What Thousands of Real Pitches Taught Us About What Editors Answer

Most published outreach benchmarks are averages of averages, collected by survey, from people with an incentive to round up. We don’t have that problem: we run our own outreach system, it logs every send and classifies every reply, and this year we put our own pitch copy on trial. Literally, with blind judges.

This is what the data says. Where it embarrassed us, we’ve left that in.

First, an honest baseline: your reply rate is lying to you

Between February and July 2026 our headline reply rate was 7.1%. The human reply rate, after stripping out-of-office messages, unsubscribes, and spam auto-responses, was 5.8%.

That’s an ~18% relative inflation from messages nobody wrote. Every tool dashboard we’ve seen reports the first number. If you’re comparing vendors, or your own campaigns month over month, insist on the second one. We now track three rates for every cohort: any-reply, human-reply, and positive reply (someone actually engaging with the ask).

Positive is the only one that pays invoices.

The motion matters more than the template

The single biggest gap in our data isn’t between subject lines or personalization tricks. It’s between kinds of ask.

Over the same period, one of our pitch motions, guest-post proposals, ran at a 9.1% reply rate with 2.5% positive. Another, pitching inclusions in existing listicles, ran at 5.4% reply and 0.6% positive.

Same system, same quality bar, same senders. A four-times difference in positive outcomes, purely from what was being asked of whom.

What we did about it matters more than the numbers: we paused the weaker motion instead of letting a blended average hide it. If you buy outreach from anyone (agency or in-house team) ask for rates per motion, not the blend. A healthy blend can hide a dead channel indefinitely.

The counterintuitive one: plain asks beat free drafts

Common sense says: reduce the editor’s work, attach the draft, make it a one-click yes.

Our corpus says the opposite. In the only comparison clean enough that we’d defend it statistically, plain asks converted 2.5% positive versus 1.1% for pitches that led with an unsolicited draft. More than double.

Our best explanation, having read a lot of these threads: an unsolicited draft doesn’t read as helpful. It reads as pre-commitment. The sender has already invested, so the reply now carries an obligation. It also quietly signals that the same draft went to twenty other sites. A plain, specific ask respects the editorial gate, and editors guard that gate for a living.

One honest caveat, because our own engineers wrote it into the codebase: this is the only pitch class in our corpus where the data is unambiguous. We treat everything else as hypothesis, not doctrine.

We put our pitch copy on trial, blind

In July 2026 we ran a two-round tournament of pitch structures against our incumbent playbook. Multiple candidate rewrites, judged blind: the judges didn’t know which version was the incumbent.

Every finalist shared one skeleton we now consider settled. The gating question moves into the subject line and first sentence, with an explicit early exit, a “if this isn’t relevant, stop reading here” ramp. Respecting the reader’s time structurally, not rhetorically.

The versions varied on exactly one dimension: how much of the proposed content rides in the first email. Nothing? A teaser? The complete, ready-to-use entry?

The winner, ranked first by both independent judges, put the complete entry in the opening email, formatted in the page’s own style, behind the permission gate. Not a teaser, not a “happy to send more.” The full thing, after asking permission to show it.

Which sounds like it contradicts the previous section, until you notice the mechanism. The plain-ask finding is about unsolicited drafts as a door-opener. The tournament winner gates first: the subject and first line ask, the body delivers only if they’ve kept reading past the exit ramp. Consent, then substance. The failed pattern is substance instead of consent.

What actually kills pitches (from our own audit)

The tournament came out of an internal audit of sent pitches. The kill list, verbatim from our own operation:

• A falsifiable claim, disproven by the recipient’s own page. We shipped a pitch hooked on “none of the eight tools do X offline” while the recipient’s page listed offline support for one of them. That single line torched the thread. Every factual claim now gets machine-checked against the live page before anything ships.
• A 197-word first email. Nobody owes you three paragraphs of reading. Our initials now run a fraction of that.
• Invented specifics. One batch went out with two contradictory made-up pricing details. Confident invented details read fine to the sender and fatal to the recipient who knows the real number.
• The wrong human. Full editorial pitches went to a Customer Success Manager and a Senior Business Analyst. They don’t own the blog; the best case is silence. Role-matching is now enforced by a classifier, not by discipline.

None of these are exotic. They’re what actually happens at volume when nobody is auditing.

Timing: we were doing it wrong at scale

The same audit caught something dumber: 68% of our first-touch emails were going out Wednesday through Friday, and those replied at 6.5%. The ones sent Monday and Tuesday replied at 13.4%. More than double, and the gap held up even when we compared emails inside the same campaign, so it was the timing, not the list. Worse, sends went in bursts of 111 to 137 emails on a single day, alternating with days of nothing, which is exactly how you teach a spam filter to distrust you.

The fix was boring and effective: a cap of ~20 new first-touches per day, Monday to Wednesday, in the recipient’s morning. Follow-ups moved to a day-3 and day-7 cadence, replacing an old sequence that crammed three touches into four days. That reads less like diligence and more like pestering.

While we were in there, we stopped sending to unverified addresses entirely. Verified recipients bounced 1.2% of the time; unverified ones bounced 18.5%, fifteen times more. Every bounce taxes your ability to reach the next real person.

If your outreach volume graph has spikes, your reply rate is paying for them.

What editors’ replies actually look like

Every reply we receive is classified into one of 17 intents before a human touches it. The taxonomy itself tells you what the real inbox looks like: seven intents are commercially positive (interest, acceptance, pricing questions, counter-offers, terms discussions, alternative offers, placement confirmations), three are pure auto-noise (out-of-office, unsubscribe, spam), and the rest are the textured middle: questions, referrals to a colleague, polite declines, “that person left the company.”

Two patterns from that textured middle are worth stealing:

The decline-then-pivot rule. When an editor passes on the idea but not the relationship (a polite “not this one” with the door ajar) we allow exactly one sharper re-pitch, matched to what they actually publish. A second decline settles it. And a policy decline (“we don’t accept contributed content”) gets no pivot at all. Editors remember who respected the first no.

“That person left” is an opportunity, not a dead end. A departed contact triggers finding the replacement politely, later, instead of a soft-close email to a mailbox nobody reads.

The takeaways, if you only skim

1. Track human reply rate and positive rate. Headline rates are ~18% noise.
2. Measure per motion. A blend of 9.1% and 5.4% channels looks fine while half your spend is dead.
3. Don’t lead with unsolicited drafts. Ask plainly (2.5% vs 1.1% in our corpus), gate first, deliver substance after consent.
4. Every claim in a pitch must survive the recipient opening their own website.
5. Under 100 words beats 197. Every time.
6. Send Monday to Wednesday mornings, in small daily batches, follow up day 3 and day 7.
7. One re-pitch after an idea-decline. Zero after a policy-decline.

We publish numbers like these because we think proof beats promises in this industry. It’s the same reason every placement we deliver ships with verified traffic data and a daily live-check. If you want to see what that looks like for your own brand, the baseline below is free and takes us a day.