A leader's message written by AI: what happens to the team's trust

In a congratulations message written by a boss with little AI help, 83% of the professionals surveyed considered the manager sincere. With a lot of AI help, that rate fell to the range of 40% to 52%. The numbers come from a study of 1,100 professionals conducted by Anthony Coman, of the University of Florida, and Peter Cardon, of the University of Southern California, published in the International Journal of Business Communication. If you use AI to write emails to your team, recognition messages and tribute speeches, take time to consider this finding.

In this fifth article of the series, we talk about the relationship between you and the people who work with you. In speeches and internal messages, the audience knows the author, works alongside them and will keep doing so after the event. This is the ground where generic text costs the most.

The Coman and Cardon study on bosses' messages written with AI

According to the University of Florida, the researchers asked the 1,100 participants to evaluate a congratulations message sent by a manager, written with a low, medium or high level of AI assistance. The study also measured perceived professionalism. With little AI, 95% saw the manager as professional. With a lot of AI, the rate was between 69% and 73%.

The study brought another finding that matters to you as a manager. Participants judged their own use of AI more leniently than the use made by their boss. The bar rises when the writer holds a leadership position. The authors' recommendation, according to the university, is to avoid heavy AI use precisely in messages of praise, congratulations and personal feedback. We did not consult the original article directly, and the numbers here come from the University of Florida's press release.

Look at the list of situations. Praise, congratulations and recognition are the raw material of many corporate podium speeches: the award for the best salespeople, the tribute to the employee with 30 years at the company, the thank-you to the team that delivered a difficult project. In these moments, your team expects to hear your voice.

Vanderbilt's letter after a shooting

In February 2023, after the shooting at Michigan State University, the Office of Equity, Diversity and Inclusion at Peabody College, part of Vanderbilt University, sent an email to the entire academic community. The text spoke of mutual care and a sense of community. At the bottom, in small print, was a note that the message was a paraphrase from ChatGPT, date included. According to ABC News, the email also did not include the list of student support resources. The case was revealed by the student newspaper, the Vanderbilt Hustler.

The students' reaction was recorded by the American press. A senior whose sister studies at Michigan State said there was "a sick irony in having a computer write your message about community." According to CBS, other students called the letter impersonal and lacking in empathy, and some accused it of being a public relations move.

The next day, associate dean Nicole Joseph apologized and wrote that using ChatGPT went against Peabody College's values. The college's dean, Camilla Benbow, said the email had not gone through the several layers of review that were supposed to be in place, acknowledged that the communication failed to meet the need for personal connection and empathy and apologized "to all those who deserved better from us." Nicole Joseph and assistant dean Hasina Mohyuddin were temporarily removed from their roles while the university investigated the case.

The reports do not say how much of the text was written by people, and we did not consult the original Vanderbilt Hustler article. CBS and ABC recorded the episode as a case in which the text's origin took the place of the message.

An apology written by AI loses authenticity

Vanderbilt had to apologize for its own text. A study by Glikson and Asscher, published in 2023 in the journal Computers in Human Behavior, helps explain why an AI-mediated apology runs a similar risk. In three studies with workplace conflict scenarios, AI involvement reduced the perceived authenticity of the apology and, with it, the willingness to forgive. Disclosing that AI had been used did not offset the effect. The study's abstract does not report sample sizes.

The same work recorded a useful detail: when only one of the three tools tested was part of the process, the negative effect did not appear. The finding suggests that the effect grows as the machine takes over the part of the message that should carry the speaker's responsibility.

Research from Ohio State University, conducted by Liu, Kang and Wei with 208 adults, reached a similar conclusion about messages between friends: whoever used AI was seen as someone who made less effort. With your team, your perceived effort also counts, and the second article in the series details this study.

Trust in the leader: what is at stake in a message to the team

The 2025 Edelman Trust Barometer, according to a summary published by the Australian Institute of Company Directors, found that 68% of people believe business leaders deliberately mislead the public, a record. The same survey indicates that 75% trust their own employer. The figure comes from a secondary source, because Edelman's page could not be consulted, and it should be checked against the original report.

With that caveat, the numbers help you gauge what you are managing. Trust in one's own employer is high and coexists with deep distrust of business leaders in general. Every internal message weighs on that balance. A text that sounds like a standard statement brings the boss the team knows closer to the generic figure of the executive, precisely the one that triggers distrust.

Leader messages that call for your own words

Putting the research and the Vanderbilt case together, we list the situations in which the team most expects the real presence of whoever signs the message:

• Recognizing and awarding people by name, at conventions or all-hands meetings.
• Messages of solidarity after a loss, an accident or a tragedy.
• Apologies from leadership for a company mistake.
• Farewell, retirement and tribute speeches.
• Announcements about layoffs, restructurings and changes that affect people's lives.

In these situations, the content needs to come from you: the memories, the names, your honest opinion about what happened. With that material in hand, a scriptwriter helps organize the speech, find the right story and adjust the rhythm for the stage. The sixth article in the series applies this logic to convention and year-end speeches, when the whole team is in the audience.