WRITE FOR THE READER  //  relevance beats clever every timeApply
Cold email copywriting

Cold email copywriting: the rules our grader scores against

Updated July 2026  //  by Mark Glazer  //  the same 11 checks the grader runs

The short answer: A first touch is 50 to 80 words, aim around 65, with one soft ask as its last line and a subject that passes 2 of 3 tests: what is this, who is it for, what does it promise. Our grader scores these exact 11 checks out of 100 in about 30 seconds.

Most copywriting advice is adjectives. Ours had to compile: the cold email grader scores pasted copy against the playbook we run for clients, and code cannot check "be relevant". Every rule below is a rule because it has a number.

Where these rules come from

The grader runs 11 checks that sum to exactly 100 points, entirely in your browser, in about 30 seconds, then applies the mechanical fixes in one click. Building it imposed a useful discipline: any rule we could not reduce to a threshold did not ship. This page is the published version of those checks, with the numbers the code actually uses.

The subject line must pass 2 of 3 tests

The subject check is worth 12 of the 100 points. A line passes when it clears at least 2 of 3 tests: it says what the email is, it says who it is for, and it makes a promise, either a figure or a near-money word such as meetings, deals, revenue or pipeline. Five things zero the check outright no matter what else the line does: a fake "Re:" on a first touch, an em dash, ALL CAPS, a merge tag standing alone as the whole subject, and a template opener.

"Quick question" is on that banned-opener list. An earlier version of this page used it in its example subject line. Openers that once read as casual now read as template, because every template uses them.

Length: a hard cap of 80 words, an aim of 65

50 to 80 words earns the full 10 points; over 80 scores zero, not a small deduction, because past the cap the problem is never one extra sentence. When you are over, cut everything that is not identification, mechanism, or the ask.

The follow-up rule is stricter and surprises people: one sentence, 20 to 28 words, no greeting, no sign-off, and it must still name the concrete offer, a teardown, an audit, a report. A follow-up reduced to "is now a good time?" fails the offer check outright.

One ask, and it is the last line

The ask is worth 10 points and the discipline is a gradient: reply first, resource second, time last. Asking a stranger for a 30-minute call, or dropping a booking link into a first touch, fails the check; a soft reply-level question as the final line passes it. A question buried mid-message gets docked too: the last thing a skimming reader sees should be the ask.

Follow-ups end on a value-check instead: "worth it?", "useful?", "valuable?", ideally with an easy out such as "or not a fit?". A cheap way to say no makes saying yes cheap too.

Sound like a person: 606 AI tells, 480 fluff phrases

The vocabulary check (12 points) carries 606 AI-writing tells in three tiers. Tier 1 words are always flagged: delve, leverage, robust, comprehensive, cutting-edge. Tier 2 words fire only when two or more cluster in the same email. Tier 3 words fire only on density, at roughly 3 percent of the body or more. The tiering is the point: "streamline" once in a year of email is a word; five of its cousins inside 70 words is a fingerprint.

Two related checks are pass or fail. A single em dash anywhere, including the double hyphen typed as two minus signs, zeroes an 8-point check; use a colon, period, comma or parentheses instead. And the 480 tracked fluff phrases split into hard cliches ("just checking in", "hope you're well", "circling back"), which fail, and soft filler, which only warns.

Specificity, mechanism, offer: 30 points on substance

Specificity (10 points) passes on a concrete figure with no platitudes. "Our net profit is 3 percent" is believed; "lowest prices" is not. Weak outcomes such as "save time" and "streamline operations" fail the check; near-money outcomes pass it.

Mechanism (8 points) demands a "because": one mechanism named in 3 to 7 words, not a comma-separated feature soup. A claim with no mechanism attached does not move a buyer who has seen it before.

The offer check (12 points) looks for three things: who it is for, a near-money outcome, and, only if your pricing genuinely supports it, a payment condition such as "we only get paid when deals close". That risk reversal is the largest slice of the check, and it is the one line the grader will never insert for you, because it is a business commitment only the sender can make. It is also why we run a revenue-share model ourselves: the strongest sentence in the playbook has to be true.

Tone: the reader outnumbers you

The reader-focus check (5 points) counts pronouns: "you" and "your" must appear at least as often as "I", "we", "our" and "us" combined. The tone check (5 points) penalises more than one exclamation mark, any emoji, ALL CAPS, and known spam triggers, which are scanned everywhere except inside a risk-reversal clause. The full trigger list is public: our spam checker highlights all 1,130 of them live as you type. One house rule worth stealing: in every rewrite the grader swaps the four-letter pricing word that spam filters weigh most heavily for "complimentary".

What a grade means, and what copy cannot fix

85 or above is an A, 70 to 84 a B, 55 to 69 a C, and below that the built-in copy doctor is the point: it repairs em dashes, deletes filler sentences and swaps flagged vocabulary deterministically, then leaves the judgment calls flagged.

One ceiling worth stating: an A-grade email still fails on a bad list or from a burned domain. Copy sits on top of list building and deliverability, which is why the full playbook treats them as one system.

Two format-level companions to these rules: email preview text, the inbox snippet your first sentence becomes; and plain text email, why campaigns send just words and what that trades away.

Common questions

How long should a cold email be?

50 to 80 words for a first touch, with 80 as a hard cap and about 65 as the aim. Going over 80 scores zero on the length check rather than losing a point or two. A follow-up is stricter: one sentence of 20 to 28 words, no greeting, no sign-off, still naming the concrete offer.

What makes a good cold email subject line?

One that passes 2 of 3 tests: it says what the email is, who it is for, and what it promises. A fake "Re:" on a first touch, ALL CAPS, an em dash, a bare merge tag, or a template opener such as "quick question" fails the subject outright regardless of the rest of the line.

How many asks should a cold email have?

One, and it should be the last line. On a first touch keep it at reply level; asking a stranger for a 30-minute meeting or dropping a booking link fails the gradient of reply first, resource second, time last. Follow-ups end on a value-check question such as "worth it?", ideally with an easy out like "or not a fit?".

What words should you avoid in cold email copy?

Template openers ("just checking in", "circling back", "hope you're well"), AI-tell vocabulary (the grader tracks 606 terms such as "delve", "leverage" and "robust"), superlatives, and the 1,130 spam triggers the spam checker highlights live. The house rule on pricing words: say "complimentary".

What is a good score on the cold email grader?

85 or above is an A, meaning the copy passes the playbook. 70 to 84 is a B with a flagged item or two, 55 to 69 is a C, and below 55 the copy doctor earns its keep: it repairs em dashes, deletes filler sentences and swaps flagged vocabulary in one click, then leaves the judgment calls to you.

Grade the email you were about to send

Thirty seconds for the score and the one-click fixes. Or skip the learning curve entirely: we write, test and run the whole campaign, paid from the revenue we help close.

Apply to work with us Run the grader