Cold email subject lines: 24 examples that work

Most cold email subject line advice optimizes the wrong field. In the inbox list, your subject renders beside two things the recipient weighs first: a sender name and a domain. A plain subject from a credible-looking sender beats a clever subject from a stranger, and no wording on earth recovers from mail47.outreach-blast.net.

The subject is the third thing they read

Scan your own inbox and watch what your eyes do. Sender name first. Then, if the name is unfamiliar, the domain. The subject gets whatever attention survives those two checks, which for an unknown sender is roughly half a second.

That ordering has a practical consequence: the best subject line for cold emails is usually the plainest one sent by the most credible sender. Credibility is built weeks before the send, with an authenticated domain, a warmed history, and a complaint rate near zero. That groundwork is the subject of our cold email infrastructure guide; this article assumes it is done.

We run our own mail platform and read the cold pitches that land in it. The ones we open share a shape. A human name. A domain that matches the company being described. A subject that sounds like a sentence someone typed, about us specifically, rather than a headline someone A/B tested.

Internal-style subjects, and the line you cannot cross

Real one-to-one mail is typed fast and rarely capitalized. Marketing mail gets title case, polish, and a preheader. Recipients have absorbed that split without ever thinking about it, so budget question for q3 reads as a colleague and A Question About Your Q3 Budget Strategy reads as a campaign. Choose the register of a person. Lowercase when your own voice suits it, sentence case when it does not, title case never.

Mimicking the tone of internal mail is fair play. Forging its context is not. A "Re:" on a thread that never existed, or a "Fwd:" implying someone passed your pitch along, is a fabricated header, and the CAN-SPAM Act prohibits exactly that: the FTC's compliance guide requires that the subject line "accurately reflect the content of the message" and that header information not be misleading. The penalty runs up to $53,088 per individual email under the FTC's inflation adjustment effective January 2025, and there is no cap on the total.

The commercial case against faking it is just as sharp. The recipient opens, finds no prior thread, and feels tricked, and tricked people click "Report spam." Under Google's bulk-sender rules (February 2024, tightened since), a complaint rate at or above 0.3% means Gmail withholds delivery mitigation until you sit below that line for 7 consecutive days. Three angry recipients per thousand is all it takes. The legal side has its own page: is cold email legal covers the full checklist.

Never fake a "Re:" or a "Fwd:". It is the one subject-line rule with a statute attached.

24 subject lines, grouped by angle

We wrote these. Take the reasoning, not the string: a subject copied verbatim from a listicle is already in a thousand other outboxes. Braces mark what you replace, and each of these deserves a body to match, which is what our cold email templates are for.

Specific value

The strongest angle. You looked at their business, found something concrete, and named it.

• your unsubscribe link is broken. A checkable fact about them, found by you. Nobody deletes this unread, and the body writes itself. Only send it if it is true.
• idea for {company}'s onboarding emails. Names the artifact, promises a contribution, asks nothing yet.
• the {competitor} comparison page you don't have. Specific gap, named rival. Works because it is the recipient's job to care about that gap.
• saw the outage postmortem, one suggestion. Proves you read something they published this week. Recency is the personalization that cannot be faked in bulk.
• {company}'s careers page lists 6 SDR openings. Their own number, quoted back. Signals research and implies a relevant problem (six new senders, one domain).
• a faster first reply for {team}. Plain promise, no adjectives. Works when the body immediately says how.
• you're paying for 40 seats, using 12. Only when you can actually see their usage. Devastatingly effective and instantly spam-reported if you guessed.

Question

Questions get opened because unanswered questions itch. They fail when generic, so aim the question at their org, not at the universe.

• who owns deliverability at {company}?. Doubles as a routing request. Even a "not me, ask Dana" is a reply, and replies are the engagement signal mailbox providers weight most.
• is {project} still on the roadmap?. Uses a fact from their changelog or blog. Reads like a stakeholder, not a vendor.
• are you still on {tool}?. One fact of research, one plausible business question. Get the tool name from their job posts or their stack page; guess it and the email is dead on arrival.
• what happens when {tool} sunsets?. Ties to a real deprecation date. Urgency you did not invent.
• open to a teardown of {site}'s signup flow?. An offer disguised as a question. Sets up a body that delivers the first finding free.
• wrong person for {topic}?. Self-deprecating, honest, and very hard to be annoyed at. Pairs with a referral ask in the body.

Referral

The highest-trust angle and the easiest to abuse. A referral subject with no real referral behind it is a lie with a witness, and the witness gets asked.

• {name} pointed me your way. The classic. Only if the pointing happened.
• {name} said you own {area}. Adds the reason the referral routes to them specifically.
• after {name}'s intro. For when the intro email exists and this is the follow-through.
• your talk at {event}. Context, not referral, but works the same way: shared ground before the ask.
• we both know {name}. The weakest of the five. True mutual connection, no actual referral. Use only when the name is genuinely close to them.

Brevity

One to four words. These survive every client's truncation and read as confident, but they carry zero context, so the sender name and domain do all the trust work.

• {company} + {yourcompany}. States the topic honestly in the fewest possible characters.
• deliverability question. Names your lane. Works when the domain backs it up.
• {city}, next thursday. For a real visit. Concrete time and place reads as a person with a calendar.
• onboarding idea. Two words, one promise. The body has to deliver it in the first line.
• hiring?. One word attached to a real signal, like a job posting. Cheeky; know your audience.
• spam folder problem. Blunt, relevant to anyone who sends mail, and exactly what our own inbound pitches should have said.

What phones actually show

Truncation points are measurable. Email Tool Tester's client-by-client measurements (checked August 2026) put the Gmail app on a phone at roughly 30 visible characters and Apple Mail on an iPhone at roughly 41 to 46 before truncation. Desktop clients show more. You do not get to pick which client your recipient opens, so write for the smallest window on the list.

So the practical rule: a 1-to-4-word subject always survives. A 6-to-10-word subject loses its tail on phones, which is fine only if the specific noun comes first. {company}'s careers page lists 6 SDR openings truncates to {company}'s careers page lists... and still works; flip the clause order and the phone shows only your setup, not their name.

The test costs nothing: send the draft to yourself and look at your own phone. Thirty seconds, and it settles every length debate you were about to have.

What triggers filters, honestly

The word-list theory of spam filtering is mostly dead, and the people who run the filters say so by omission. Google's sender guidelines (February 2024, with unauthenticated mail moving to outright 550 rejection in November 2025) list requirements in full: SPF and DKIM aligned, a DMARC policy published, complaint rate under 0.3%, one-click unsubscribe for bulk senders. Not one banned word appears anywhere in them.

The measurable word effects are small. SpamAssassin publishes its ruleset, so you can read the scores yourself: an all-caps subject costs between roughly half a point and two points depending on the version, against a default block threshold of 5. FREE and currency symbols score similarly. Weak signals, nowhere near fatal alone.

Reputation is the strong signal, and we have the scar to prove how silently it operates. In June 2026 we watched Gmail reject 73% of mail we were mirroring through a forwarding path that broke DMARC alignment. No bounce we could see, no warning, mail simply gone. No subject line survives failed authentication, and no subject line saves a domain the filters already distrust.

Here is the twist that keeps the old advice half-alive: caps, FREE, and dollar signs still hurt you, through humans. They read as mass mail, mass mail gets reported, and reports feed the 0.3% math above. Skip them because they cost trust, not because a filter keyword list exists.

Write the subject last

Draft the body first. Then pull the most specific noun phrase out of it, cut it to under 30 characters if you can, drop it to lowercase if that is how you write, and check it against two questions: is every word true, and would I open this from a stranger?

Then send it inside a volume your domain can carry (how many cold emails per day has our actual caps), and have the follow-up sequence written before the first send, because the follow-up thread reuses this subject and the reply, when it comes, arrives on it.

One good subject, sent 30 times to the right people from a warm domain, beats a brilliant one blasted to 3,000.