November 6, 2026 · 8 min read
Email tone checker: what it measures and what it misses
Tone checkers read word choice, punctuation and length. They miss your history with the reader, and on short emails they often never run at all. Here is the ninety second audit that covers the gap.

A tone checker scans the words in your draft, scores punctuation, hedges and sentence length, then labels the result: formal, friendly, confident, worried. It measures text, not context. On short emails it often doesn't run at all, because the two best known checkers need 100 to 150 characters before they say anything.
What an email tone checker actually measures
Under the label, a tone checker is a text classifier. It reads surface features it can count, then maps them onto a small set of tone words. Nothing about your relationship with the reader goes into the calculation, because none of it's in the text field.
- Word choice: formal or conversational verbs, apology words, thanks words, and absolutes such as always and never.
- Punctuation and sentence length: exclamation marks, question marks, capitals used for emphasis, and whether a sentence ends hard or trails off. The same rhythm signals that detectors describe as burstiness.
- Hedges and intensifiers, the words that decide whether a request reads as a suggestion or an order.
Those four features are also what your reader uses, which is why the label sometimes lands. They aren't the whole of tone, which is why it sometimes misses by a mile. Vendors also blur one distinction worth keeping: a detector labels the whole text, while a suggester rewrites individual sentences toward a target.
Grammarly splits the two in its own support documentation: the tone detector identifies the overall tone of your text, while tone suggestions offer rewrites of individual sentences to make writing sound more personable, positive and confident (Grammarly Support, accessed 6 November 2026).
Why most work emails never trigger a check
Both major implementations have a length floor, and the floors sit above a normal reply. Grammarly's support page states that its tone detector needs at least 150 characters to activate. Microsoft's support page for coaching by Copilot in Outlook asks for at least 100 characters before it gives any feedback.
Here is the awkward part. Four sentences you have probably sent this week: "Can you review this by 3?" (25 characters), "Quick question about tomorrow" (29 characters), "Thanks, that works for me." (26 characters), "Sent from my phone, reply tonight." (34 characters). Every one sits under 100 characters, so neither checker would have said anything at all.
The emails that get misread are short by nature. A three word reply with no greeting reads as curt, and it's exactly the draft no tool will score. Length floors are a quiet reason those threads go wrong, and two more blind spots sit alongside them.
- A reply inside a long thread is scored on its own text, with the earlier messages, the unanswered question and the history all invisible to it.
- The checker reads the body, not the subject line, which is where a lot of tone is actually set before anyone reads a word.
What a tone checker can't see
Context isn't a missing feature waiting for the next release. it's information the tool was never given.
- Your history with the reader: whether you owe them an answer, whether the last message went unanswered, whether a joke is allowed.
- The reader's state today: end of quarter, a bad week, a message they were dreading.
- Your intent, including irony, plus register differences across cultures and first languages, where directness is normal and hedging reads as evasive.
- The decision the email is trying to move, which is the real measure of whether the tone worked.
There is decent evidence that senders overrate their own clarity, which is why a confident label feels so convincing. Across five experiments, participants predicted they would communicate their tone successfully 88.8 percent of the time and managed it 70.4 percent of the time.
Kruger, Epley, Parker and Ng measured that gap directly: predicted success 88.8 percent against actual success 70.4 percent, in “Egocentrism over e-mail: can we communicate as well as we think?”, Journal of Personality and Social Psychology, 2005, volume 89, number 6, pages 925 to 936, doi:10.1037/0022-3514.89.6.925.
How do you check tone when the tool stays silent?
You read the email as the recipient would, against a fixed order of questions. The order matters, because the usual failure is answering the easy question and calling it a review. The test isn't whether the draft sounds fine to you. You wrote it, so it does. The test is what the reader can infer in thirty seconds with no context. Picking the register before you type a word is its own craft, and how to match email tone to the reader covers it.
The six-check tone audit, scored
Run these six checks on any email under 150 characters. Score each one 0, 1 or 2.
- First line: does it name what the reader needs to act on, inside the first eight words, with a greeting chosen on purpose rather than inherited from a template? Score 0 if the ask arrives at the end.
- Request shape: is there one clear ask, phrased so the reader knows what yes means? Score 0 for two asks in one sentence.
- Hedges: count them. Score 2 for none, 1 for one, 0 for three or more, since stacked hedges read as unwillingness.
- Warmth: is there one human detail, a name, a reference to the last conversation, or a plain thanks? Score 0 if the text would fit any recipient.
- Specifics: is there a date, a number or a next step in the body? Score 0 if the only deadline is implied.
- Close: does the last line commit to a next step? Score 0 if it leaves the thread hanging.
Add the scores. Ten or more means send it. Seven to nine means fix the check you scored lowest, then resend to yourself. Six or below means the first sentence is the problem, so rewrite that and leave the rest alone. The audit takes about ninety seconds on a three line email, and it works because it's coarse: a rubric forces the three judgements a checker skips, namely who is reading, what you want them to do, and what happens if they misread it.
Which tone checkers are worth installing in 2026?
Install one, use it as a nudge, and don't let it decide. The realistic read on the three shapes available:
- A browser based checker that overlays webmail. Grammarly's detector runs from 150 characters and labels overall tone, which covers longer drafts and misses quick replies.
- A coaching feature inside Outlook. Tone detection reached Outlook on the web through Microsoft Editor in October 2021, and the current coaching flow asks for 100 characters before it comments on tone, clarity and reader sentiment.
- A general assistant you paste into. It can rewrite toward any register you name, which gives you more control than a checker, and it still has no memory of how you talk to this particular person.
None of the three scores you against your own register. That gap is what a written tone guide fills, and it's the reason the guide matters more than the label.
Where tone suggestions make you sound less like yourself
Suggestions have a direction. Grammarly's documented target is writing that sounds more personable, positive and confident. that's a fine default for a product update and a bad default for a complaint, a delay notice or an apology that needs to stay plain. Take a line that has to be firm, written the way the tool prefers:
“I completely understand your frustration, and I am so excited to share that we're actively working on a fix and will keep you posted!”
The same line with the facts left where they're:
“We missed the date we gave you. The fix ships Thursday, and I will confirm here when it does. If you need a workaround before then, I can send one today.”
The second version scores worse on positivity and lands better, because it answers the question the reader asked. Accepting suggestions one by one keeps that choice. Accepting them all at once hands the choice to the tool.
A tone audit on a real three line email
Start with the draft most people send without reading twice:
“Hi, just wanted to check in on this and see if there is any update? Sorry to bother you, no rush.”
- First line: the ask is a check in, so the reader works out the request. 0.
- Hedges: just, any, sorry, no rush. Four in two sentences. 0.
- Warmth and specifics: no name, no date, and the apology is doing the work of a deadline. 0.
Rewrite, same facts, same politeness, one clear ask:
“Hi Sam, checking on the invoice from the 14th. Do you have a date for it? I need to close the month on Friday, so a yes or no today is enough.”
Nothing was added except a name, a date and a real deadline. The hedges went because a deadline does the hedging for you. that's the whole technique: swap softeners for specifics.
When to skip the checker entirely
Some emails should not be optimised for tone at all, because text is the wrong medium for what you have to say.
- Bad news about a person: a rejection, a performance message, a redundancy. Say it live, then write the summary.
- Anything readable as a commitment, a warning or a legal position. Clarity beats warmth, and a suggester pushes the other way.
- A thread that has already gone wrong. A third careful paragraph reads as evasive, and a five minute call resets it.
For everything else the check is worth ninety seconds, and the habits that make a draft read as robotic are the same ones the audit catches. If the draft still reads thin afterwards, the problem is usually phrase level rather than tone level, and the routine for fixing tone on longer emails covers that case.
Frequently asked questions
Is a tone checker accurate on short emails?
Usually it does not run. Grammarly's detector needs at least 150 characters to activate and the coaching flow in Outlook asks for at least 100, so a three line reply falls below both floors and gets no score at all. Short emails are exactly where a manual audit earns its keep.
Does an email tone checker measure politeness or formality?
It labels both from the same surface features: word choice, punctuation, sentence length and hedges. That means it can call a message formal without noticing it is cold, and call a warm message unprofessional because it carries two exclamation marks.
Can a tone checker tell whether an email was written by AI?
No. Tone detection and AI detection are different classifiers. A tone checker answers which register the text is in. An AI detector answers how predictable the text is. Neither one reads the relationship you have with the recipient.
What should I check if I only have thirty seconds?
The first line. If the reader cannot tell what you want from the opening sentence, warmth later in the message does not fix it. Name the thing, then say what you need and by when.