Get mentioned on the blogs AI cites

Start Free

Cold Email Grader

Paste a cold email and get a 0 to 100 score from twelve rule-based checks, each calibrated on what earned replies across 8,531 real sends. Nothing leaves your browser. · Updated September 14, 2026

Try:

    Or let the drafter write it to these rules

    MentionAgent finds the sites, writes each email to the pattern that replied best, and holds every draft for you to read before it sends.

    Start Getting Mentioned On Autopilot
    Our data

    The score is calibrated on 8,531 cold first-touch emails MentionAgent customers sent, measured on 31 August 2026, counting only a human reply within three days. Emails under 50 words replied 6.39%; 90 to 104 words, 3.19%.

    Paired check: inside the same account and week, under 65 words replied 3.92% against 2.62% for over 80, and the short arm won in ten of twelve accounts.

    The other checks: the drafter's own gate, which rejects em dashes, twenty-two words, "Hi there", first-touch offer language and anything over 90 words.

    The data the grade is built on

    Most email graders score against somebody's opinion of good copy. This one scores against reply rates. On 31 August 2026 we measured every cold first-touch email MentionAgent customers had sent in the preceding weeks, 8,531 of them, counting a reply only if a person answered within three days of the send, with auto-replies, out-of-office notices and bounces removed and institutional recipients (universities, government, nonprofits) excluded so the comparison is like for like. The single strongest variable was length.

    Body lengthCold emails sentHuman reply rate
    Under 50 words3606.39%
    50 to 64 words1,7795.62%
    65 to 79 words3,7765.06%
    80 to 89 words1,6124.84%
    90 to 104 words1,0043.19%

    The obvious objection is that short emails come from different customers or different weeks. We checked: paired inside the same customer account and the same calendar week, over the thirteen cells with at least twenty sends in each arm, emails under 65 words replied 3.92% against 2.62% for emails over 80, and the short arm won in ten of the twelve accounts with enough volume to compare. The word count includes the greeting and the sign-off, because the measurement did; the grader counts the same way.

    The other checks come from the gate our drafter runs before any email is queued. That gate rejects em dashes, twenty-two specific words, "Hi there" in place of a name, offer language on a first touch, copy that tells the recipient there is no fit, and anything over 90 words. Each of those rules exists because the version without it measurably lost replies; the grader applies them to your email exactly as the gate would.

    What the twelve checks are

    CheckPointsPassFail
    Length25Under 65 words, greeting and sign-off included90 or more, the drafter's hard cap; 25 or fewer is too short to say what you want
    Greeting10A name: "Hi Dana,""Hi there", "Hey there", "Dear Sir or Madam", or no greeting
    Em and en dashes5NoneAny, in any encoding
    Drafter's banned words15None of the 22Two or more; one is a warning
    Spam trigger words10NoneThree or more; one or two is a warning
    Offer on a first touch5No price, trial, day count or "send the link"Any of those
    The ask10Exactly one question, and it is the last thing before the sign-offNo question, or three or more
    Links10One at mostThree or more; two is a warning
    Written for this person10Names something of theirs: a post, page, episode, talk, productNothing that could only be said to this recipient
    No-fit copyfailAbsent"not a good fit", "no real overlap", "best of luck with the work"
    Formatting5Sentence case, one exclamation mark at most, no paragraph over 60 wordsShouted words, "!!!", a wall of text
    Subject (if pasted)10Six words or fewer, plain, no fake "Re:"Ten or more words, capitals, spam words, a fake reply prefix

    The score is points earned over points available, so leaving the subject blank does not cost you; it just removes ten points from both sides. The twenty-two banned words are: comprehensive, streamline, streamlining, leverage, delve, meticulous, elevate, revolutionize, holistic, empower, seamless, game-changer, unprecedented, cutting-edge, supercharge, laser-focused, bespoke, dofollow, backlink, resources page, return the favor, link exchange, partnership opportunity. The first seventeen are what machine-written copy sounds like; the last five tell a publisher, in the first sentence, that the email is about a link and nothing else.

    Reading a low score

    A cold email that fails several checks at once almost always has the same shape: a greeting with no name, a paragraph about the sender, a paragraph about the product with a price in it, three links, and a sign-off with an exclamation mark. Fixing the checks one by one produces a shorter version of the same email. Rewriting produces a different one: one line about something specific the recipient made, one line on why a link to your page would fit there, one question. That version passes every check without trying, and in our data it is the version people answer.

    Two checks deserve a caveat. "Written for this person" looks for words like your post, your guide, your episode, and it can be fooled by an email that says "your blog" and means nothing by it; it can also miss a genuinely personal email that quotes a title without those words. Read it as a prompt, not a verdict. And the length check is a cold first-touch rule; a reply in an existing thread can and should be as long as the question it answers.

    Frequently asked questions

    What does the cold email grader check?

    Twelve things, all rule based: body length against the reply rates we measured by word count, whether the greeting uses a name, em dashes, twenty-two words our own drafter refuses to send, spam trigger words, offer language on a first touch (prices, trials, day counts), whether there is exactly one question, how many links, whether anything in the email could only have been written to this recipient, no-fit phrasing, formatting (capitals, exclamation marks, wall-of-text paragraphs) and, if you paste one, the subject line.

    Does it use AI to grade my email?

    No. Every check is a rule that runs in your browser; nothing is sent to a server and no model is called. That is deliberate: the rules are the same ones MentionAgent's drafter enforces before a send, so the grade tells you how far your email is from the pattern that earned replies in our data, not what a model happens to think of it today.

    Where does the length data come from?

    From 8,531 cold first-touch emails sent by MentionAgent customers and measured on 31 August 2026: replies counted after a three-day window, auto-replies and bounces excluded, commercial recipients only. Under 50 words replied 6.39%, 50 to 64 words 5.62%, 65 to 79 words 5.06%, 80 to 89 words 4.84%, and 90 to 104 words 3.19%. Paired within customer and week, the short arm beat the long arm in 10 of 12 accounts.

    Why does it fail an email for an em dash?

    Because our drafter does, and for a reason a reader would recognise: em dashes are the most reliable tell of machine-written copy in 2026, and a publisher who spots one stops reading. The grader treats it the way the send gate does, as a hard fail, and the fix is a comma or a full stop.

    Why is a price or a free trial a problem?

    On a cold first touch it reads as an offer to a stranger, and our reply rates fell when the drafter leaked that phrasing into first emails that were meant to be a soft ask. The send gate now rejects dollar amounts, the word trial, for N days and send me the link on those emails. Save the offer for the second message, after they have replied.

    What is a good score?

    80 or above means the email would clear our send gate and sits inside the length band that replied best. 60 to 79 usually means one structural problem: too long, no question, or no sign it was written for this person. Below 60 the email has more than one of those, and rewriting from the greeting down is faster than patching.

    Is a longer email ever better?

    In our data, no, on a cold first touch. The email that got the most replies referenced one thing on the recipient's site, said in one line why a link made sense, and asked one question, in 50 to 60 words. Longer emails add the product description and the paragraph about why their audience would care, and those are the sentences readers skip.