will They Fix Workplace Feedback or Break It?
• The AI Job Search Mistakes That Get You Screened Out
• Writing Networking Messages With AI That Get Replies
• Prompt Writing Skills for People Who Are Not Engineers
• Email Etiquette Rules Every Job Seeker Needs Before Hitting Send
• Why Gen Z Workers Are Pushing Back on AI at Work
• AI Skills for Managers Leading AI-Augmented Teams

People Also Asked

Q: Do coworkers really judge people who use AI at work?

A: Yes, according to peer-reviewed research. A 2025 PNAS paper by Duke researchers reporting four preregistered experiments with 4,439 participants found people who use AI at work anticipate and receive negative evaluations of their competence and motivation. Duke's researchers note the penalty occurs more strongly among evaluators who are less frequent users of AI and on work tasks where AI is not directly of use.

Q: What is workslop?

A: Workslop is AI-generated work that looks polished but lacks the substance to move a task forward. BetterUp Labs and Stanford researchers coined the term, and in their September 2025 survey of 1,150 full-time U.S. desk workers, 40% said they had received workslop in the past month. The researchers say it shifts the burden of the work downstream, requiring the receiver to interpret, correct, or redo the work.

Q: Should I tell my manager when I used AI on a piece of work?

A: The research is split. A June 2026 Atlassian experiment found colleagues rated a peer who disclosed AI use as 10x lazier, even when the output was identical, though framing the use as team-focused softened the penalty. Duke's researchers suggest transparency works best when you can clearly demonstrate how it enhances rather than replaces your unique skills. A 2025 email study covered by Euronews found AI use was viewed positively when a message was purely informative, but far less so for relationship or motivational messages.

Future-proof your career. The workers who get ahead with AI are the ones whose judgement stays visible in everything they send. Metaintro matches people with strong AI and communication skills to companies hiring now. Create a free profile and find a role that rewards how you think.