ArticlesWriting tools

Do my English texts sound polite? What a non-native writer can and cannot check

Why a correct English message can still read as abrupt, why no single polite English exists, ten messages rewritten, and the part no tool can see.

By Samet Durgun · Co-founder of Subtext · 39 min read

In short

Correct grammar doesn’t settle it. In most of the studies cited here, mostly of student emails to lecturers, native readers judged abruptness on directness and softening more than on errors: a bare request, no greeting or closing, and no sign the writer knew they were asking for something. Those conventional phrases can be checked and learned. Whether this reader, in this relationship, will take your words as you meant them can only be estimated, by you or by a tool working from what you tell it.

Three things to check

  1. Does the request give the reader a way to say no, as “Could you send it when you have a chance?” does?
  2. In a first message to someone senior, keep a greeting and a closing.
  3. Watch for deadlines you didn’t mean to set, such as “as soon as possible”. If there’s no rush, say so.

Most of the research behind this is about emails from university students to lecturers. Everyday texting in a second language has barely been studied.

Is your English draft too direct?

Subtext marks where a request reads as an order and offers politer versions.

Your grammar can be correct and the message can still sound abrupt. In most of the studies behind this piece, native readers judged an email as abrupt because of how direct it was and how little softening it had, not because of its errors. So a grammar checker cannot tell a non-native English writer whether a message sounds polite. What marked the abrupt emails out was a request made too directly, with no softening words around it, no greeting or closing, and no sign that the writer knew they were asking for something. Those habits are often shaped by a first language, and they are invisible to anything that checks spelling and verb agreement. In the one experiment that varied spelling and grammar errors, most of the negative impressions the errors created were reduced once readers knew the sender was from another culture (covered below).

Most of the tools people reach for check grammar, fluency and formality. A smaller group reads tone. None of them knows who you are writing to unless you say, how big the favour is, or what counts as polite in that person’s culture, and that last layer is where the research places most of the damage.

There is also no single standard to aim at. British and American writers use please about equally often, for different jobs. Indian English has well-documented forms of its own. A Slack thread between a German and a Brazilian colleague probably runs on the team’s habits more than on either country’s, though I found no study of that I could check. So the useful question is whether your message sounds natural and polite to the person reading it, for what you meant. Whether it sounds native matters much less.

I co-founded Subtext, a tone checker, so I have an interest in how the last four sections read. They are where I describe what it does, and before them it comes up four times in passing, once for what its terms say. The research before them comes from sources I opened myself, in some cases only the abstract or a page describing the study, as the Sources list says. The ten rewrites and the checklists are my own. For the update of 13 October 2026 I merged two AI research reports into the original piece, traced every claim from them back to its source, and dropped what did not hold up. The first of those four sections shows what Subtext flagged in the made-up request “Send me the report until tomorrow.” and the versions it offered.

Four different things can be wrong with a message

Grammar, naturalness, register and politeness are four separate checks, and a message can pass one and fail the next.

  • Grammar is whether the forms are right: tense, articles, word order, spelling. “Sorry for late” fails it.
  • Naturalness is whether people who use English in that setting would phrase it that way. “I have another plan” is clear and slightly off, because English has settled on “I already have plans”. These fixed combinations differ between Englishes, which is why naturalness always means natural to someone.
  • Register is whether the formality fits the channel and the relationship. “I would like to inform you that I will be late” is right for a formal letter and odd in a group chat.
  • Politeness is whether the message handles the relationship the way you meant: how much you are asking, how much choice you leave, how much warmth you show. “Send me the file today” passes the first three checks and can still fail this one.

Grammar checkers are good at the first. Rewriters help with the second and third. The fourth depends on who is reading, and most of the research below is about it.

Why a correct English message can still read as rude

A message can be grammatically perfect and still fail, because politeness in English lives mostly in the words around the request. Economidou-Kogetsidis studied 200 email requests that Greek Cypriot university students sent to faculty at a university in Cyprus that teaches in English, one email per student, sent to 11 lecturers over 18 months1. In 75 of the 82 requests for information, about 91%, the students asked directly. Only 16.5% opened with a greeting such as “Hello” or “Good morning”. Most of the rest opened with a form of address alone, such as “Dear” and a name, and 14.5% with nothing. Only 23% ended with a sign-off such as “Best” or “Sincerely”, though 70% did say thanks. The emails were marked by “significant directness”, especially in requests for information, “an absence of lexical/phrasal downgraders”, omitted greetings and closings, and forms of address the author judged inappropriate or unacceptable. Her argument is that such emails can be read as impolite because “they appear to give the faculty no choice in complying with the request and fail to acknowledge the imposition involved”1. The paper’s title is one of the lines from the data, “Please answer me as soon as possible”, and it shows the pattern in seven words. The writer means to be courteous, uses the word please, and still produces a sentence that sets the reader a deadline.

A downgrader is a small word that lowers the pressure. “I was wondering whether”, “if you have a moment”, “would it be at all possible”. Many British and American writers reach for these. Writers coming from languages where directness is the polite default may skip them and add a please instead. In a 2013 study of 4,353 Wikipedia requests, each rated by five US annotators, Danescu-Niculescu-Mizil and colleagues found that a please in the middle of a sentence averaged +0.49 on their politeness scale, while a request that opened with “Please” averaged -0.302. They read an opening please as a direct strategy that can seem insincere. The study does not report the writers’ first language, so it shows how US raters hear a leading please, not who writes one.

Native speakers write directly too, with more ways to soften

Directness is common among native writers as well, and what separates the groups is how much softening they add. Biesenbach-Lucas compared email requests to professors from graduate students who were native and non-native speakers of English at an American university3. She analysed 533 emails that TESOL graduate students sent to a single professor over six semesters, 382 from native and 151 from non-native speakers, and every non-native writer came from an Asian background. Both groups wrote more direct requests and hints than the conventionally indirect forms found in older speech act studies, so the drift is not confined to learners. But native speakers “demonstrate greater resources in creating e-polite messages to their professors than nonnative speakers”3. They had more ways to soften, and they used them more. The groups are unequal in size. The study lists the single professor and the narrow sample of non-native speakers (all from Asian language backgrounds, all in one degree programme) as limitations, so it describes one setting and does not settle the question.

What native readers actually hold against you

The 2011 study also asked 24 British lecturers from 12 UK universities to rate six of the student emails1. Their lowest politeness rating and highest abruptness rating (1.38 and 4.67, on scales of five) went to an email that put a please in front of an imperative and had no greeting, no closing and no thanks. Two other emails built on the same please and imperative were rated more polite, and the author puts the difference down to the thank you before the sign-off.

Native readers judge the sender as well as the sentence, and they judge the register more harshly than the grammar. In a later perception study, Economidou-Kogetsidis showed the same direct, unmodified student emails to Greek Cypriot learners and to native English lecturers4. The two groups had “significantly mismatching perceptions of what is appropriate”. In most cases the native lecturers rated the same email as “significantly more abrupt, less polite and as not adequately acknowledging the imposition involved”, and they rated the personality of the sender “significantly less favorably” than the learners did4.

Hendriks ran two online experiments in which native English speakers read email requests written by Dutch learners, with the softening words varied. The 110 raters in the first and 158 in the second rated comprehensibility and the sender’s personality5. Underusing those softening words “had a negative effect on participants’ evaluation of the personality of the sender of the e-mail”5. The message was understood either way. The effect was on how agreeable the sender seemed, it came from one phrase, and it was small, about a third of a point on a scale of five points.

A 2023 study by the same lead author found the penalty narrower still. It gave 120 Dutch and 131 British readers one of four versions of a request email from a colleague6. The curt version (“Can you send me…”) made the sender seem bossier. But a Dutch writer’s curt email was also rated more authoritative and drew more willingness to comply, while an English writer’s curt email was rated less competent. The authors say leaving out softening words “may have a limited but important impact”. In that study, readers held the native writer to the softening convention and gave the Dutch writer more room, which is worth remembering before assuming every reader is grading your English.

Readers forgive grammar more readily than tone

Readers forgive grammar errors more readily than a curt tone, and Vignovic and Thompson tested whether knowing the sender is from another culture buys any forgiveness7. Readers judged an email sender’s conscientiousness, intelligence, agreeableness and trustworthiness, among other traits, after reading messages with spelling and grammar errors, or messages that broke etiquette norms, which the authors defined as short messages lacking a conversational tone. Grammar errors produced negative impressions, and most of those impressions were reduced when readers learned the sender was from a different culture. Etiquette violations were different. The negative attributions “were not significantly mitigated by knowledge that the e-mail sender was from a foreign culture”7. A reader will largely excuse your grammar when they know English is your second language. They will not excuse a curt tone in the same way. The paper frames this as attribution, with grammar errors put down to the situation and a curt tone put down to the person.

Whose English sets the standard?

Most of the studies above put British or American readers in the reader’s chair. That makes them good evidence about those readers and weak evidence about English in general, because the varieties differ in exactly the places a message gets judged.

Take please. Earlier studies of requests found it about twice as often in British directives as in American ones8. Murphy went to web English from both countries, about 386 million words each, and found the word itself almost equally common, with 1.5% more in the British half8. What differed was the job it did. British please was more often routine, the please of instructions and notices. American please turned up more often where more was at stake between writer and reader, and the dismissive kind, as in “oh, please”, was more than twice as frequent in the American sample, though the counts are small, 20 against 9 in 800 uses each. The same word can read as a habit to one reader and as a pointed choice to another.

Indian and Sri Lankan English have been measured too. Degenhardt compared requests in recorded speech from Britain, India and Sri Lanka and found Indian English about as direct as British English overall, 45.8% of requests made directly against 44.8%, with Sri Lankan English more direct at 56.5%9. The differences were matters of degree, and they depended more on the kind of sentence than on the country. Xie’s comparison of politeness formulas in British, Australian and Hong Kong web English found Hong Kong English the most tentative, mostly because of please, and Australian English ahead of British on both deference and friendly solidarity forms10. Sorry turned up about 87 times per million words in the British sample, 77 in the Australian and 40 in the Hong Kong one.

Some forms belong mainly to one variety. Oxford’s learner’s dictionary marks “do the needful”, meaning to do what is necessary, as especially Indian English11. In that English it is an everyday phrase. A client in Ohio may not know what the needful refers to, and that is a reason to name the action for that reader. The phrase itself is fine.

A great deal of everyday English is also written in groups where it is nobody’s first language: the Slack channel in a Berlin startup, the supplier thread between Istanbul and Seoul. There the reader’s expectations come from the team. I did not find a study of politeness in that kind of everyday chat that I could check properly, so I will leave it at a rule of thumb: ask what your actual reader is used to before you ask what a native speaker would write.

Phrase problems and situation problems

Thomas drew the line that explains why some of this can be checked and some cannot12. She called the inability to understand “what is meant by what is said” pragmatic failure, and split it in two. Pragmalinguistic failure is a matter of “highly conventionalized usage which can be taught quite straightforwardly as ‘part of the grammar’”. Sociopragmatic failure “is much more difficult to deal with, since it involves the student’s system of beliefs as much as his/her knowledge of the language”12.

In practice the first kind is the missing “I was wondering if” or the “please answer me as soon as possible”. It is a phrase problem, and a tool or a phrasebook can catch it. The second kind is deciding that a request of one line to a senior colleague needs no softening at all, because in your first language it would not. No phrase fixes that, because the words are fine and the situation was misread.

Learners also notice these two kinds of error differently depending on where they live. Bardovi-Harlig and Dörnyei asked English learners to judge utterances that were either grammatically wrong or contextually inappropriate13. Learners studying English in a country where it is spoken rated the pragmatic errors as more severe than the grammatical ones. Learners studying English in their home country rated the grammatical errors as more severe13. A learner who studied English in their home country therefore tends to rank grammar errors as the worse kind, while the studies above found native readers reacting more to directness and softening.

Three questions that decide how much softening a message needs

Brown and Levinson’s Politeness: Some Universals in Language Usage, the best-known model of politeness, gives a practical handle on the second kind of problem14. Requests, refusals and complaints put “face” at risk, the public self-image everyone wants left intact. Positive face is the wish to be liked and approved of. Negative face is the wish not to be imposed on. How much a message threatens either depends on three questions: how well do you know the reader, how much power do they have over you, and how big is the thing you are asking? The bigger the total, the more softening, warmth or indirectness a writer reaches for.

The three questions travel better than the theory behind them. Brown and Levinson’s theory of face has been argued over for decades, mostly by researchers writing about languages other than English. Matsumoto argued that in Japanese the main concern is your position relative to others, so honorifics mark the relationship even when nothing is being imposed15. Ide made a related case that much Japanese politeness is discernment, choosing the form the situation calls for, more than a strategy chosen to save face16. Mao argued that the Chinese ideas of face, mianzi and lian, do not map onto positive and negative face17. Fukada and Asato answered that the model can explain Japanese honorifics after all, because a higher-status person pushes power and distance up18. The argument is still open. For a message, the safe reading is that distance, power and the size of the ask matter nearly everywhere, and the words that show you noticed them change from language to language.

What instruction and higher proficiency help with

Instruction and higher proficiency both help with the conventional phrases, and neither closes the whole gap. Usó-Juan collected 110 naturally occurring email requests that Spanish learners of English as a foreign language sent to three faculty members, before and after a period of strategy instruction, and found the requests more appropriate afterwards19. The abstract reports a before and after comparison, with no control group or delayed test mentioned, so it shows the phrases can be taught and not that the gain lasts. Qin, Jia and Ren analysed 306 request emails from 32 teachers who speak English and 121 learners whose first language was Chinese, French or Spanish20. The learners’ first language had a significant effect on their writing. Learners with higher proficiency used more indirect request strategies, but they had not developed the pragmatic competence to adjust their usage across written contexts. A more advanced learner uses more indirect requests and still does not yet adapt to the situation, which resembles Thomas’s second category12.

The larger picture is encouraging, with a catch. A meta-analysis of 29 instruction studies covering 1,898 learners found a large average effect from teaching pragmatics (g = 1.66), and the biggest gains showed up on written discourse completion tasks, the describe-a-scenario tests the field relies on21. Gains on a test of that kind tell you less about messages to real people. Nguyen followed 32 Vietnamese English majors, 13 of whom had six hours of explicit teaching on email requests, and found their advantage still there eight months later on elicited emails22. Greetings, indirect requests and leaving the reader a choice improved and stayed improved. Closings improved at first and had slipped most of the way back by eight months, and giving reasons, disarming phrases and compliments did not change at all.

Most of this research is about emails to professors

The evidence on requests is good because researchers could collect real emails, and almost all of those emails were sent by university students to lecturers, a relationship with a large power gap and fixed conventions. Refusals and apologies have mostly been studied through discourse completion tasks, where people write what they would say in a described situation. The classic study of refusals by Japanese learners of English worked that way, with 60 participants in three groups of 20 responding to written scenarios23. Golato compared answers to a written task with recordings of the same kind of moment in real talk and found they captured different things: the recordings showed how people respond in the moment, and the written answers reflected their accumulated experience of the language24. Everyday texting in a second language, with friends, landlords or colleagues, has barely been studied.

The field also warns against the shortcut this kind of research invites. House and Kádár put it bluntly: “By degrading learners to cultural robots, the researcher precludes unearthing fine-tuned pragmatic conventions”25. Your first language may shape which English phrases feel natural to you. People who share a first language still write in very different ways, and nothing in these studies predicts how any one person writes.

Why you cannot hear it yourself

Rereading your own message may not catch this, for two reasons. The first is the general one, covered in does this text sound rude. Kruger and colleagues ran five experiments on tone over email and their results suggest that people tend to believe they communicate over email more effectively than they do. Their fourth and fifth experiments suggest the cause is that writers “hear” a statement as they meant it, which makes it hard to see that the audience may not26. A 2025 study of everyday text messages (347 receivers, 171 senders) and work or education emails (361 receivers, 61 senders) found no sign of misread positive or negative tone27. It rated positive or negative tone, not politeness, so it leaves this section’s question open. The second reason is an inference from the studies above, not a finding, and is specific to writing in a second language. The silent voice you hear is reading with your first language’s politeness rules. A request that is perfectly courteous in Greek or Dutch sounds courteous to you in English too, because the judgement is being made by the same ear.

This is why the perception gap in the second Economidou-Kogetsidis study is so wide4. The learners rated their own kind of email as appropriate because, by the standards they had grown up with, it was. The lecturers rated it as abrupt because, by theirs, it was. Each group read the email accurately by its own standards.

So asking whether your English sounds polite means asking how it sounds to a reader whose expectations you cannot check, which is what a second reader is for, whether that is a fluent friend or a tool such as Subtext. The reader’s side of the same gap, working out what a message you received was meant to say, is in what does this text mean.

Second-language writing anxiety, measured

Second-language writing anxiety has been measured for twenty years. Cheng built a scale for it from the reports of 65 learners and validated it with 421 students majoring in English as a foreign language at seven colleges in Taiwan, splitting the anxiety into physical symptoms, worried thoughts and avoiding writing altogether28. A 2022 meta-analysis pooled 84 effect sizes from 22 studies and found writing anxiety moderately linked to weaker writing (a correlation of about 0.3) and to lower confidence in one’s own writing (about 0.4)29. Both are correlations, mostly from students writing for class. They do not show that anxiety causes worse writing, and they say nothing direct about a two-line text to a landlord. The everyday version is in why texting gives you anxiety.

Ten messages, before and after

Each pair below starts with a message that is grammatical, or close, and likely to read as blunter or stiffer than its writer meant. The second version is one option among several. A version for your reader may be warmer, shorter or more formal, and in some relationships the first version is fine.

1. Asking a colleague for something, on Slack

Before: “Please send me the final numbers today. I need them for the meeting.”

After: “Could you send me the final numbers today? I need them for the 4pm meeting. Thanks.”

To the US raters in the Wikipedia study, a request that opened with please read as less polite than neutral2. The question form leaves room for a no. Your reason was already doing good work, so it stays.

2. Following up with a landlord

Before: “Why did you not answer my email? The heating is still not working. Please fix it as soon as possible.”

After: “Hi Alex, I’m following up on my email from Tuesday about the heating. It still isn’t working. Could you let me know when someone can come and look at it this week? Thanks.”

“Why did you not” reads as an accusation, and “as soon as possible” is the phrase the 2011 paper took its title from1. The rewrite keeps the pressure through a date and a concrete next step. If the repair is the landlord’s legal duty you can be firmer, and polite wording never means giving up the claim.

3. Replying to a manager

Before: “OK. I will do it.”

After: “Sure, happy to take it. I’ll go through the notes before Thursday.”

Two short sentences with full stops and no contractions can read as reluctant in a chat. The etiquette violations readers did not excuse in Vignovic and Thompson’s study were of this kind, short messages without a conversational tone7. Whether the full stop makes it worse is covered in does a full stop make a text sound angry.

4. Declining an invitation from someone you just met

Before: “No, I cannot come. I have another plan.”

After: “Thanks so much for the invite. I’m sorry, I already have plans on Saturday. I hope it’s a great party.”

“I already have plans” is the set phrase. Thanks, a short regret and a good wish are the usual cushion around an English no. Refusals have mostly been studied through elicited tasks, so treat that cushion as a habit worth knowing, not as a measured effect.

5. Asking a friend for a favour

Before: “Can you bring me to the station tomorrow?”

After: “Any chance you could give me a lift to the station tomorrow?”

“Bring me to the station” is understandable, and ordinary in Irish English, though many British and American readers would say take. “Give me a lift” is the British idiom and “give me a ride” the American one, and neither variety is the correct one. “Any chance” makes the favour easy to turn down.

6. Apologising to a client for a slow reply

Before: “Sorry for my late answer. I was very busy.”

After: “Sorry for the slow reply. Here are the answers to your questions.”

A reason helps when it tells the reader something. “I was very busy” tells a client only that they came second, so it goes, and the reader mainly wants the answers. One clear sentence of apology is usually enough at work. More on this in how to apologise over text.

7. Chasing a refund with customer service

Before: “I am still waiting for my money. When will you send it?”

After: “Hello, I’m writing about the refund for order 4812, which I returned on 2 October. Could you tell me when it will be processed? Thank you.”

With customer service, politeness mostly means making the agent’s job easy: the order number, the date and one clear question. Specifics do more here than warmth.

8. Saying no to extra work

Before: “I don’t have time for this.”

After: “I’d like to help, but the launch has taken my whole week. Could I look at it on Monday?”

The no becomes a not this week. A reason and an alternative show you took the request seriously. Research on when declining costs you is in how to say no to extra work.

9. Asking an international client to fix an invoice

Before: “Kindly do the needful.”

After: “Could you please correct the VAT number on the invoice and send it back?”

“Do the needful” is established in Indian English, and Oxford’s learner’s dictionary labels it especially Indian English11. A reader outside it may not know what the needful is, so the rewrite names the action. If your reader writes Indian English too, the original works.

10. Telling a group chat you are late

Before: “Dear all, I would like to inform you that I will be late to the meeting.”

After: “Running 10 minutes late, start without me.”

The original is correct and natural in a formal letter. In a chat where everyone writes in fragments it reads stiff, and matching the channel fixes it.

Checks that don’t need an app

Four checks catch a good share of this, and each has a blind spot.

  • A phrase bank, such as a university’s guide to email, gives you tested conventional phrases. It cannot tell you whether a phrase fits this reader, and many are written for one variety and one formal register.
  • A fluent friend may know the reader and the setting. Their sense of polite is still their own, and you may not want to show them the message.
  • Reading it aloud catches missing words and a clipped rhythm. Reading it in the tone you are afraid of is the version with some evidence behind it, covered below.
  • Mirroring the other person, their length, greeting, punctuation and emoji, is the quickest local cue you have. It only works once they have written, and copying a manager’s terse style upward can still land badly.

What grammar and fluency tools check, in their own words

The large writing tools mostly promise correctness and fluency, and some promise formality. Each of the four descriptions below quotes the tool’s own page, so none of them is a feature I inferred.

LanguageTool’s homepage describes a checker for “grammatical errors and spelling mistakes across languages”, says it “masters more than 30 languages and dialects”, and offers a paraphraser that rephrases “to be more formal, fluent, simple or concise”30. Its product menu says the checker “helps you find the right tone”; nothing on the page says it detects politeness.

DeepL Write, by its own blog, supports English, German, French, Spanish, Italian, Portuguese, Chinese (Simplified), Japanese and Korean, and, for French and Spanish at least in that post, lets you pick a style (Business, Academic, Simple, Casual) and a tone (Friendly, Diplomatic, Confident, Enthusiastic)31. Both are settings you apply, and the tool does not detect whether your draft was diplomatic.

Wordtune promises to “Switch between casual and formal tones with the click of a button”, says it “can only write in English”, and claims its translation feature “will make you sound like a native-English speaker with just a click”32. The first two are checkable. The third is marketing.

QuillBot’s paraphraser, per its own blog, offers two free modes (Standard and Fluency) and nine preset modes for paying users, works “in over 25 languages”, and is pitched at people writing “in a first, second, or third language”33. Again, the modes are instructions you give it.

Two more kinds of tool get used for this without being built for it, and here I have no page to quote, only my own observation. A translator tends to keep the register you wrote in, so a direct request in your first language usually arrives as a direct request in English. General chatbots such as ChatGPT can critique a message well if you tell them who the reader is and what you want, and they work only from what you tell them. Several of the large ones train on what you paste unless you switch that off, as a check of thirteen writing assistants’ policies found.

All of these can make a sentence more fluent and more formal. Formal and polite are different qualities, and the research above is about the difference. A formal sentence that still gives the reader no choice is still the kind of email the 2011 paper took its title from.

What the tone tools check

Two tools claim to read tone before you send and a third rewrites towards it, and their claims are narrower than the question you are asking. A fourth, QuillBot, scores tone after a paraphrase.

Grammarly’s tone page shows a set of labels for the overall tone of a text, such as Formal, Confident, Accusatory and Worried, with no statement about languages other than English, about limits, or about who the reader is34; what the detector reads is covered in AI tone detection. How that compares with a reader built for messages is in Subtext against Grammarly.

Tonely, on its App Store listing, is “the world’s first On device Ethical AI-powered iMessage extension and keyboard designed to help you understand how your words may come across before you press send”, runs entirely on the device, lists English as its only language, and is free to download with a subscription of $9.99 a month or $99.99 a year35. Processing on the device is a privacy advantage, and English only is a limit for anyone who also messages in their first language.

ConfiText’s homepage says it “rewrites your messages so they come across clear, confident, and exactly how you meant them”36. It names no languages and makes no claim about reading how a message will land before rewriting it.

QuillBot’s help page, last updated on 10 June 2026, describes Premium tone settings that “help you see how the tone of your text changes after paraphrasing”37. A panel charts the original and the paraphrase on five scales: Casual to Formal, Unfriendly to Friendly, Wordy to Concise, Complex to Simple and Unclear to Clear. Those score the text itself, and the page says nothing about politeness or how a reader will judge it.

A wider comparison of 17 tools, grouped by what each one does first, is in the tone checker comparison.

What each kind of tool checks

Kind of tool What it checks What it cannot know
Grammar checker Spelling, grammar, punctuation, some style Whether the message suits this reader
Paraphraser or rewriter Fluency and formality, in modes you pick How your original draft reads
Translator Meaning across languages Whether the politeness carried over
General chatbot Whatever you ask, with the context you give Anything you leave out
Tone checker How the draft is likely to read Your relationship with the reader, unless you tell it
Fluent friend Naturalness, register, sometimes the reader Whether their sense of polite matches the reader’s

Can a program rate politeness at all?

A program can rate the politeness of English requests close to a person’s accuracy within one site, and it is still behind across sites and across languages. Danescu-Niculescu-Mizil and colleagues built a politeness classifier from a corpus of annotated requests, using features that stand in for indirection, deference and impersonalisation, and reported that it “achieves close to human performance and is effective across domains”2. The base is narrower than the quote. The 2013 paper had five US annotators rate each of 10,957 English requests (4,353 from Wikipedia, 6,604 from Stack Exchange) and tested only on the most and least polite quarter of each set. Within one site the classifier scored 83.79% on Wikipedia and 78.19% on Stack Exchange, against human accuracy of 86.72% and 80.89%.

Trained on Wikipedia and tested on Stack Exchange it scored 67.53% against 80.89% for people, and in the reverse direction 75.43% against 86.72%. When Srinivasan and Choi built TyDiP, a politeness dataset with 500 annotated examples in each of nine typologically diverse languages, multilingual models carried their English training over to the other languages reasonably well in the authors’ judgement, “yet fall short of estimated human accuracy significantly”38. The same paper also asked whether each English politeness strategy keeps its effect when mapped into other languages, and whether formality and politeness line up. Neither does so every time.

Can a tone tool judge a non-native writer’s English?

A tone reading of your English is an estimate, because neither classifier study cited above reports whether its requests came from writers using a second language. The tools above that read tone are doing a job of this kind, whatever their internals. The published evidence says the models tested so far are good at English and measurably behind people in other languages, and that the strategies which make an English sentence polite do not transfer one for one. Even where a model is strongest, it is modelling a generic reader.

The nearest test is adjacent evidence, not a tone test. Liang and colleagues ran seven detectors of AI text on 91 TOEFL essays written by humans and 88 essays by US eighth graders in 202339. The detectors were nearly perfect on the eighth graders’ essays and misclassified more than half of the TOEFL essays as written by AI, an average false positive rate of 61.3%. It says nothing about tone tools. Rereading the message as its recipient would, whenever a tone score on your own English worries you, is a sensible precaution regardless.

What AI feedback does to a second-language writer’s English

Research on generative AI giving feedback to second-language writers has grown fast, and it points two ways. A scoping review of 51 studies found gains in writing quality reported in 17 of them, error detection more reliable in English than in Greek or Chinese, and only about half the studies reporting the prompts they used40. AI and teacher feedback together did better than either alone.

The fixes are real and shallow in a specific way. Chon and Shin had 40 Korean English-education students write, let ChatGPT proofread, then revise with its help41. Direct proofreading resolved 86% of the original errors and the students’ own AI-supported revisions 77%. The errors that survived were mostly word choice and sentence structure, the ones that needed a judgement about context, and the authors describe the gains as short-term. In a case study of four Chinese learners, the two who got 11 or 12 corrections per task revised correctly about three-quarters of the time, and the two who got more than 22 managed about 60%42. It is tempting to blame the sheer number of corrections, but the two who got more were also the weaker writers, which is how the authors explain the gap, and four learners cannot separate the two.

Voice is a cost too. In Agarwal, Naaman and Vashistha’s experiment with 118 writers in India and the US, AI autocomplete gave Americans bigger efficiency gains and pulled Indian writers towards American styles43. A 2026 preprint by Liu and Zhu had four AI models revise 106 essays by Chinese university students44. Every model fixed the grammar and kept the meaning, and every model cut roughly three-quarters or more of the pronouns that address the reader, such as “we”, “you” and “our”. Rhetorical questions fell too, by a quarter to more than four-fifths depending on the model. The models had been told to meet the standards of elite American and British academic writing, which the authors list as a limitation, so part of the flattening is what was asked for. It is not peer reviewed, uses no statistical tests and involved no writers using AI, so it shows what the models do to a finished text. What happens to writers who use them is a separate question. The wider evidence on rewrites and voice is in does AI make your texts sound like a robot.

None of these studies tested whether using a tone or rewriting tool makes someone better at writing messages without it. That question is still open.

What no tool can check

No tool knows the one thing the research says matters most, which is who is reading, beyond what you tell it. Every perception study above got its result by holding the message constant and changing the reader. The same email was fine to a learner and abrupt to a lecturer4. The same grammar error was judged less harshly when the reader knew the sender’s background and not otherwise7. A tool sees the message. It does not see the lecturer, or how heavy the favour is to ask.

Whether your relationship with this reader allows a bare request, whether a message of two lines to this person reads as efficient or as cold, and whether the imposition is small enough to skip the apology are all beliefs about the situation, which makes them Thomas’s second category again12. A tool can tell you the sentence is direct. It cannot tell you whether direct was the right call, and the field does not have that certainty to give. Rewriting tools share that limit, because most hand back new text without saying how the reader will take it, which the review of AI message rewriters found across the main kinds of app.

What a tool can do is read the draft as someone other than you. It can notice that a request has no softening, that a message has no greeting, that a line sets a deadline you did not mean to set. That catches the first category, the conventional phrases, which Thomas said is “fairly easy to overcome”12. For someone writing in a second language that covers most of the easy fixes, and it is the part your own ear is worst at.

How can you check your English yourself before you send it?

Start with the shape of the request. Does it give the reader a way to say no? The 2011 finding was that emails can read as impolite when they “give the faculty no choice”1. “Could you send it when you have a chance” does. “Send it by Friday” does not, and an opening “Please” is not a reliable fix either. Requests that began with “Please” averaged -0.30 on the politeness scale of the US raters in the Wikipedia study2.

Use a softener that changes how the request reads. In Hendriks’s experiments, swapping “can” for “could” changed nothing, and “I was wondering if” was the one modifier tied to a more agreeable sender, by a small margin. Adding “possibly” made no difference5.

Acknowledge the ask. One clause that shows you know you are asking, “I know you’re busy”, does the work the lecturers in the perception study felt was missing4.

Keep the greeting and closing in a first message. Their absence was one of the features the 2011 study found in the emails that read as rude1. Nobody expects them in an ongoing chat, but a first email to someone senior is read for them.

Watch your deadlines. “As soon as possible” was the phrase that gave the paper its title1. If you mean no rush, say so.

Then read it as if you were annoyed. The reason is in the article on why you cannot tell if your own text sounds rude. In Kruger and colleagues’ fourth experiment, 54 students read their statements aloud, half in the tone they meant and half in the opposite one, and only that second half stopped overestimating how well a partner would decode them26. It was one experiment, on sarcasm, so treat it as a precaution.

A checklist for the minute before you send:

  1. Who is reading, how well do you know them, and how much power do they have over you?
  2. How big is the ask? A bigger ask needs more cushioning.
  3. Does the reader have a way to say no?
  4. Is there a greeting and a closing that fit the channel?
  5. Did you give a reason for the deadline, the refusal or the delay?
  6. Does any line sound like an accusation when you read it annoyed?
  7. If they wrote first, does your message match their length and formality?
  8. Is there one sign of warmth, if warmth is what you feel?
  9. Are you apologising or explaining more than the situation needs?
  10. If a tool rewrote it, does it still sound like you and still say what you meant?

Pasting a thread means pasting someone else’s words

Any checker works better with context, and context usually means the other person’s messages. A screenshot carries their name, their words and sometimes their number, and they never agreed to it going anywhere. Crop out names and numbers you do not need, paste only your draft when the draft is enough, and keep work messages out of any tool your employer has not approved. Subtext’s terms put that responsibility on you. They note that screenshots and messages often contain someone else’s words, and that by submitting them you confirm you may lawfully do so45.

How much of this is settled

Claim Evidence behind it How strong
Grammar and politeness are separate problems Foundational theory and decades of learner research Strong
Direct, unsoftened emails to lecturers read as rude to British lecturers and other native English raters Real emails and rating studies, mostly Greek Cypriot and Dutch writers Moderate, for that setting
Readers excuse grammar errors more readily than a curt tone One experiment Limited
Readers always judge non-native writers more harshly One experiment found the reverse for curt emails Not supported
There is one British or American standard of polite English Corpus studies of several Englishes Contradicted
The conventional phrases can be taught A meta-analysis, one eight-month study and one before-and-after study, mostly on test tasks Moderate to strong
Findings on refusals and apologies predict everyday texting Mostly elicited tasks Weak
AI feedback fixes surface errors A scoping review and two small studies Moderate
AI suggestions pull writing towards American norms One experiment and one preprint Limited
Tone apps, Subtext included, make people better writers No study found None

What Subtext checks in a draft you are about to send

Subtext, the app I co-founded, names issues such as pressuring or overly demanding wording, which overlaps with part of the first category in Thomas’s split, the conventional slips. Its prompts do not list missing softening words or greetings as a check, though on the draft below it flagged a bare command for having no courtesy. It works in 20+ languages and answers in the language you write in46. I read the app’s backend prompts for what it returns; the on-screen marking is as the app shows it, and what it does with your data is as its privacy policy states it. What the app returns for each draft is described in full in what the Subtext app does.

For a message you are about to send, Subtext names up to three issues in your draft, most serious first, marks the words at fault, reads the main emotion, and offers up to three versions, each carrying a safe-to-send score, shown as a badge, for how likely it is to land as you meant. Each issue has a short explanation that quotes your own words where it points at a phrase, and pressuring, overly demanding, condescension and a robotic tone are among them. Each version fixes what was flagged while keeping your meaning and your register, and carries a sentence of its own naming what makes it land or what could be misread, whether its score is high or low. Subtext is paid after the first few analyses.

After the read, one-tap changes include Politer, Simpler, Warmer, Shorter, Add confidence and More casual. The More casual chip sends the instruction “Sound more natural”, which is the closest the app comes to the naturalness check at the top of this piece46. The score runs from 5 to 100. Subtext’s own FAQ says the score reads only the words in front of it, cannot know the person you are writing to, and should be treated “as a second reading rather than a measurement”46. That line predates Memory. With a person saved, the note behind the safe to send badge can draw on what you saved about them, and it still knows no more than those notes.

A made-up request of the kind you get by translating word for word shows the read, the versions and a one-tap change in order. On the draft “Send me the report until tomorrow.” Subtext marked the whole line and returned the tags Overly demanding and Clarity. The first note called the line “a bare command with a deadline and no courtesy”. The second said that “Until tomorrow” is unclear, because it could mean by the start or the end of tomorrow. Its closest version read “Could you send me the report by tomorrow, please?” One tap on Politer produced “Would it be possible for you to send me the report by tomorrow? I’d really appreciate it.”, marked safe to send. That is the app’s real output, unedited. The versions say “by tomorrow”, and nothing told the writer that “until” was the mistake. Neither version settled the question its own Clarity note raised, and the app said so, noting that “Tomorrow” still has no time of day.

Send me the report until tomorrow.

  • Overly demanding: "Send me the report until tomorrow." is a bare command with a deadline and no courtesy.
  • Clarity: "Until tomorrow" is unclear: it could mean by the start or the end of tomorrow.
  • Urgency: The message is curt and sets a deadline without any softening.

Deadlines are easier to meet when the ask is friendly and clear. Here are a few ways to say it.

  • Close to yours (Polite, safe to send): Could you send me the report by tomorrow, please? Keeps your ask and deadline but adds a polite "could you" and "please". Tomorrow still has no time of day.
  • Warmer (Appreciative, safe to send): Could you send me the report by tomorrow? Thanks a lot, it really helps me. The thank-you and the note that it helps you make the request feel collaborative rather than like an order.
  • Clear and friendly (Collaborative, safe to send): I'd love to have the report by the end of tomorrow. Does that work for you? If something gets in the way, just let me know and we'll figure it out. It clarifies "end of tomorrow", checks that it's workable and leaves room to flag problems, so it should land well.

I assumed "tomorrow" means end of day. Tell me if you need it earlier. Want it shorter, warmer, or more confident?

Politer

Here's a more diplomatic take on your selected draft.

  • More diplomatic (Courteous, safe to send): Would it be possible for you to send me the report by tomorrow? I'd really appreciate it. Asking "would it be possible" and adding appreciation makes the deadline feel like a request, so it should land well.

"Tomorrow" still has no time of day, so add one if it matters. Want it shorter or warmer?

Subtext also has a narrower mode, Proofread only, chosen at the bottom left of the input box, which checks spelling and grammar and leaves the wording alone. One example from 13 October 2026, with a made-up draft and the app’s real answer, shows where that line falls. The draft was “Hi Mr. Alvarez, thanks for you’re email. I have attach the signed form and will sent the rest untill friday. Best, Lena”. Subtext put no tag on it and wrote “Your message is polite and clear, it just needs a few small spelling and grammar fixes.” The version reads “Hi Mr. Alvarez, thanks for your email. I have attached the signed form and will send the rest until Friday. Best, Lena”. Five slips are gone. The sixth is the one from the example above, and it stayed: Subtext left “until Friday” in place and added that it is fine, though “by Friday” is “the more usual way to give a deadline.” So with Proofread only, read its note as well as its version.

The Subtext app showing the draft “Hi Mr. Alvarez, thanks for you’re email. I have attach the signed form and will sent the rest untill friday. Best, Lena” with the feeling word Courtesy and the marker Proofread only The corrected version, “Hi Mr. Alvarez, thanks for your email. I have attached the signed form and will send the rest until Friday. Best, Lena”, in a card headed Corrected, with the feeling word Professional, marked safe to send

Does Subtext ask who you are writing to?

Subtext can ask about the reader. The prompts tell it to read the message “the way a native speaker of its language would hear it”. For languages where politeness is grammatical, Japanese, Korean, Turkish and German among them, it asks about the relationship, or writes at the more polite level when it has to draft without an answer. English does not mark politeness in its grammar, so that rule does not apply to English drafts. In any language, when a draft pushes back on someone (a refusal, a boundary, a confrontation) and nothing says who it is for, Subtext asks who the recipient is before it drafts, while everyday messages get a draft straight away. How this reader, in this relationship, will take your words is still up to you, which is Thomas’s second category. No study cited here tested Subtext, or any tool above, on English written as a second language, so apart from the two drafts above, neither of which is a test of second-language English, what I describe is what its prompts say it does.

You can also tell Subtext who the reader is and have it kept. Since October 2026 Memory holds notes on the people you save.

What Subtext does with a received message

For a message you received, Subtext gives a summary in one line of what the other person wants from you and two to five action points, plus a reply if you ask for one, with no label on the sender’s tone, so the tone reading runs only on your own draft.

It knows your reader only as far as you tell it. Tell it once that your manager likes short messages and ask it to remember that, and Memory holds it as a note on how you write to that person, which you can read, edit or delete, and which stays until you delete it. It does not know your field’s conventions. It can flag a request that pressures or demands, one of the conventional slips, as in work messages that are firm without being cold. Use it as a second reader, and still reread the message as the recipient would.

Subtext sends the text you submit to the providers its privacy policy names, so it leaves your device. Nothing you submit is used to train an AI model, and the policy says none of those providers may train on it either45. A conversation is deleted from Subtext’s servers five days after you last used it, or 90 days if you pin it. On the provider side, the policy says Anthropic deletes inputs and outputs within 30 days and may keep a flagged request for up to 2 years and the related safety scores for up to 7 years, Deepgram (voice transcription), with Subtext’s opt-out set, keeps audio only as long as transcription takes, OpenAI keeps read-aloud requests for up to 30 days, and Subtext has no zero-retention arrangement with any of them45. A check of thirteen writing assistants’ policies, which also details the providers’ retention, found several tools above train on pasted text by default, including Grammarly and QuillBot on individual accounts and DeepL Write’s free tier.

What Subtext can and cannot do for a non-native writer

Subtext can:

  • read a draft and mark the words likely to land as pressuring, demanding, condescending or robotic
  • offer up to three versions and one-tap changes such as Warmer, Shorter and More casual
  • answer in the language you write in, in 20+ languages
  • ask who a message is for when the draft pushes back on someone

Subtext cannot:

  • guarantee a message suits a particular reader or variety of English
  • know your relationship beyond what you tell it
  • show that using it improves your own writing over time
  • point to an independent study of how accurate its reads are

Treat what Subtext says as a second reading of your draft. Try Subtext in your browserDownload on the App StoreGet it on Google Play

Sources

Numbered in the order they appear above. Every DOI was resolved against Crossref. Sources from the first version were opened on 4 and 5 October 2026, with the Kruger and Pollmann studies added and checked on 10 October 2026. Sources added in the update were checked on 12 and 13 October 2026, when the Subtext pages were also opened again. No retraction, correction or expression of concern was found for any study cited, from publisher records, Crossref and OpenAlex.

  1. Economidou-Kogetsidis, M. (2011). “Please answer me as soon as possible”: Pragmatic failure in non-native speakers’ e-mail requests to faculty. Journal of Pragmatics, 43(13), 3193 to 3215. 200 emails from 200 Greek Cypriot students to 11 faculty over 18 months, plus ratings of six emails by 24 British lecturers from 12 UK universities. Figures checked in the full text (author’s upload) on 12 October 2026; no funding stated.
  2. Danescu-Niculescu-Mizil, C., Sudhof, M., Jurafsky, D., Leskovec, J., Potts, C. (2013). A computational approach to politeness with application to social factors. Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics. Corpus of 10,957 requests rated by five US annotators each. Funded by NSF, ARO, DARPA and other grants and company support.
  3. Biesenbach-Lucas, S. (2007). Students writing emails to faculty: An examination of e-politeness among native and non-native speakers of English. Language Learning & Technology, 11(2), 59 to 81. 533 emails (382 native, 151 non-native) sent to one professor over six semesters; no funding stated.
  4. Economidou-Kogetsidis, M. (2016). Variation in evaluations of the (im)politeness of emails from L2 learners and perceptions of the personality of their senders. Journal of Pragmatics, 106, 1 to 19. Participant numbers were not visible in the parts I could open.
  5. Hendriks, B. (2010). An experimental study of native speaker perceptions of non-native request modification in e-mails in English. Intercultural Pragmatics, 7(2), 221 to 255. Two online experiments with 110 and 158 native English raters; no funding stated.
  6. Hendriks, B., van Meurs, F., Kakisina, B. (2023). The effects of L1 and L2 writers’ varying politeness modification in English emails on L1 and L2 readers. Journal of Pragmatics, 204, 33 to 49. 120 Dutch and 131 British participants; no funding stated.
  7. Vignovic, J. A., 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. Sample size is not stated in the abstract, the only part I could open.
  8. Murphy, M. L. (2025). Separated by a common im/politeness marker: please in American and British web-based English. English Language and Linguistics, 29(4), 781 to 804. GloWbE web corpus (collected 2012), about 386 million words per variety, 800 coded uses of please per variety, plus a survey of 1,325 US and 1,293 UK adults. Open access; research assistance funded by the University of Sussex. Full text read 12 October 2026.
  9. Degenhardt, J. (2023, online 2021). Requests in Indian and Sri Lankan English. World Englishes, 42(3), 523 to 543. Requests from the dialogue sections of ICE-GB, ICE-India and ICE-Sri Lanka (recorded speech). Funded by the German Research Foundation (DFG). Open access; full text read 12 October 2026.
  10. Xie, F. (2024). Politeness Variation: Politeness in Britain, Australia, and Hong Kong. Corpus Pragmatics, 8(4), 313 to 333. GloWbE web corpus, 21 politeness formulae, frequencies per million words. Open access funded by the University of Zurich; no competing interests declared. Full text read 12 October 2026.
  11. Oxford Learner’s Dictionaries. needful (noun), idiom “do the needful”. Checked 12 October 2026.
  12. Thomas, J. (1983). Cross-cultural pragmatic failure. Applied Linguistics, 4(2), 91 to 112.
  13. Bardovi-Harlig, K., Dörnyei, Z. (1998). Do language learners recognize pragmatic violations? Pragmatic versus grammatical awareness in instructed L2 learning. TESOL Quarterly, 32(2), 233 to 262. The finding is quoted as stated on the page for its replication, Niezgoda and Röver (2001), in Pragmatics in Language Teaching, Cambridge University Press.
  14. Brown, P., Levinson, S. C. (1987). Politeness: Some Universals in Language Usage. Cambridge University Press. Reissue of their 1978 essay with a new introduction. Bibliographic details and table of contents checked 12 October 2026; the book text itself was not opened, and the definitions of face are as quoted from the 1978 edition by Matsumoto (1988).
  15. Matsumoto, Y. (1988). Reexamination of the universality of face: Politeness phenomena in Japanese. Journal of Pragmatics, 12(4), 403 to 426. Full text read 12 October 2026.
  16. Ide, S. (1989). Formal forms and discernment: two neglected aspects of universals of linguistic politeness. Multilingua, 8(2 to 3), 223 to 248. Not open access; the argument is as quoted, with page numbers, by later authors (Mao 1994, Pizziconi 2003).
  17. Mao, L. R. (1994). Beyond politeness theory: ‘Face’ revisited and renewed. Journal of Pragmatics, 21(5), 451 to 486. Partial full text read 12 October 2026.
  18. Fukada, A., Asato, N. (2004). Universal politeness theory: application to the use of Japanese honorifics. Journal of Pragmatics, 36(11), 1991 to 2002. Full text not opened; their position is confirmed through the record and later citations.
  19. Usó-Juan, E. (2022). Exploring the role of strategy instruction on learners’ ability to write authentic email requests to faculty. Language Teaching Research, 26(2), 213 to 237. 110 emails before and after instruction; funded by Spain’s Ministerio de Ciencia e Innovacion, Universitat Jaume I and Generalitat Valenciana per Crossref; only the abstract was available to me.
  20. Qin, W., Jia, R., Ren, W. (2024). Pragmatic competence in an email writing task: Influences of situation, L1 background, and L2 proficiency. Written Communication, 41(4), 726 to 755. 306 emails from 32 English-speaking teachers and 121 learners; funded by the Shanghai Institute of Educational Science (C2023155) per Crossref; only the abstract was available to me.
  21. Ren, W., Li, S., Lü, X. (2023). A meta-analysis of the effectiveness of second language pragmatics instruction. Applied Linguistics, 44(6), 1010 to 1029. 29 studies, 1,898 learners, 54 effect sizes; overall g = 1.656. Abstract only; it does not say whether the difference between outcome measures was statistically significant.
  22. Nguyen, T. T. M. (2018). Pragmatic development in the instructed context: A longitudinal investigation of L2 email requests. Pragmatics, 28(2), 217 to 252. 32 Vietnamese EFL students (13 instructed, 19 comparison), emails elicited through three scenarios, tested up to eight months after instruction. Full text read (author’s upload).
  23. Beebe, L. M., Takahashi, T., Uliss-Weltz, R. (1990). Pragmatic transfer in ESL refusals. In R. C. Scarcella, E. S. Andersen, S. D. Krashen (Eds.), Developing Communicative Competence in a Second Language (pp. 55 to 73). Newbury House. Design and sample as described in the CARLA speech act bibliography; the chapter itself was not read. That page has a certificate fault in some browsers, and an archived copy is available.
  24. Golato, A. (2003). Studying compliment responses: A comparison of DCTs and recordings of naturally occurring talk. Applied Linguistics, 24(1), 90 to 121. Abstract only.
  25. House, J., Kádár, D. Z. (2023, print 2025). Speech acts and interaction in second language pragmatics: A position paper. Language Teaching, 58(3), 396 to 407. Quote from p. 403. Open access; funded by Hungary’s National Research, Development and Innovation Office; no conflicts declared.
  26. Kruger, J., Epley, N., Parker, J., 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; the fourth, on reading statements aloud, had 54 University of Illinois students; funded by a University of Illinois Board of Trustees grant and NSF grant SES-0241544. DOI resolved against Crossref on 10 October 2026; abstract and fourth experiment read in a full-text copy of the paper, because the publisher page did not load.
  27. Pollmann, M. M. H., Roos, C. A. (2025). “I get u”. People correctly interpret the tone of text messages and emails. Computers in Human Behavior Reports, 18, 100689. Study 1: text messages, 347 receivers and 171 senders. Study 2: work or education emails, 361 receivers and 61 senders. Rates positive or negative tone only. Crossref lists no funder. DOI resolved against Crossref on 10 October 2026; abstract read via OpenAlex, because the publisher page did not open.
  28. Cheng, Y.-S. (2004). A measure of second language writing anxiety: Scale development and preliminary validation. Journal of Second Language Writing, 13(4), 313 to 335. 421 EFL majors at seven colleges in Taiwan. Abstract only.
  29. Li, R. (2022). Understanding foreign language writing anxiety and its correlates. Frontiers in Psychology, 13, 1031514. Meta-analysis of 22 studies, 84 effect sizes; its total of 24,290 participants counts a sample once for each effect size; pooled r = -0.298 with writing performance and -0.382 with writing self-efficacy. Funded by the Social Science Foundation of Hunan Province. Open access; full text read 12 October 2026.
  30. LanguageTool. Free AI Grammar Checker. Checked 4 October 2026.
  31. DeepL. DeepL Write: your AI-powered writing assistant speaks French and Spanish. Checked 4 October 2026.
  32. Wordtune. Express yourself with confidence. Checked 4 October 2026.
  33. QuillBot. QuillBot’s Paraphraser: The Best AI Paraphrasing Tool. Checked 4 October 2026.
  34. Grammarly. Writing Tone Detector and Tone Suggestions. Checked 4 October 2026.
  35. Tonely AI Ltd. Tonely AI, App Store listing. Checked 4 October 2026.
  36. ConfiText. AI Texting Assistant. Checked 4 October 2026.
  37. QuillBot. How can I check and use Tone settings in Quillbot? Last updated 10 June 2026; checked 5 October 2026.
  38. Srinivasan, A., Choi, E. (2022). TyDiP: A dataset for politeness classification in nine typologically diverse languages. Findings of the Association for Computational Linguistics: EMNLP 2022.
  39. Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. 91 TOEFL essays and 88 US eighth-grade essays; funded by NSF, NIH, the Silicon Valley Foundation and the Chan Zuckerberg Initiative. No retraction or correction found on 12 October 2026.
  40. Crosthwaite, P., Sun, S. (2025, issue 2026). Generative AI and L2 Written Feedback Studies: A Scoping Review. RELC Journal, 57(1), 207 to 219. 51 studies, searched to April 2025. Open access; no funding and no conflicts declared. Full text read 12 October 2026.
  41. Chon, Y. V., Shin, D. (2026). AI Chatbot as a Revision Aid in Second Language Writing: From Error Correction to Lexical Sophistication. Journal of Computer Assisted Learning, 42(3), e70259. 40 Korean university English-education students, 39 of them analysed, one session. Funded by Hanyang University. Abstract read 12 October 2026.
  42. Yan, D., Zhang, S. (2024). L2 writer engagement with automated written corrective feedback provided by ChatGPT: A mixed-method multiple case study. Humanities and Social Sciences Communications, 11, 1086. Four learners at one Chinese university, five weeks; descriptive only. Full text read 12 October 2026.
  43. Agarwal, D., Naaman, M., Vashistha, A. (2025). AI Suggestions Homogenize Writing Toward Western Styles and Diminish Cultural Nuances. Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 118 participants in India and the US. Funded by Infosys, the Microsoft AFMR Program and Global Cornell, with partial NSF support.
  44. Liu, A., Zhu, S. (2026). The Cost of Perfect English: Pragmatic Flattening and the Erasure of Authorial Voice in L2 Writing Supported by GenAI. arXiv preprint 2605.13055, not peer reviewed. 106 essays from the ICNALE learner corpus, 424 AI revisions by four models; descriptive statistics only; no funding statement.
  45. Subtext, Terms, privacy and your account. Terms last updated 26 July 2026, privacy policy last updated 10 October 2026. Checked 12 October 2026.
  46. Subtext, home page and FAQ. Checked 12 October 2026.