How to Make a Slack Message Firm but Not Rude

Two messages go to the same colleague on the same Thursday. One reads “Could you send the signed copy by 3pm? Legal closes at 4.” The other reads “Still waiting on the signed copy, as I said yesterday.” Both senders want the file by the same afternoon, but only the first says when. The second also tells the reader what you think of them, and “as I said yesterday” is the part to cut.

I co-founded Subtext, an app that marks the words in a draft that could read as harsh or vague, so I gain if you use it. Further down, Subtext reads a friendly chase that says both “soon” and “no rush at all”, and I show the tags it gave and the three versions it offered. If you only want wording to copy, go to the four parts and the rewrites.

I think many messages that read as rude are firm messages with one extra sentence about the person, and the fix is to cut that sentence and keep the firmness. In experiments with invented email lines, writers overrated how clearly their tone came across¹. In real messages between people who knew each other, positive or negative tone mostly arrived as meant². And in email, one line saying a message was not urgent narrowed the gap in how fast readers thought a reply was expected³.

Why does a firm message come across as rude?

Firm is about what you ask for or refuse. Rude is about what the message adds about the person, or how it lands. The research complicates that split in three ways.

First, rude is the reader’s judgement. Andersson and Pearson’s 1999 definition of workplace incivility has three parts: low intensity, ambiguous intent to harm and a breach of norms of mutual respect⁴. In a 2025 interview study, nine mediators described disrespectful exchanges with no intent to harm in which the people involved still, in most cases, read malice into the other’s behaviour⁵. These are mediators’ recollections of workplace behaviour in general, not of written messages, yet they fit the premise that something can be taken as rude without being meant that way.

Second, you may not hear the edge yourself. In five experiments, US students wrote or picked lines for a partner to read by email, and in two of them also to hear¹. In the largest, 154 pairs tried to convey four tones. Among the roughly one third of those pairs who used email, senders expected 88 per cent of their lines to be understood and readers, picking from four tones, managed 62.8 per cent, where guessing would score 25 per cent. The lines were invented and the writers students, and the explanation, that you hear the tone you meant, rests on two small tests of 54 and 58 students, so I would treat the percentages and the explanation as tentative. The authors suspected that chat’s speed might widen the gap but did not test it¹. Does this text sound rude? sets the experiments out study by study.

Third, in real exchanges positive or negative tone mostly arrives as meant. Pollmann and Roos asked both senders and receivers of real messages between people who knew each other, and receivers’ ratings of positive or negative tone matched what the senders intended, for 347 receivers of text messages and 361 of work or education emails². The study rated positive against negative tone only, so it says nothing about irony, firmness or coldness. It also tested six possible moderators, such as closeness, and the authors report no evidence that any of them changed how well the ratings matched².

Which small cues tip a message?

In lab studies, small cues can change how a short message reads, although the real-message study above found no evidence that message length or emoji changed how well ratings matched². The lab studies are narrower than the headlines.

The full stop. In one lab study, 126 undergraduates rated one-word replies such as “okay” and “sure” as less sincere when they ended in a full stop, in text messages and not in handwritten notes⁶. A 2025 study with 200 Australian undergraduates found that a final full stop made invented replies read slightly more negative at one word and at three to four words, and not significantly at six to eight⁷. Both used invented exchanges, one of them about a fictitious friend, so they say little about work. Does a full stop make a text sound angry? has the detail.

Emoji. In three experiments on an unfamiliar sender’s work email, smileys did not raise perceived warmth and lowered perceived competence, and in the last one the penalty appeared in formal messages only, while informal ones gained warmth⁸. A preregistered replication of that one, with 847 people against the original 85, found smileys lowering competence and raising warmth whatever the formality⁹. In a study where 243 adults read simulated workplace instant messages, a positive emoji raised perceived competence next to positive or neutral text and made no difference next to negative text¹⁰. So I would not trust any single rule, and as far as I could tell none of these tested teammates who already know each other.

Capitals, length and greeting. I found no peer-reviewed study of capitals in workplace chat, and the authors of the full stop study note that capitals may convey a range of emotion beyond anger⁷. The studies I read give no word count above which a chat message stops reading as curt, and none shows that it needs a greeting on every turn. “Yes, Tuesday works” is complete when it answers a simple question.

What is different about chat at work?

Chat tends to run faster than email, and Slack itself is missing from the tone studies I found. Almost every study I found has students or online panels rating invented emails or texts, and none tested threads, channels, @mentions, reactions or edits, so what follows about chat is inference.

Part of the pace is expectation. In a 2000 study of early instant messaging, most of the 20 workers interviewed and observed felt they could ignore a message without offending the sender, possibly because the sender could not be sure they were there¹¹. It predates Slack, and it points to something I think still holds, that how fast an answer is expected is negotiated between people.

The one piece of advice here that has an experiment behind it concerns that expectation. In eight preregistered studies, 4,004 people in all and mostly imagined scenarios about email, readers overestimated how quickly senders expected a reply to a non-urgent message outside work hours. In the study that tested it, a one-line note from the sender saying it was not urgent reduced the gap to a difference that was not significant³. For Slack that is an inference, and a cheap one to act on with a line like “For tomorrow, no reply needed tonight.” What makes a reply land well covers how fast to answer yourself.

Slack’s own help page describes the dot beside a name as availability, active when Slack is open and away after a period of inactivity, and it does not say the person has seen your message¹².

Does the same message read differently depending on who sends it?

The same words may read differently from a manager, a peer and a contractor, but I found no study that held the wording fixed and changed the sender’s role. The nearest evidence compares native and non-native senders, in email and short written scenarios. I would expect that “Could you do this tonight?” may be heard as an instruction from a manager and as an invitation to negotiate from a peer, and that a junior colleague can still state a dependency.

In one study, 169 businesspeople found errors by non-native speakers less bothersome than the same errors from native speakers, and found errors of politeness and tone more bothersome than grammatical ones¹³. In an email experiment, knowing the sender was from another culture reduced most of the bad impressions from spelling and grammar errors, and did not significantly reduce those from short messages with no conversational tone¹⁴. Do my English texts sound polite? has the longer account of the email experiment.

What goes in a firm Slack message?

A firm message has four parts, and a short one can hold all four in two sentences. The parts are my own framework.

The ask, or the no. What exactly needs to happen, or what you will not do. If you cannot point to it in your draft, the message is a hint.

The reason. The dependency the reader needs to know about, such as “the printer needs the file”. Leave out any case against their behaviour.

The time. A deadline, with the time zone if it matters, and whether it is a requirement or a preference. Making an expectation explicit is the nearest I can get to a tested ingredient, since in one email study a line saying the message was not urgent narrowed the gap³. It did not test deadlines.

The way out. One real option for when they cannot, such as who else can decide or the earliest they can manage. One, so that it does not turn into hedging.

The test is functional. Can the reader find the action and the time, see why it matters, and answer without having to answer an accusation? Not every message needs all four, and “Yes, Tuesday works” needs none. I think softening that removes the ask or the time makes a message vaguer without making it kinder, so if either is missing, add it before you soften anything.

Chasing a reply and a dropped follow-up

These messages are made up for this post, and each is one way through, not the correct text. The line above each pair holds the facts the after uses, and the times are placeholders you would replace with ones you will hold. I picked befores that look harmless, because an edge in a phrase you would not think twice about is the one you are least likely to hear yourself. The afters drop any smiley only to keep the comparison on the wording, and keeping one is your call.

Chasing a reply. You asked in a thread whether the old or the new logo goes on the flyer. Two days on there is no answer, and the printer needs the file by Thursday noon.

Before “Bumping this again. Two days now.”

After “Back to my question above: old logo or new on the flyer? The printer needs the file by Thursday noon. If it’s someone else’s call, who should I ask?”

“Again” and the count of days say nothing the reader needs and a lot about how you feel. The pressure stays, now with a reason and a time, and the last sentence is the way out. It turns “you have not answered” into a question about whose call it is.

A dropped follow-up. A colleague said on Monday’s call that they would send the client the revised quote. It is Wednesday and the client has written to ask where it is.

Before “Funny, the client says they’ve heard nothing about the quote.”

After “The client has asked for the revised quote and has not received it. Could you send it today, or tell me if something is holding it up so I can reply to them?”

The before can read as a dig dressed as an observation, and it asks for nothing. The after states the fact, says what is needed and by when, and offers a second way to answer.

Saying no to scope or a meeting, and correcting a mistake in public

The same four parts apply to a no and to a correction.

Changing the scope. Someone senior asks in a channel for a competitor summary on top of the update due Friday, and you can manage one or the other.

Before “Sure, I can fit that in 🙂 I just won’t be sleeping much this week.”

After “I can finish the update or do the competitor summary by Friday, not both. Which should come first? If both have to land Friday, which one can move?”

The joke is the edge, and it does not say no. The after refuses the impossible version, offers two things you can really do, and hands the choice to whoever owns the priorities.

Declining a meeting. The Thursday sync clashes with the release notes, which are due that morning.

Before “Ugh, another sync 😅 I’ll try to pop in if I can.”

After “I won’t be at Thursday’s sync, because the release notes are due that morning. If you need a decision from me, please put it in the thread.”

The before sounds light but declines nothing, and “another” carries a small dig. The after is a plain no with its reason, plus a way to reach you, which is worth offering only if you will really check the thread.

Correcting a mistake in public. Someone posts the 6th as the client demo date to a channel of forty. It is the 9th, and the invite says so.

Before “Tiny thing 🙂 the demo is the 9th, not the 6th, like I mentioned on Monday. Worth double checking before posting next time.”

After “The client demo is on the 9th, not the 6th, as the calendar invite shows. Worth a look if you planned around the 6th.”

The after corrects the fact and leaves the person out. “Like I mentioned on Monday” and “next time” can put one person on the record in front of forty. The invite stays because a correction needs a source. If the slip keeps happening, that is a private conversation, covered in giving critical feedback to a coworker in writing.

Which phrases are worth swapping?

In my reading, many of the phrases people call rude leave out the ask or point at the person. I call a message passive aggressive only when it hides irritation behind polite words, and I do not call the sender that.

Some lists of phrases that are supposedly rude at work rest on vendor polls about email, such as a 2023 survey of 1,006 US adults by a language-tutoring company, which recorded what respondents said they had received or sent and tested no effect¹⁵. Is that email passive aggressive, or just short? goes through those lists, and the email version of this table has twenty rows.

If you were about to send	What the phrase often does	One way to put it
“Any update?”	Leaves out the topic and the time	“Has the venue booking gone through? I need to know by Wednesday.”
“Just checking in”	Leaves out the ask	“Can you tell me by 2pm who signs off the translation?”
“Quick question” and nothing else	Makes the reader wait for the question	Send the whole question in that message
“As I said”	Puts a rebuke in place of information	Say it again plainly, as in “The cut-off is Friday.”
“Per the thread”	Sends the reader searching	“We agreed to launch on Monday. The decision is in the thread above.”
“ok.”	Leaves three readings open (receipt, agreement or displeasure)	“OK, that works for me.”
“Noted”	Leaves the same three readings open	“Thanks, I’ve read this.”
“No worries if not”	Undoes the request it follows	“I need it by Friday. If you can’t, please say so by noon.”
@channel for one person’s task	Interrupts everyone for one owner	“@Zainab, can you take the venue booking?”
A thumbs-up and nothing else	Leaves open whether it means seen, agreed or done	“Done.” or “I’ll do it by Friday.”

In my reading, the edge can sit in one small word, such as “actually”, “again”, “still” or “obviously”. If it added only a rebuke, cut it. Keep the original when it fits.

Thread, channel, DM or call: where should it go?

The same words do different work in a thread, a channel and a direct message. These checks come from how Slack works.

In a thread, restate the open question. “Any update?” under forty replies can leave readers guessing which work is blocked.
In a channel, give enough context for everyone reading, and correct the shared fact and leave the person out.
In a direct message, put the ask and the time in the first line.
With an @mention, name the person who owns the task. A channel-wide ping is a decision to interrupt everyone, not a way to strengthen a weak deadline.
After hours, say whether it can wait. If it cannot, use whatever your team treats as the urgent route.
After an edit that changes an action or a deadline, post a visible follow-up.
With reactions, agree what they mean, because a tick can be received, done or approved.

Some exchanges need a voice. In the email experiments, voice narrowed the gap in how well tone came across¹, and I would expect a short call to be the quickest way to repair a thread that keeps going round without an answer, if a call suits you both. Say what the call is for, because “let’s take this offline” can sound like a shutdown, and put the decision in writing afterwards.

Before you send anything sharp, draft it outside the send box and read it aloud in the voice you fear it has. That is my adaptation of the fourth experiment, in which 54 students were split in two. Half read their statements aloud in the tone they meant and half in the opposite tone. The opposite-tone half lost their overconfidence about how it would come across, and the half who read as meant kept it¹. That was a sarcasm task with students, so treat it as a habit to try.

Where Subtext fits, and what to paste

Subtext gives a draft a second reading. For AI-polished messages I found two randomised experiments on suggested replies, in short chats between strangers. People suspected of using the replies were rated lower, which is an association, because suspicion itself was not randomised, while partners who really used them were rated more cooperative¹⁶. Does AI make your texts sound like a robot? covers how readers take it.

Subtext reads a draft you wrote for a work chat and marks the words that could read as curt, pushy, passive aggressive or unclear, with a short reason for each. It offers up to three versions, the first staying close to your wording, each marked safe to send or heads-up. A safe to send badge is Subtext’s own estimate from the words in front of it, and a heads-up is a reason to read that version again, not proof of a problem. You can ask for a shorter or more confident version, or type what you want changed. A version written from an instruction defaults to warm, plain and short, so a firm one may need you to say so. It cannot tell what your colleague meant, as the site’s FAQ says, it does not know who holds the power in the thread, and it is not a Slack integration or a keyboard. It sends nothing, so you copy the version you pick and send it yourself.

On a draft made up for this run, “Hi Nadia, just circling back on the Q3 numbers 🙂 Not sure if you saw my last message? Would be great to have them soon if you get a chance, no rush at all.”, Subtext named the feeling as Patience and returned two tags, Mixed messages and Passive-aggressive. The note on the first named “Would be great to have them soon” and “no rush at all”, which “send two different signals, so Nadia can’t tell how urgent this is”. The second went on “Not sure if you saw my last message?”, which “can read as a quiet dig about her not replying”. The first note speaks for Nadia, which the app cannot know, so read both notes as comments on the wording.

All three versions were marked safe to send. The first, Close to yours, kept the smiley, dropped the “Not sure if you saw my last message?” line and asked “Could you send them over when you get a chance?”. The second, Warmer polish, added “hope your week went well” and “in the next few days”. The third, Fresh rewrite, offered help, “If anything’s holding them up, just tell me and we can sort it out together”, which the draft never said. Its note says the version “keeps pressure low and invites a reply”, again an estimate and not knowledge of Nadia.

None gives a specific deadline, and under the versions Subtext said “I didn’t add a specific deadline since you didn’t mention one.” Before the versions it had said it made the timing clearer, yet the only timeframe is that soft “next few days”. So it took out the part that could read as a dig and left the firm part, a specific deadline, to the sender. That is the app’s real output, unedited.

Hi Nadia, just circling back on the Q3 numbers 🙂 Not sure if you saw my last message? Would be great to have them soon if you get a chance, no rush at all.

Mixed messages: "Would be great to have them soon" and "no rush at all" send two different signals, so Nadia can't tell how urgent this is.
Passive-aggressive: "Not sure if you saw my last message?" can read as a quiet dig about her not replying.
Patience: You're gently nudging while trying not to sound pushy.

A friendly nudge like this is totally fine, and your tone is already kind. I just made the timing clearer and took out the "did you see it?" bit.

Close to yours (Friendly, safe to send): Hi Nadia, just circling back on the Q3 numbers 🙂 Could you send them over when you get a chance? Keeps your friendly opener and emoji, and makes one clear, light ask without the mixed signals.
Warmer polish (Warmth, safe to send): Hi Nadia, hope your week went well 🙂 Just checking in on the Q3 numbers. Would you be able to send them over in the next few days? Thanks so much. Warm and polite, with a soft "next few days" timeframe so she knows what you need.
Fresh rewrite (Supportive, safe to send): Hi Nadia, quick one: could you send me the Q3 numbers when you have a moment? If anything's holding them up, just tell me and we can sort it out together. 🙂 Leads with the ask and offers help if something is blocking her, which keeps pressure low and invites a reply.

I didn't add a specific deadline since you didn't mention one. Want it shorter, warmer, or a bit more confident?

That was one run on one draft. The badge is a second reading of the words, and it read safe to send for all three versions although none held a specific deadline, so here it was no firmness check. I found no independent evaluation of Subtext, or of any tone tool, on workplace messages, so read each version against what you will commit to before you copy it.

Try Subtext in your browserDownload on the App StoreGet it on Google Play

What should you paste from a work chat?

Check your employer’s rule on AI tools before you paste work chat anywhere. Canada’s national cyber security centre warns that staff who type work content into these tools may leak personal and confidential commercial information outside approved frameworks¹⁷. Subtext’s terms ask you to confirm you may use it for work. Paste only the lines the task needs, crop names and numbers, and leave out anything about a colleague’s health, money or private life. Nothing you paste into Subtext is used to train an AI model, though the text does go to the AI providers its privacy policy names. A Subtext conversation deletes itself five days after you last use it, or 90 days if you pin it, and notes you save to Memory stay until you delete them¹⁸. Other tools handle your text differently, and which AI writing assistants train on your messages sets out how.

The order to do it in

Decide what you are asking for or refusing, and write it as one sentence. If you cannot, the message is not ready.

Add the reason the reader needs, a time you will hold and one real way out. Cut the sentence that scores a point against them and keep the deadline.

Check the place and the hour. Name the person who owns the task, correct the fact in a channel and leave the person out of it, and say whether it can wait until morning.

Read it aloud in the voice you fear it has. If it still carries an edge, or the thread keeps going round without an answer, call. If a tool helped, check that the ask and the time survived and that every detail is one you can stand behind.

If they answer sharply, I would stay level, reply to the substance in one line, and call if it goes round again.

None of this makes a colleague answer, or read your message kindly on a bad day. The aim is that whatever still stings is the thing you meant to ask for.

Questions people ask

Is a short reply like “noted” or “ok.” rude? It can be, for some readers. In lab studies a final full stop made one-word replies read less sincere⁶ and slightly more negative⁷. Both replies are ambiguous, so saying what you mean, as in “Thanks, I’ve read this”, leaves less to guess. If you are the reader and it matters, ask what they meant.

How do I say no to extra work from my boss without sounding difficult? Say what you can do and name the trade, as in the competitor summary above. That keeps the no and gives the person who owns the priorities a real choice. How to say no to extra work covers the research on refusing.

Can an AI rewrite my work message so it doesn’t sound passive aggressive? Subtext marks wording that could read that way and offers versions for a draft you wrote. A tool does not know the time you can hold, and in the run above one version added a soft “next few days” you had not given, so check that the ask and the deadline survived before you send it.

What if they still don’t answer? Ask once more with the same four parts, in a direct message or a short call. Silence in chat is not necessarily rudeness, since in a 2000 study of early instant messaging, which predates Slack, most of the 20 workers felt they could ignore a message without offending the sender¹¹. For ordinary blocked work, I would tell your colleague first that you are taking it to whoever sets the priorities. Tell them what is blocked and since when, and leave out what your colleague did. If the situation involves harassment, safety or discrimination, I would skip those steps and start with your employer’s own policy. If the silence happens every time, giving critical feedback to a coworker in writing is the next piece.

What if this is not a wording problem? If a colleague threatens you, demeans you or keeps contacting you after you asked them to stop, a better message will not fix it. Use your employer’s reporting process, or talk to whoever handles complaints at work, and if you are in danger, call your local emergency number. Please do not paste those conversations into Subtext or any other AI tool.

Sources

Numbered in the order they appear above. Where a figure could not be verified against the primary source, the text says so. Checked 6 November 2026.

Kruger, J., Epley, N., Parker, J., and Ng, Z.-W. (2005). Egocentrism Over E-Mail: Can We Communicate as Well as We Think? Journal of Personality and Social Psychology, 89(6), 925 to 936. Five experiments with Cornell and University of Illinois students, all by email and two also by voice, with statements the students wrote or chose from lists. Study 3: 154 pairs, about a third of them by email, 70.4 per cent against 88.8 across email, voice and face to face, and 62.8 against 88.0 by email alone; receivers picked from a list of four tones. Study 4: 54 students, half reading aloud as meant and half in the opposite tone, the read-aloud experiment. Study 5: 58 students, humour. Supported by a University of Illinois Board of Trustees grant and NSF grant SES-0241544.
Pollmann, M. M. H., and Roos, C. A. (2025). “I get u”. People correctly interpret the tone of text messages and emails. Computers in Human Behavior Reports, 18, article 100689. Study 1: text messages, 347 receivers and 171 senders. Study 2: work or education emails, 361 receivers and 61 senders. Real messages between people who knew each other; positive or negative tone only. Only the abstract was read. Recent and not yet replicated.
Giurge, L. M., and Bohns, V. K. (2021). “You don’t need to answer right away!” Receivers overestimate how quickly senders expect responses to non-urgent work emails. Organizational Behavior and Human Decision Processes, 167, 114 to 128. Eight preregistered studies, 4,004 participants in all: two recall studies (Study 1, Spanish public-sector employees, and Supplemental Study 1A) and six built on imagined scenarios, all about email, not chat.
Andersson, L. M., and Pearson, C. M. (1999). Tit for Tat? The Spiraling Effect of Incivility in the Workplace. Academy of Management Review, 24(3), 452 to 471. A conceptual paper with research propositions and no sample. Definition quoted from p. 457 as later papers quote it, and the original page was not opened.
Garant, J., Hausdorf, P. A., González-Morales, M. G., and Gill, H. (2025). Understanding workplace incivility dynamics from the perspective of conflict mediators: a qualitative study. Frontiers in Organizational Psychology, 3, article 1683369. Interviews with nine workplace mediators (five women, four men) about parties they had mediated; the parties themselves were not interviewed. Funded by a Social Sciences and Humanities Research Council of Canada grant.
Gunraj, D. N., Drumm-Hewitt, A. M., Dashow, E. M., Upadhyay, S. S. N., and Klin, C. M. (2016). Texting insincerely: The role of the period in text messaging. Computers in Human Behavior, 55, 1067 to 1075. 126 Binghamton undergraduates read 16 invented exchanges ending in a one-word reply, shown as text messages or as handwritten notes; the sample and design come from the university’s news release, the finding from the abstract. It measured sincerity, not rudeness.
Kemp, N., Kovacic, R., and Beyersmann, E. (2025). Is the period really “pissed”? The effect of punctuation and message length on perceptions in digital communication. Telematics and Informatics, 97, article 102241. 200 Australian undergraduates, first language English, rated how a fictitious friend felt in 30 invented exchanges; a final full stop lowered ratings at one word and at three to four words, by 0.11 to 0.37 of a point on a seven-point scale, and not significantly at six to eight words.
Glikson, E., Cheshin, A., and van Kleef, G. A. (2018). The Dark Side of a Smiley: Effects of Smiling Emoticons on Virtual First Impressions. Social Psychological and Personality Science, 9(5), 614 to 625. Three experiments on first impressions of an unfamiliar sender’s work email: 203 Amsterdam undergraduates, 90 online participants of 29 nationalities and 85 US online workers; not established colleagues and not chat. Funded by a Netherlands Organisation for Scientific Research grant.
Lai, L., and Mayiwar, L. (2023). The Dark Versus Bright Side of a Smiley: A Preregistered Replication of Experiment 3 in Glikson et al. (2018) “The Dark Side of a Smiley”. Collabra: Psychology, 9(1), article 90195. Preregistered replication with 847 participants against the original 85; smileys lowered perceived competence and raised perceived warmth whatever the level of formality. The publisher page was behind a bot check, so only the abstract was read.
Courtice, E. L., Lawrence, M., Collin, C. A., and Boutet, I. (2026). Emojis at Work: The Effects of Emoji Use on Perceptions of Competence and Appropriateness. Collabra: Psychology, 12(1), article 147309. Experiment in which 243 adults (134 men, 109 women) read simulated workplace instant messages that varied in tone, emoji and sender gender; only the abstract was read.
Nardi, B. A., Whittaker, S., and Bradner, E. (2000). Interaction and Outeraction: Instant Messaging in Action. Proceedings of the 2000 ACM Conference on Computer Supported Cooperative Work (CSCW), 79 to 88. Interviews and observations of 20 workers (seven at a telecommunications company, twelve at an internet company of about 700 employees, one contractor) using early one-to-one instant messaging; two of the three authors at AT&T Labs; it predates Slack.
Slack. Set your Slack status and availability. Slack Help Center. Undated; checked 6 November 2026. Describes the dot beside a name as availability, active when Slack is open and away after a period of inactivity.
Wolfe, J., Shanmugaraj, N., and Sipe, J. (2016). Grammatical Versus Pragmatic Error: Employer Perceptions of Nonnative and Native English Speakers. Business and Professional Communication Quarterly, 79(4), 397 to 415. 169 businesspeople commented on three versions of an email with different kinds of error; one email task, not chat; only the abstract was read.
Vignovic, J. A., and Thompson, L. F. (2010). Computer-mediated cross-cultural collaboration: Attributing communication errors to the person versus the situation. Journal of Applied Psychology, 95(2), 265 to 276. An email experiment on spelling and grammar errors and on etiquette violations, defined as short messages lacking a conversational tone; only the abstract was read.
Zajechowski, M., for Preply (2023, updated 3 July 2024). Survey: 83% of Americans hit by passive-aggressive emails from coworkers. Survey of 1,006 US adults aged 18 to 76, fielded 22 June 2023, by a language-tutoring company; sampling and weighting are not published. A vendor survey of self-reported perception, not evidence of effect.
Hohenstein, J., Kizilcec, R. F., DiFranzo, D., Aghajari, Z., Mieczkowski, H., Levy, K., Naaman, M., Hancock, J. T., and Jung, M. F. (2023). Artificial intelligence in communication impacts language and social relationships. Scientific Reports, 13, 5487. Two randomised experiments with Mechanical Turk workers, 219 and 291 pairs, in short chats about a policy issue. A Publisher Correction of 3 October 2023 added the omitted funding statement, NSF grant CHS 1901151.
Canadian Centre for Cyber Security. The threat from large language model text generators. Guidance, page last modified 17 January 2024; checked 6 November 2026. Names the risk that staff who type work content into such tools may leak personal information and confidential commercial information outside approved security and policy frameworks.
Subtext. Terms and privacy. Privacy policy and terms last updated 22 October 2026. First party.