If your open rates on Outlook and Microsoft 365 addresses fell off a cliff sometime this quarter, you're not imagining it and you didn't do anything uniquely wrong. Microsoft blocked more than 3 million outbound sending accounts in Q1 2026 for sending-pattern violations, rolled out in two distinct enforcement waves. Most deliverability advice still floating around dates to 2022 — before Microsoft tightened its complaint threshold, capped new-domain volume, and turned on machine-learning engagement scoring. Here's what actually changed, the exact numbers, and what to do about it this week.
What actually changed: Microsoft's two 2026 enforcement waves
Wave one hit in February 2026: volume-based caps on new and low-reputation sending domains. Wave two landed in April: ML-based engagement-signal flagging that looks past raw volume and complaint counts to how recipients actually behave — opens, deletes-without-opening, moves-to-junk, and reply patterns aggregated across every Microsoft 365 tenant that receives your mail.
The practical effect: a domain that would have sailed through in 2024 on 'good' metrics — low bounce rate, DKIM/DMARC passing, decent open rate — can now get throttled or junked purely because Outlook's model decides your traffic pattern looks like cold outreach at scale, independent of whether recipients complain.
The new numbers: caps, thresholds, and the flag line
These are the figures teams are hitting in the wild this quarter:
| Metric | Old threshold (pre-2026) | 2026 threshold |
|---|---|---|
| New-domain daily send cap (first 30 days) | ~500/day | 200/day |
| Spam complaint rate before throttling | 0.20% | 0.10% |
| Open-rate floor before ML engagement flag | not scored | ~8% |
| Recommended inboxes per sending domain | 5-10 | 2-3 |
| Warmup period before full volume | 2-3 weeks | 4-6 weeks |
Why Outlook is stricter than Gmail right now
Gmail's spam filtering leans heavily on domain and IP reputation built over time, plus Google Postmaster Tools giving senders visibility into their own standing. Microsoft 365 leans more on real-time ML engagement scoring layered on top of reputation — which means Outlook reacts faster to a bad week of sending than Gmail does, but also recovers faster once you fix the underlying behavior.
| Gmail | Microsoft 365 (2026) | |
|---|---|---|
| Primary signal | Domain/IP reputation, historical | ML engagement scoring, real-time |
| Visibility tool | Google Postmaster Tools | Smart Network Data Services (SNDS) |
| Complaint threshold | ~0.30% | 0.10% |
| New domain ramp | Gradual, less rigid caps | Hard 200/day cap, first 30 days |
| Recovery speed after fix | Slower (weeks) | Faster if engagement improves |
The 3-stage flag system
Microsoft 365 enforcement isn't binary — you move through three stages, and each has different symptoms:
- Capped — your domain hits the daily volume ceiling and excess sends queue or bounce with a 4xx soft-fail. Reversible within days once volume drops.
- Junked — mail delivers but lands in the recipient's Junk folder rather than inbox. This is the ML engagement flag in action; it's a scoring problem, not a blocklist problem.
- Suspended — the sending account or tenant is blocked outright, usually after sustained complaints above 0.10% or repeated junk-stage sending with no improvement. This requires a formal reputation review to reverse.
Diagnostic checklist
Run through this before you touch your outreach copy:
- SNDS status: Register your sending IPs in Microsoft's Smart Network Data Services and check for red/yellow reputation flags — this is the closest thing to a real-time answer.
- DMARC alignment: Confirm SPF and DKIM both align with your From: domain, not just pass individually. Partial alignment is a common silent cause of junk placement.
- Inboxes per domain: If you're running more than 2-3 mailboxes per sending domain for cold outreach, you're now over Microsoft's implicit density threshold.
- Warmup age: Any domain under 4-6 weeks old sending above 200/day is very likely already capped, whether or not you've seen a bounce.
- Engagement rate, last 14 days: Pull opens and replies specifically for Microsoft 365 recipients. Below 8% open rate on that segment is your early warning, weeks before complaints accumulate.
The fix: a 2026-safe Microsoft 365 sending setup
- Cap new domains at 200/day for the first month, full stop — don't scale on 'it seems to be working.'
- Run 2-3 inboxes per domain maximum for cold outreach; buy more domains rather than stacking inboxes.
- Extend warmup to 4-6 weeks with gradually increasing volume and forced engagement (replies, not just opens).
- Split volume across Google Workspace and Microsoft 365 sending infrastructure so a single provider's enforcement wave doesn't take out your whole pipeline.
- Segment lists so Outlook/Microsoft 365 recipients get slightly lower daily volume and more personalized subject lines than Gmail recipients — you're now optimizing for two different scoring systems, not one.
- Monitor complaint rate daily, not weekly. At 0.10%, you need to catch a bad batch before it compounds.
Why email-only outbound is the wrong bet this quarter
Even a perfectly configured Microsoft 365 setup is still exposure to a single provider's enforcement decisions — and Microsoft has shown it will change the rules mid-quarter. Teams that paired email with LinkedIn touches and phone follow-up saw response rates lift by up to 287% over email-only sequences, largely because a prospect who junks your email but sees a LinkedIn message from the same rep still counts as reached.
The deeper issue is that email deliverability crackdowns are now a recurring cost of doing business, not a one-time fix. Every quarter that email is your only channel, you're fully exposed to the next Microsoft or Google enforcement wave. Multichannel sequencing isn't a nice-to-have anymore — it's the hedge.
Building a resilient, multichannel motion with AI-scored leads
The teams handling this best aren't just fixing their SPF records — they're rebalancing which leads get email at all. If you can score lead quality before you send, you can route your best-fit, highest-intent prospects to a multichannel sequence (email plus LinkedIn plus a call attempt) and reserve pure high-volume email for lower-priority segments where inbox placement matters less.
This is where a lot of manual list-building falls apart at scale — nobody has time to research and personalize for every lead by hand. Autonomous prospecting tools like [AutoReach](https://www.autoreach.work) address this from the other end: they find and score leads with AI before a single email goes out, then draft A/B-tested, personalized outreach so the emails you do send earn opens and replies instead of junk-folder placement — which matters even more now that Outlook's ML engagement flag rewards exactly that behavior.
If you're also building in-house reply classification or lead-scoring models rather than paying per-seat for a black box, fine-tuning an open-weight model on your own historical outreach and reply data — an approach [InfoPlatform.ai](https://infoplatform.ai) is built around — lets you own the weights and adapt scoring as Microsoft's engagement signals keep shifting.
This quarter's action plan
- Pull your SNDS status and Microsoft 365 recipient engagement rate this week — don't wait for complaints to show up.
- Cap any domain under 60 days old at 200 sends/day and cut inboxes-per-domain to 2-3.
- Extend warmup schedules to 4-6 weeks for any new sending infrastructure.
- Split at least 20% of prospecting volume into a non-email channel (LinkedIn or phone) for your top-scored leads.
- Re-audit DMARC/DKIM alignment, not just pass/fail status.
- Set a daily (not weekly) complaint-rate alert at 0.08% so you catch problems before hitting the 0.10% threshold.
FAQ
Why are my cold emails suddenly going to Outlook junk in 2026?
Microsoft rolled out ML-based engagement-signal flagging in April 2026 that scores your sending pattern based on real recipient behavior — opens, deletes, and junk moves — not just complaint counts. If your open rate on Microsoft 365 recipients falls below roughly 8%, the model treats subsequent sends as low-value bulk mail and routes them to junk, even with clean DKIM/DMARC and low bounce rates.
What is the new Microsoft 365 spam complaint threshold for 2026?
Microsoft cut the complaint threshold from roughly 0.20% to 0.10% — about one complaint per 1,000 emails sent. Crossing it triggers throttling or junk-folder placement on that domain, and repeated crossings can escalate to a full sending suspension.
How many emails can a new domain send per day on Microsoft 365 now?
New or low-reputation domains are capped at roughly 200 emails per day for their first 30 days under the February 2026 volume enforcement wave, down from around 500/day previously. Extending warmup to 4-6 weeks before scaling past that cap is the safest path.
Is Gmail or Outlook stricter for cold email in 2026?
Outlook is currently stricter and reacts faster because it leans on real-time ML engagement scoring rather than purely historical domain/IP reputation, which is Gmail's primary signal. Outlook also recovers faster once engagement genuinely improves, so the fix is to segment sending strategy by provider rather than using one blanket approach.
Does multichannel outreach actually help with Microsoft's email restrictions?
Indirectly, yes — pairing email with LinkedIn and phone touches has been shown to lift response rates by up to 287% over email-only sequences, and it reduces your total exposure to any single provider's enforcement wave. It doesn't fix Outlook deliverability directly, but it means a Microsoft crackdown doesn't take out your entire pipeline.