Deliverability

AI made everyone a mass emailer. Most inboxes are broken and nobody’s talking about it.

In 2024, Google rewrote the rules for bulk email senders. In the same year, AI outreach tools made sending at scale trivially easy. Most teams did both without understanding what the combination does to a sender reputation. The damage is invisible — dashboards still say green while pipelines quietly dry up.

Chris AylottSep 27, 2026 · 7 min read

Key takeaways

In 2024 Google rewrote the rules for bulk senders — in the same year AI outreach tools made sending at scale trivially easy.
AI outreach tends to produce weaker engagement signals because targeting precision drops when list-building is fast and frictionless, and ISPs score sender reputation on those signals.
A warmup score measures delivery to a controlled seed network, so it can stay green while real inbox placement erodes. A placement test against real inboxes shows the gap.

The pattern that keeps appearing

Something happened to cold email performance in 2024 and 2025 that doesn’t show up in warmup dashboards or authentication checks. Teams that had been running cold email successfully — decent open rates, consistent replies, a functioning pipeline — started seeing those metrics quietly collapse. Not overnight. Gradually. One month a little worse, the next month a little worse again.

The most common diagnosis was copy or targeting. Teams rewrote sequences. Bought new lead lists. Switched SDRs. The metrics kept falling.

The actual problem, in most cases, had nothing to do with the email content. It was the sender reputation — and it had been eroding for months before anyone noticed.

What changed in 2024

Two things happened at almost exactly the same time.

ISPs raised the bar

In February 2024, Google began enforcing new requirements for anyone sending more than 5,000 emails per day to Gmail addresses. DMARC authentication became mandatory. One-click unsubscribe became required. The spam rate threshold was set at 0.1% — exceed that, and Google starts throttling your delivery. Microsoft made parallel changes. The era of “send a lot and hope for the best” ended.

AI tools made volume trivially easy

In the same period, AI prospecting tools matured to the point where a single rep or founder could build a personalised outreach list of thousands in an afternoon. The friction largely disappeared. Building volume no longer required a team or a budget — just a tool subscription and an afternoon.

The combination is what broke things. Higher volumes of outreach, sent by teams who had never had to think about sender reputation at scale, into a filtering environment that is significantly more aggressive than it was in 2022.

The mechanism

ISPs — Google, Microsoft, Yahoo — don’t just check whether your DKIM signature is valid. They build a reputation score for your sending domain based on what happens after your email arrives. Do recipients open it? Reply? Move it from spam to inbox? Or do they delete it unread, report it as spam, or let it sit untouched?

AI-generated outreach tends to produce worse engagement signals — not necessarily because the copy is bad, but because targeting precision tends to drop when list-building is fast and frictionless. Emails that reach the right inbox but from the wrong sender, at the wrong time, for the wrong reason, get ignored. Ignored email is a negative signal. Enough ignored email and your reputation starts to move.

Volume compounds this. One email ignored is noise. Thousands of emails ignored is a data point ISPs take seriously.

The two types of broken

Not everyone who hits this problem got there the same way.

Type 1: The scaled sender

This is the team that was doing cold email successfully before 2024, adopted AI outreach tools to increase volume, and has been slowly losing performance ever since without a clear explanation. Their warmup score looks fine. Their SPF, DKIM, and DMARC are configured. But their emails are landing in spam folders at a rate their warmup dashboard doesn’t reflect — because warmup tools measure delivery to a controlled seed network, not to real Gmail and Outlook inboxes with real filtering behaviour.

Type 2: The returning sender

This is the team that was living on inbound — SEO, content, paid search — and is now pivoting to cold email as those channels become less predictable. They’ve never had to think about deliverability because they’ve never sent at volume before. They pick up an AI outreach tool, set up a domain, start sending, and encounter problems that have nothing to do with their copy.

Both types need the same thing: a placement test against real inboxes with real filtering — not a warmup score from a controlled seed network.

Why warmup scores don’t show you this

This is worth understanding before moving on, because the gap between a good warmup score and broken inbox placement is exactly where the problem hides.

A warmup tool measures whether your emails land in the inboxes of the seed network it controls. Those are not real Gmail or Outlook users with real spam filters, real engagement patterns, and real user-behaviour signals. They are accounts specifically configured to receive warmup emails. They will not filter you based on your domain reputation. They will not ignore your email based on irrelevant targeting. They will not report you as spam.

So a warmup score tells you something real — that your authentication is set up correctly and that your messages can be physically delivered. What it doesn’t tell you is whether Google’s actual filtering, applied to real recipients with real behavioural data about your domain, is letting them through.

Chris Aylott, founder of InboxPlease
“Every dashboard told me I was fine. The open rates told me something completely different. I kept asking the same question: which one is lying? It took me too long to realise they were both telling the truth about different things.”

What this looks like in practice

Broken sender reputation doesn’t announce itself. It shows up as:

Reply rates falling with no change in copy or targeting
Open rates dropping in a way that doesn't correlate with subject line changes
Performance that's significantly worse on Gmail than on other providers, or vice versa
A warmup score that looks healthy while a placement test shows spam folder landing
A pipeline that's gradually drying up with no identifiable cause

The common thread is the dashboard saying one thing while actual results say another. That is the specific gap InboxPlease was built to close.

What to do about it

Three steps, in order.

1. Run a real placement test

Not a warmup score check — a test that sends to real inboxes across Google, Microsoft, Yahoo, and other providers and shows you where your email actually lands. This is what the InboxPlease free test does: inbox placement rates broken down by provider, not a synthetic score from a controlled network.

2. Verify your authentication

SPF, DKIM, and DMARC all need to pass. If you're meeting Google's February 2024 requirements, you have these. If you're not certain, the placement test will surface any gaps. Authentication problems are fixable in hours — reputation problems take weeks.

3. Slow down before scaling back up

If your placement test shows spam folder landing, increasing volume is the wrong response. The mechanism that moved you into spam will keep you there. Reduce volume, let reputation recover — which takes weeks to months depending on severity — and run placement tests as you scale back up. Recovery is real, but it requires stopping the thing that caused the damage.

Find out where your emails are actually landing

Run a free placement test against real Gmail, Outlook, and Yahoo inboxes. Two minutes. No card required.

Run a free test

Written byChris Aylott

Founder & CEO of InboxPlease. Over a decade building email infrastructure for SaaS businesses; writes about deliverability, inbox placement and sender reputation.