Arriving Before Anyone Is Listening

Growing the company got hectic and outbound was the part that frustrated me most. I would send a batch, get nothing back, and I could not work out what was wrong with it. The copy was fine. I read it back a week later and it was still fine.

So I went looking for the reason. The reason everybody gives turned out to be the one part I could not find any evidence for.

The story everyone tells

The story goes like this. AI made outreach basically free, so everyone turned the volume up. Buyers drowned in it, stopped replying, and reply rates fell off a cliff. The sequences that worked in 2019 return nothing now. So outbound is dead.

It is a good story and it explains the feeling exactly. That is probably why it travels. Every part of it that can be checked against a primary source, except the first part, turns out to be unsupported.

I went looking for the decline curve and it is not in any dataset I could open. The largest one I found is Instantly's cold email benchmark, built from what they describe as "billions of cold email interactions across thousands of active workspaces" over calendar 2025. It reports an average reply rate of 3.43%, and it says, plainly, that "reply rates remain stable despite growing volume proving that relevance, not quantity, drives conversations."

Stable. Not collapsing.

The figure I kept seeing quoted for the collapse - a drop from 0.50% to 0.40% over the year - is not in that report, and it is not in any other report I could open either. It is attributed to Instantly by a lot of pages, with methodology detail the report does not contain. It looks like it was invented by a search summary and then repeated by people who never clicked through. I am being a bit honest with ourselves here, because our own first draft of this piece was built on exactly that number, and the draft had to be thrown away.

So the premise is wrong. That is more interesting than if it had been right, because it means the thing making outbound painful is not the thing everyone is blaming, and blaming the wrong thing is how you end up buying a tool that makes it worse.

Outbound is not failing, it is expensive

Gong analysed more than 28 million cold emails and came out with one number that has stayed with me since I read it. The average rep sends 344 cold emails to land one meeting.

Worth being careful about what that number is and is not. Gong does not publish a date range for those emails and does not give a prior baseline, so it cannot tell you anything got worse. It is not a decline. It is a unit cost. Three hundred and forty-four attempts, one conversation, and the other three hundred and forty-three were a stranger interrupting someone who had not asked.

Once you read it as a cost rather than a verdict, the obvious question changes. Not "why did this stop working", which assumes it used to be cheap. The question is what exactly you are paying for.

The buyer was not listening yet

6sense surveyed around 4,766 B2B buyers, with a median purchase value somewhere between $200,000 and $300,000, and the finding they lead with is that buyers "are deliberately ignoring unsolicited outreach until they're ready."

Deliberately. Not accidentally, not because your subject line was weak.

Three numbers from that report do most of the work.

First contact with a seller happens at 61% of the way through the buying journey, which is later than it sounds, most of the deciding is already done by then. Buyers start 79% of those conversations themselves. And the one that does the damage: 95% of the time, the vendor who wins was already on the shortlist the buyer wrote on day one.

Read those three next to the 344 and it stops being mysterious. Most outbound is not being rejected on merit. It is arriving during the long silent stretch where the buyer has decided not to listen, addressed to someone who is going to buy from a name they had already thought of before the process started.

The message was not bad. The moment was.

I think this is the part that is hard to accept, because it removes the comfortable explanation. If reply rates were collapsing you could fix it with better copy, better data, a better tool. If the buyer is not listening yet, better copy does nothing, it just arrives early in nicer language.

Where AI actually does the damage

This is where the popular story gets closest to right and still gets the mechanism backwards.

The assumption is that buyers can smell AI-written outreach and hold it against you. A 2025 study of 261 participants across 990 responses found the first half of that is false. People "are increasingly unable to distinguish AI-generated text from human writing without explicit disclosure." Nobody is catching you.

That sounds like good news and it is not, because the same study found what happens when AI involvement is disclosed. Disclosure "generally erodes perceptions of trustworthiness, caring, competence, and likability, with the sharpest declines in social and interpersonal writing."

There is a number on the sincerity cost too. Cardon and Coman, writing in the International Journal of Business Communication, studied 1,100 professionals and found that only 40% to 52% of readers judged a heavily AI-assisted message sincere, against 83% for a message written with little AI help. That study measured supervisor-to-employee email rather than sales outreach and I am not going to pretend it measured mine. Take it as directional. Outreach is still interpersonal writing sent to somebody deciding whether you are worth their attention, and interpersonal writing is exactly the category where the study found the damage concentrates.

Put the two findings together and the risk is not the one people plan for. It is not that a prospect reads your message, spots a machine, and deletes it in irritation. They mostly cannot spot it. The risk is that the penalty is delayed and it runs backwards. It lands on the day it becomes obvious - the second identical message from a different sender, the reply that does not follow the conversation, the moment on a call where it clicks - and then it reprices every message you ever sent that person.

You do not get caught. You get repriced, backwards, across everything already sent.

So volume is not free. It is a loan against trust, and you do not pick the repayment date.

What I would do instead

If the constraint is timing rather than reach, more sending cannot fix it. The industry's current answer is the same outreach generated faster, which solves the wrong variable and pays for it with the one asset outbound still has. The alternatives are unglamorous and slow, and I would rather say that than dress them up.

Be on the day one list. 95% of deals are won there. That is not an outbound motion at all, it is being known before the buyer starts looking. It is also why content and network compound while sequences do not, you rebuild a sequence every quarter and the sequence never gets more valuable.

Watch for the 61%, do not guess at it. The moment a buyer starts looking leaves traces. A role change, a funding round, a post about the problem, someone coming back to your site for the third time. Outreach that lands inside that window is not an interruption. Everything outside it is you paying the 344.

Spend the trust on purpose. If sincerity is the asset being drawn down, the test for any message is whether you would still defend it if the recipient learned exactly how it was written. That is a hard test and it rules out most of what is currently being automated, including things I have automated myself.

None of this is an argument against using AI for outreach. It is an argument that the useful job for AI is deciding when and whether, not producing more of what is already arriving too early. That is the part we have been building at Linkenite, an agentic layer that watches for the moment and holds the send until there is a reason, rather than a faster way to write the same email. It is also why it has to be inspectable. You are the one who repays the loan, so you should be able to see what was sent in your name and why it went out on that day.

344 attempts per meeting is not a volume problem you can solve with more volume. It is the price of arriving before anyone is listening.

But yeah, nothing extraordinary otherwise. Read the sources, they are all linked, including the one number in this whole area that everybody quotes and nobody can find.