Somewhere between the inbox and the heart, a small question has taken root in working life: when a manager's warm words are shaped by a machine, do they still belong to the one who sends them? A study of 1,100 professionals, led by researchers at the University of Florida and the University of Southern California, suggests that people will forgive the machine for making a message tidy, yet hold the sender to a stricter account of sincerity once its hand is heavy on the page. The lesson reaches beyond email, touching an old human need to be seen by someone who actually took the time, and it ask
AI-Polished Manager Emails Can Erode Employee Trust, Study Finds
Polished, professional, and less trusted: the cost of an AI-written thank-you
Mark
So the headline finding is that AI makes emails better but makes the boss look worse?
Mimi
Roughly, yes. People rated AI-assisted messages as professional and effective, but when a supervisor used a lot of AI, employees trusted the sender less.
Luke
Before we go further, what did participants actually read? Real emails from real bosses?
Mimi
Not exactly. They evaluated versions of a congratulatory message and were told how much AI assistance had been used. So the reaction is to the stated level of AI help.
Mark
That's a notable detail. They were told, not guessing.
Luke
Right, and the source also says employees can often detect AI content. That's a claim I'd hold loosely, since the survey design tells people the answer.
Mimi
Fair. The study measures how disclosed AI use is judged, which is still relevant, because disclosure is increasingly likely.
Mark
What are the numbers?
Mimi
Sincerity: 83% for low-assistance messages, 40 to 52% for high. Professionalism: 95% versus 69 to 73%.
Luke
Those are survey percentages from one study of 1,100 people. They show a strong gap, but not how it plays out in an actual office over months.
Mimi
And the perception gap is interesting. People judge their own AI use gently, but not their boss's.
Mark
So the practical rule?
Mimi
Use AI for reminders and announcements, keep praise and motivation mostly in your own words.
O Pulso
- A congratulatory note from a manager now arrives polished, yet a hidden chatbot authorship can quietly unsettle the person reading it.
- Employees readily accept AI for grammar and editing, but their goodwill thins sharply when supervisors lean on it heavily.
- Sincerity ratings collapse from 83% for lightly assisted messages to between 40% and 52% for heavily assisted ones, cutting into trust in leadership.
- Workers judge their own AI use leniently while holding supervisors to a harsher standard, a double standard that widens the perception gap.
- The researchers' emerging guidance points toward reserving AI for routine notices and keeping praise, motivation and congratulations in human hands.
Somewhere between the inbox and the heart, a small question has taken root in working life: when a manager's warm words are shaped by a machine, do they still belong to the one who sends them? A study of 1,100 professionals, led by researchers at the University of Florida and the University of Southern California, suggests that people will forgive the machine for making a message tidy, yet hold the sender to a stricter account of sincerity once its hand is heavy on the page. The lesson reaches beyond email, touching an old human need to be seen by someone who actually took the time, and it asks leaders to remember that efficiency and care do not always travel together.
Picture a congratulatory email from your manager: warm, well-organized, free of typos. Now picture learning that a chatbot wrote most of it. A study of 1,100 professionals suggests that this knowledge changes how the message lands, and not in the sender's favor.
Anthony Coman of the University of Florida and Peter Cardon of the University of Southern California designed the research, published in the International Journal of Business Communication. Participants read congratulatory messages labeled as having low, medium or high AI assistance, then rated both the message and its sender. The messages themselves were generally judged efficient and professional. The senders were judged differently.
The drop is stark. When a supervisor's message involved only light help, such as grammar fixes, 83% of employees saw them as sincere. Under heavy AI use, that figure fell to between 40% and 52%. Professionalism slipped too, from 95% to between 69% and 73%.
The researchers also found a perception gap. People rated their own AI use as acceptable at nearly every level, but applied a far stricter standard to supervisors using the same help. Heavy AI writing led employees to question a manager's authorship, integrity, caring and competence, the qualities that anchor trust in ability.
The authors' advice is practical: consider the message type, the amount of assistance and the relationship before writing with AI. Meeting reminders and factual announcements tend to be received well. Empathy, praise, congratulations, motivation and personal feedback are better written by hand. For managers, the real question is less whether to use AI than which messages to keep for themselves.
Imagine a congratulatory email from your manager landing in your inbox, warm and tidy and free of typos. Then imagine learning that a chatbot wrote most of it. A study of 1,100 professionals suggests that this knowledge changes how the message is received, and not in the sender's favor.
The research was carried out by Anthony Coman, Ph.D., of the University of Florida's Warrington College of Business, and Peter Cardon, Ph.D., of the University of Southern California. It appears in the International Journal of Business Communication. The authors start from the observation that AI is now a routine part of workplace writing, with most professionals using tools such as ChatGPT and Gemini.
Their method was straightforward. Participants were shown versions of a congratulatory message and told that it had been written with low, medium or high levels of AI assistance. They then rated the message itself and the person who sent it. On the message, the verdict was mostly positive: AI-assisted writing was generally seen as efficient, effective and professional.
The sender fared differently. Coman describes a tension between how people judge the quality of a message and how they judge the person behind it. Managers who used medium to high levels of AI assistance for routine communication put their trustworthiness at risk, he said, despite the polished impression the writing made.
The numbers show the size of the drop. When a supervisor's message involved low AI help, such as grammar or editing fixes, 83% of employees saw the supervisor as sincere. With high levels of AI, only 40% to 52% did. Professionalism followed the same pattern: 95% rated low-AI supervisor messages as professional, but that fell to between 69% and 73% when supervisors leaned heavily on the tools.
The researchers also found what they call a perception gap, which depends on who is using the tool. People tended to rate their own AI use about the same whether it was low, medium or high. When judging others, the amount of assistance mattered a great deal. Coman summarized it this way: professionals view their own AI use leniently, but they are more skeptical of the same levels of help when supervisors use it. Heavier AI writing led employees to question a manager's authorship, integrity, caring and competence. Coman said it can undermine perceptions of ability and integrity, both central to cognitive-based trust.
The findings indicate that employees can often detect AI-generated content, and that they may read its use as laziness or a lack of caring. The effect appeared when supervisors relied heavily on AI for messages such as team congratulations or motivational notes, where employees saw less sincerity and began doubting leadership ability.
The authors' practical advice is to weigh the type of message, the level of AI assistance and the relationship before using AI to write. Informational or routine messages, such as meeting reminders or factual announcements, appear to be received well when AI helps. Messages that carry empathy, praise, congratulations, motivation or personal feedback are better written with minimal technological intervention. For managers, the question is less whether to use AI than which messages to leave to themselves.
Citações Notáveis
Managers who use AI for routine communication tasks put their trustworthiness at risk when using medium- to high-levels of AI assistance.— Anthony Coman, University of Florida's Warrington College of Business
Quer a matéria completa?
Leia o original em ScienceDaily ↗
Education