Teams running cold email automation at scale are seeing a strange pattern this year: authentication is clean, lists are verified, copy is personalized — and reply rates have still dropped 30-40% since early 2025. The senders aren't doing anything new wrong. Gmail and Outlook are doing something new right, from their perspective. Both mailbox providers have continued tightening bulk sender enforcement well past the 2024 baseline requirements, and most B2B outreach teams are still running a deliverability playbook that was written for the old rules.
This isn't a minor tweak. It's a structural shift in how inbox placement gets decided, and it changes what "good deliverability practice" actually means in 2026. Teams that don't adjust aren't getting flagged as spam in an obvious way — they're getting quietly filtered into tabs and folders nobody checks, which is worse, because the dashboards still show the email as "delivered."
The 2024 bulk sender requirements from Google and Yahoo set a floor: SPF, DKIM, DMARC, a sub-0.3% spam complaint rate, and one-click unsubscribe for high-volume senders. Most cold email automation platforms, OnyxSend included, built compliance into onboarding so customers cleared that floor by default. That was table stakes two years ago.
What's changed since is the weighting behind the floor, not the floor itself. Three shifts matter most for B2B outreach:
1. Domain reputation now aggregates faster across a sender's full domain portfolio. Rotating sending across multiple subdomains used to isolate risk — burn one, keep sending from the others. Both Gmail and Outlook now correlate registrant and infrastructure signals across subdomains and related domains far more aggressively than they did in 2024. A reputation hit on one sending domain propagates to its siblings in days, not months.
2. Engagement recency carries more weight than engagement volume. A list with a 40% historical open rate from six months ago no longer protects a domain the way it used to. Mailbox providers are weighting the last two to three weeks of engagement heavily, which means a domain can look fine in a quarterly report and still be actively throttled based on this week's behavior.
3. "Delivered" and "inbox" have diverged further. Bounce rate — the metric most sequence tools surface first — tells you almost nothing about inbox placement anymore. An email can be accepted by the receiving server (no bounce) and still land in Promotions, Spam, or a quarantine folder the recipient never opens. Teams optimizing against bounce rate are optimizing against the wrong number.
Most of the deliverability advice still circulating in cold outreach tips content was written against the 2024 rules and hasn't been updated for how enforcement actually works now. Three specific practices that used to be safe are now actively risky:
- Warming a new domain for 2 weeks and calling it done. That timeline assumed engagement signals accumulated slowly. With recency-weighted scoring, a domain needs sustained positive engagement much closer to send time, not just a historical warm-up period. We cover the mechanics of this in our cold email warmup playbook, but the timeline itself needs to run longer than it did eighteen months ago — closer to 4-6 weeks for a domain that will carry meaningful volume.
- Buying a block of look-alike domains and treating them as independent risk pools. With cross-domain correlation, a portfolio of five subdomains under one root behaves more like one reputation unit than five. Spreading volume across them still helps with raw sending limits, but it no longer isolates one domain's mistakes from the others.
- Reporting on delivery rate as a proxy for deliverability. If your dashboard's headline number is "98% delivered," it's answering a question nobody in sales cares about. The number that matters is inbox placement rate, and most teams don't have visibility into it at all.
Here's what actually holds up under the current enforcement regime, based on what we're seeing across domains running through our platform:
1. Authenticate at the standard, not the minimum. SPF, DKIM, and a DMARC policy of at least p=quarantine (moving toward p=reject once alignment is confirmed) are non-negotiable. If you haven't set this up correctly, our SPF, DKIM, and DMARC setup guide walks through the configuration end to end.
2. Track inbox placement directly, not as an inference from opens. Open-tracking pixels are increasingly unreliable as a proxy — Apple Mail Privacy Protection and Gmail's own image caching distort open data on both ends. Seed-list testing across major providers gives a real inbox-placement read instead of a guess.
3. Cap daily volume per mailbox and ramp it slowly, even on domains with sending history. A safe ceiling for an established domain is 30-50 emails per mailbox per day for cold outreach; new domains should start under 10 and increase gradually over the extended warm-up window. Multiply mailboxes, not per-mailbox volume, to scale sending — this is the single highest-leverage lever in scaling cold email automation without killing deliverability.
4. Monitor spam complaint rate in near real time, per domain. Given how fast reputation now aggregates across a domain portfolio, weekly review cycles are too slow. A spike on one domain needs a same-day response — pausing sends, reviewing recent copy and targeting, and checking list hygiene — before it drags down the rest of the portfolio.
5. Prune unengaged contacts out of active sequences aggressively. Since recency of engagement now outweighs historical volume, continuing to send to contacts who haven't opened, clicked, or replied in 60+ days actively damages the recency signal for everyone else on that domain. Suppress them, don't just deprioritize them.
6. Treat list quality as a deliverability input, not just a targeting input. Stale or scraped data drives bounces and spam complaints before it ever produces a bad reply. If your last data audit predates this year, it's worth revisiting — we go through the full process in our B2B prospect data audit framework.
There's a version of this story where the conclusion is "cold email is dead, go back to manual outreach." That's not what the data supports. B2B outreach hasn't gotten harder because cold email automation stopped working — it's gotten harder because the bar for doing it correctly moved, and a lot of tooling built for 2024 hasn't moved with it. Manual outreach doesn't escape any of the above; a human sending from Outlook is subject to the exact same reputation math, just at a volume too small to notice the damage until it's done.
The teams weighing an automated SDR replacement against headcount should treat deliverability infrastructure as part of that evaluation, not an afterthought. Automated prospecting only pays off if the emails it generates actually reach an inbox — volume without placement is just a more expensive way to talk to nobody.
Deliverability in 2026 rewards teams that treat it as an ongoing operating discipline rather than a one-time setup task. Authenticate properly, ramp slowly, watch inbox placement instead of delivery rate, and prune your list before it prunes your reputation for you. None of this requires exotic tooling — it requires actually acting on signals most platforms already surface and most teams don't check often enough.
OnyxSend builds domain rotation, per-mailbox volume caps, and real-time spam-rate monitoring into automated prospecting by default, so the deliverability checklist above runs in the background instead of landing on someone's weekly to-do list. If your reply rates have quietly dropped this year and your bounce rate looks fine, deliverability — not targeting or copy — is the first place to look. You can see how the platform handles it on our pricing page.