Will this message land badly? Check before you send.

Paste a draft message or email below — get a tone-risk score, every risky phrase flagged sentence by sentence, and a softer rewrite. Checked entirely on your device.

  • Your message text never leaves the device
  • 9 risk patterns: blame, sarcasm, shouting…
  • Tone dials: work / friend / customer
  • Softer rewrite included

Paste your draft message

Up to 5,000 characters. The check runs as you type — nothing is uploaded, and the text is never stored.

Try a sample:

Checked in your browser — nine risk patterns are matched against every sentence, with a softer rewrite. Nothing is uploaded.

Work: calibrated for colleagues, managers and clients — demanding language and curt replies weigh more.

What was flagged

    Sentence by sentence

      No message yet. Everything stays on this device. The rewrite is a suggestion, not a finished message.

      Privacy you can verify, not just trust

      Most online tone checkers send your message to their server for an AI model to read — one caps the free check at 100 words, another allows five checks a day, and a third keeps the deeper read behind an account. This one can't: your message is analysed entirely in your browser. Your message text never leaves the device — the only third-party requests are the site's own analytics, which count page views — and it uploads nothing, ever. There is no server to send it to. Don't take our word for it:

      New cross-origin requests during use: 0
      1. Press F12 to open developer tools, then the Network tab.
      2. Check a message above. Watch: apart from the site's analytics, nothing new is requested — no uploads, and no message text ever leaves your device.
      3. Extra credit: disconnect your internet after this page loads — the check still works.

      How to check a message's tone before you send it

      A message can say exactly what you mean and still sound different to the person reading it. Text strips away voice, face and timing, and the reader fills the gap with their own mood — so a rushed one-word reply reads as annoyed, "per my last email" reads as a scolding, and "you always" starts an argument about the accusation instead of the point.

      This tool gives you a ten-second second opinion where the draft already is: on your own device. When you paste a message here, your browser:

      1. Scores the tone risk — nine patterns (blame, absolutes, sarcasm, passive-aggression, over-apologising, shouting, curt replies, demands, hedging) are matched against every sentence, weighted for the tone dial you picked, and totalled into a score from 0 to 100 with a verdict band.
      2. Shows its work, sentence by sentence — each sentence carries its flags and its risk, and each flag quotes the exact phrase it matched and explains why that phrasing reads badly, with a softer alternative.
      3. Offers a softer rewrite — the stock phrases and shouting are toned down mechanically, so you have a starting point to edit into your own words.

      What makes this checker different

      The pages that top the results for this search mostly send your message to a server for an AI model to read — with a free-word cap, a daily check limit, or an account behind the deeper read — and the one free in-browser checker we found does not rewrite your message. This page is the only free, fully client-side tone checker we found that does all three: flags the exact risky phrases, breaks the message down sentence by sentence, and offers a softer rewrite — with no daily limits, no account and no upload — plus a live network-request counter so you can verify the privacy claim yourself. It runs entirely in your browser — there is no server to upload to.

      When is checking the right choice — and when is it the wrong one?

      Checking is right whenever a message has emotional stakes: a complaint, a rejection, a hard request, feedback, a follow-up after silence, or anything to a manager, a client or someone you are already tense with. It is the wrong choice for judging a message someone sent you — this tool checks drafts you are about to send, and it cannot know what the other person meant — and for jokes between close friends, where the shared context the tool cannot see is the whole point.

      Honest limitations

      Pattern-matching, not understanding

      This tool matches phrases — it does not understand what you mean. A flag is a prompt to re-read, never a verdict on your message.

      Sarcasm needs context

      Sarcasm depends on relationship and shared history. The checker flags the common markers, but it will miss sarcasm with no marker and flag jokes it cannot recognise.

      It cannot see the relationship

      The same sentence lands differently from a boss, a friend and a stranger. The tone dials adjust the weights, but they are no substitute for knowing your reader.

      English only

      Tone markers differ a lot between languages. The patterns are English-only, so checks in other languages will miss most risks.

      Warmth is only a small credit

      A "please" or "thank you" lowers the score slightly, but it cannot cancel a sharp phrase — a polite accusation is still an accusation.

      The rewrite is a starting point

      The softer version only makes mechanical changes to stock phrases and shouting. It cannot know what you mean — always read it through and make it yours.

      Frequently asked questions

      Do my messages upload anywhere?

      No. Your text is analysed entirely in your browser — your message text never leaves the device, and the only third-party requests are the site's own analytics, which count page views. You can verify it yourself with the network-request counter in the privacy section below the checker.

      How does the checker work?

      It is deterministic pattern-matching, not an AI model. A rule table of phrases and patterns that commonly read badly — blame words, absolutes, sarcasm markers, passive-aggressive stock phrases, shouting, curt replies, demands and hedges — is tested against every sentence of your draft. Each hit adds risk weight, warmth signals such as please and thank you subtract a little, and the total becomes a score from 0 to 100 with a verdict band. Every flag quotes the exact phrase it matched, so you can see the reasoning.

      What exactly does it flag?

      Nine patterns: blame language ("you always", "your fault", "you failed to"), absolutes ("always", "never", "every time"), sarcasm markers ("yeah right", "oh great", "good luck with that"), passive-aggressive phrasing ("per my last email", "as I said", "still waiting"), over-apologising (repeated sorries), shouting (ALL-CAPS words and repeated punctuation), curt or dismissive replies ("noted", "whatever", "deal with it"), demanding language ("you need to", "ASAP", "immediately"), and hedging ("just", "maybe", "I think"). Each flag explains why the phrasing reads badly and suggests a softer alternative.

      What do the risk bands mean?

      Low risk (0–19) means no risky phrasing was found — still re-read it once as the other person, because the tool cannot see your history with them. Some risk (20–39) means one or two phrases could be misread — worth a quick edit. High risk (40–59) means several risky phrases — reword before sending. Very high risk (60 and above) means the message is likely to land as hostile or accusatory — do not send it as-is. The score is a heuristic, not a measurement of what the reader will feel.

      What do the tone dials do?

      Work, friend and customer re-weight the same flags for the relationship. In work mode, demanding language and curt replies weigh more, because orders land hard on colleagues. In friend mode, sarcasm and hedging weigh less, because jokes and softeners land differently between friends. In customer mode, over-apologising, hedging and curt replies weigh more, because customers read them as indifference. The flags themselves do not change — only how much each one counts.

      Is the softer rewrite ready to send?

      No — treat it as a starting point, not a finished message. The rewrite only makes safe, mechanical changes: it swaps the worst stock phrases ("per my last email" becomes "following up on my earlier message", "you need to" becomes "could you please"), lowercases shouted words and collapses repeated punctuation. It cannot know what you actually mean, so read it through, fix anything it got wrong, and make it yours before sending.

      Can it detect sarcasm reliably?

      No — and neither can any tool that only reads words. Sarcasm depends on context, relationship and shared history, none of which a pattern-matcher can see. The checker flags the markers that most often read as sarcastic ("yeah right", "oh great", "good luck with that"), and it deliberately does not flag bare words such as "brilliant" or "interesting" on their own, because those are usually genuine. A flag is a prompt to re-read, not proof the sentence is sarcastic.

      What does this tool NOT do?

      It is not a grammar or spelling checker, it is not sentiment analysis, and it does not understand what you mean — it matches patterns. It works in English only, because tone markers differ a lot between languages. It cannot judge a message you received, only one you are about to send, and it cannot tell you what the reader will feel. The per-sentence breakdown is shown so you can judge every flag yourself.

      Why check tone before sending?

      Because text strips away everything that carries tone in person — voice, face, timing — and the reader fills the gap with their own mood. Research on email misunderstandings has found people often read a different tone into a message than the sender intended, and a short message lands hardest: a one-word reply to a long message reads as annoyed even when it was only rushed. A ten-second check before sending catches the phrases that start arguments — blame, absolutes, "per my last email" — while they are still editable.