Email deliverability for cold outreach now rests on three DNS records and one number. SPF lists which servers may send for your domain, DKIM signs each message, and DMARC tells receivers what to do when those checks fail and requires the visible From domain to match. Gmail and Yahoo have required all three from bulk senders since February 2024, Microsoft since May 2025, and all of them expect a user-reported spam rate below 0.3% (Gmail recommends staying under 0.1%). Miss any of these and mail goes to spam or bounces.
- SPF authorises servers, DKIM signs the message, DMARC checks that a pass matches your visible From domain.
- Gmail and Yahoo have enforced this since 1 February 2024; Microsoft Outlook.com since 5 May 2025.
- The hard ceiling is a 0.3% spam-complaint rate: three reports per 1,000 delivered emails.
- SPF breaks at 10 DNS lookups (RFC 7208), a limit stacked sales tools hit often.
- The most common hidden failure is misalignment: the tool's domain passes, yours doesn't.
A founder's post on Hacker News in 2026 captured how this feels from the inside: their team concluded that cold email outreach "really doesn't work", even as their own inbox filled with automated pitches the moment they mentioned their product online. Both halves of that are the same problem.
Cheap AI writing pushed outbound volume up, receivers responded by tightening authentication and complaint thresholds, and the senders who never set up their domains properly went first. This post explains the records, the thresholds and the order to fix things in, for Indian B2B teams running outbound to global and domestic buyers.
What changed for cold email in 2024 and 2025?
Gmail and Yahoo turned long-standing best practices into requirements from 1 February 2024, Microsoft followed for Outlook.com on 5 May 2025, and Gmail began ramping up enforcement, including outright rejections, from November 2025.
Before 2024, an unauthenticated or badly aligned domain could still reach the inbox if its content looked clean. Now authentication is checked first.
| Requirement | Gmail (all senders) | Gmail and Yahoo (bulk senders) | Microsoft Outlook.com (5,000+/day) |
|---|---|---|---|
| SPF | SPF or DKIM | Both SPF and DKIM | SPF must pass |
| DKIM | SPF or DKIM; 1024-bit key minimum, 2048 recommended | Both SPF and DKIM | DKIM must pass |
| DMARC | Not required | Published, p=none allowed, must pass with alignment | Published, at least p=none, aligned; p=reject recommended |
| Spam rate | Below 0.3% | Below 0.3% (Gmail advises under 0.1%) | Not stated as a number |
| Unsubscribe | Not required | One-click List-Unsubscribe (RFC 8058) for marketing mail; Yahoo: honour within 2 days | Not in the core rule |
| Other | Valid forward and reverse DNS, TLS, RFC 5322 format | Same | Junk folder first, then 550 5.7.515 rejections |
Two definitions matter for cold outreach:
- Gmail bulk sender: close to 5,000 or more messages to personal Gmail accounts in 24 hours, adding up everything sent from the same primary domain. The status never expires.
- Microsoft: more than 5,000 a day to Outlook.com consumer accounts.
Most B2B cold email goes to company domains hosted on Google Workspace or Microsoft 365, where these consumer rules don't apply word for word, but the same checks run in their filters. Treat the bulk rules as the floor for every domain you send from.
The first mass cold email went out in May 1978, on the ARPANET. Gary Thuerk, a marketer at Digital Equipment Corporation, wanted West Coast researchers to come to product presentations for the new DECSYSTEM-20 machines in Los Angeles and San Mateo. An engineer, Carl Gartley, sent it from Thuerk's account to addresses taken from a printed network directory. The mail program accepted only 320 of them, so the rest spilled into the body of the message, and people who missed it were sent it again.
The reaction set the pattern for the next 48 years. Major Raymond Czahor, who ran ARPANET management at the Defense Communications Agency, called it a flagrant violation of the network's rules and phoned Thuerk's boss to complain. A University of Utah user said it shut down his system. The agency promised action to stop a repeat. Thuerk's spamming career ended there: the receivers, not the sender, decided what was acceptable, exactly as Gmail and Yahoo do now.
Brad Templeton: Reaction to the DEC spam of 1978 (with the original message and memos)
What do SPF, DKIM and DMARC actually do?
SPF authorises sending servers, DKIM proves the message wasn't altered and ties it to a domain, and DMARC checks that one of those passing domains matches the From address the reader sees. Each one is a TXT record in your DNS that receiving servers look up when your mail arrives.
| Record | Example | What it proves | Common mistake |
|---|---|---|---|
| SPF (TXT on brand.in) | v=spf1 include:_spf.google.com include:sendtool.example -all | The sending IP is allowed to send for this domain | More than 10 DNS-querying includes, which makes SPF fail with a permanent error |
| DKIM (TXT on selector._domainkey.brand.in) | v=DKIM1; k=rsa; p=(public key) | The message was signed by a key the domain published and wasn't altered | The sending tool signs with its own domain, so DKIM passes but doesn't align |
| DMARC (TXT on _dmarc.brand.in) | v=DMARC1; p=none; rua=mailto:dmarc@brand.in | SPF or DKIM passed for a domain matching the From address; sets the policy for failures | No rua address, so you never see the reports that show what is failing |
The 10-lookup limit and the end of the record
The SPF limit is written into the standard: RFC 7208 caps the DNS-querying terms (include, a, mx, ptr, exists and redirect) at 10 per evaluation, and anything beyond returns a permanent error.
Indian B2B teams often hit this by stacking Google Workspace, a CRM, a marketing automation tool, a cold-email tool and a helpdesk on one domain. The end of the record matters too: -all tells receivers that unlisted servers are not authorised, while ~all only says they probably aren't.
What is DMARC alignment, and why does it break cold email?
Alignment means the domain that passed SPF or DKIM is the same organisational domain as the From address. A message can pass SPF and DKIM and still fail DMARC if both passes belong to your sending tool's domain rather than yours.
This is the most common hidden failure in outbound stacks. Many sequencing tools send through their own infrastructure, use their own bounce domain (so SPF passes for them), and sign with their own DKIM key unless you add yours.
The message looks authenticated in a quick test, but DMARC sees no aligned pass. Gmail's error code for this is 4.7.32: the From header isn't aligned with the SPF or DKIM domain.
The fix is to publish the tool's DKIM key under your own domain (custom DKIM) and, where offered, set a custom return-path subdomain.
How many domains actually have DMARC set up?
Far fewer than you'd expect. Even among large domains, DMARC was rare until the 2024 rules, and most domains that publish a record still sit at p=none, which reports but doesn't block.
The authors scanned the Alexa top 1 million domains in October 2017 and tested 35 popular email services with forged messages, plus a user study with 913 participants. Only 45.0% of domains had a valid SPF record, 4.6% a valid DMARC record, and 0.6% (5,614 domains) a DMARC reject policy. 33 of the 35 services let at least one forged email reach the inbox, and only 9 showed any visual warning for unverified mail. Limit: this is a 2017 snapshot; adoption rose sharply after the 2024 rules, so use it as the "before" picture.
Read the paper (free)See the data as a table
| Share of Alexa top 1M domains (%) | |
|---|---|
| Valid SPF record | 45.0 |
| Valid DMARC record | 4.6 |
| DMARC set to reject | 0.6 |
Source: Hu and Wang, arXiv 1801.00853 (USENIX Security 2018)
The team scanned DNS in September 2022 and October 2023; in the second scan, 15.9 million of 280.3 million domains had a DMARC record, up from 11.6 million a year earlier. The share of DMARC domains on p=none fell from 67.7% to 55.5%, and of domains that left p=none after 2022, 64.1% moved to quarantine and 21.8% to reject. The share with reporting (rua) addresses fell from 59% to 47%, and about 70% of domains that jumped from p=none to reject had no rua tag, meaning they enforced without watching reports. Limit: the population includes many parked and inactive domains, so the adoption rate understates what you'd see among active business senders.
Read the summary and paper (free)See the data as a table
| DMARC domains with p=none (%) | |
|---|---|
| Sep 2022 | 67.7 |
| Oct 2023 | 55.5 |
Source: Hureau et al., PAM 2024, scan of 280.3 million domains · Two scans by the same team, a year apart.
What spam rate gets your cold email filtered?
Gmail and Yahoo both set the hard line at a 0.3% user-reported spam rate, and Gmail recommends staying below 0.1%. Above 0.3%, Gmail says senders aren't eligible for mitigation help, and they only become eligible again after seven straight days below it.
The rate is spam reports divided by delivered mail, calculated daily (Yahoo measures it against mail delivered to the inbox). It sounds small until you do the arithmetic.
At 1,000 cold emails a day, 0.3% is just three people pressing "Report spam". Cold email carries a far higher complaint risk than opted-in newsletters, because nobody asked for it. Use this to check your own budget:
Your daily spam-complaint budget

Why does cold email that worked in 2023 land in spam now?
It lands in spam because authentication and complaint checks now run before content is judged, and the outbound playbook of 2021–23 (new lookalike domains, high volume, generic copy, no unsubscribe) fails all of them.
The HN thread reflects the result: the founder found cold outreach ineffective, and the replies pointed them to answering questions in the forums where buyers already are.
Authentication gets you considered; relevance gets you read.
- Misaligned tools. Sequencers sending with their own DKIM and return path fail DMARC alignment.
- Too many tools on one domain. SPF breaks at 10 lookups.
- No unsubscribe. Without an easy opt-out, annoyed recipients press "Report spam" instead, which is the one action you can't afford.
- Volume over relevance. Generic AI-written sequences raise complaint rates; three complaints per thousand is the ceiling.
- New domains, no history. Gmail says domains that hadn't sent bulk volume since January 2024 face an accelerated enforcement timetable.
How do you set up a domain for cold outreach, step by step?
Set up a dedicated sending subdomain or secondary domain, authenticate it fully with aligned SPF and DKIM, publish DMARC with reporting, then send low volumes of relevant mail and watch complaints and reports before scaling.
- Separate your traffic. Keep transactional and team mail on your main domain. Send outbound from a subdomain (for example outreach.brand.in) or a clearly related secondary domain, so a bad week doesn't damage invoices and password resets.
- Publish SPF with fewer than 10 lookups. Include only the services that send for that domain and end with -all once you're sure the list is complete.
- Turn on custom DKIM in every tool. Use 2048-bit keys where the provider supports them. Confirm the d= domain in the signature is yours.
- Publish DMARC at p=none with rua. Read the aggregate reports for two to four weeks, fix every failing source, then move to p=quarantine and later p=reject.
- Add one-click unsubscribe. Include the List-Unsubscribe and List-Unsubscribe-Post headers and a visible link, and honour opt-outs within two days.
- Register for postmaster tools. Google Postmaster Tools shows your spam rate and authentication results per domain; Microsoft's SNDS covers Outlook.com.
- Earn the open. Smaller lists, real research per account, and a reason to write. Outbound is now an account-based activity, which is the approach in account-based marketing for Indian B2B.
Authentication gets you considered; relevance gets you read. Once a prospect replies, what you send next decides the deal, which is why I pair outbound with a strong enterprise sales deck, and for Indian SMB buyers often a WhatsApp Business API flow after the first touch. If you'd like this set up for your team, see how I work on the hire page.
What's new
- Gmail ramps up enforcement on non-compliant bulk mail, including rejections Google, Nov 2025
- Microsoft enforces SPF, DKIM and DMARC for Outlook.com senders above 5,000 a day Microsoft, May 2025
- A founder on Hacker News asks how to market when cold email no longer works Hacker News, 2026
- Study of 280 million domains finds DMARC records up from 11.6 million to 15.9 million in a year APNIC Blog, Aug 2024
Questions people ask
Does DMARC p=none actually help deliverability?
Yes, as a requirement. Gmail, Yahoo and Microsoft all accept p=none for bulk senders, and publishing it with reporting lets you find misconfigured senders. It doesn't stop spoofing on its own, so plan the move to quarantine and then reject once reports are clean.
Do the Gmail rules apply if I send fewer than 5,000 emails a day?
The extra bulk-sender rules apply at roughly 5,000 a day to personal Gmail accounts, but every sender must have SPF or DKIM, valid reverse DNS, TLS and a spam rate below 0.3%. Since business filters run similar checks, set up all three records whatever your volume.
Should I buy new domains for cold email?
A separate sending domain or subdomain protects your main domain's reputation, but a brand-new domain has no history and Gmail applies an accelerated enforcement timetable to domains without a sending record since January 2024. Authenticate it fully, start with low volumes and grow slowly.
Why does my email pass SPF and DKIM but fail DMARC?
Because neither pass is aligned with your From domain. Your sending tool is probably signing with its own DKIM domain and using its own return path. Set up custom DKIM under your domain and a custom return-path subdomain.
Sources
- Google Workspace: Email sender guidelines (2024, current 2026)
- Google: Email sender guidelines FAQ (2025)
- Yahoo Sender Hub: Sender best practices (2024)
- Microsoft: Outlook.com postmaster policies, practices and guidelines (2025)
- IETF RFC 7208: Sender Policy Framework (SPF) (2014)
- Hu and Wang: End-to-End Measurements of Email Spoofing Attacks, arXiv / USENIX Security (2018)
- APNIC Blog: The D(M)ARC side of the email reporting system, on Hureau et al., PAM 2024 (2024)
- Hureau, Bayer, Duda, Korczyński: Spoofed Emails, an analysis of issues hindering DMARC deployment, PAM (2024)
- Hacker News: Ask HN, how do you do marketing in the age of slop? (2026)
- Brad Templeton: Reaction to the DEC spam of 1978, with the original message and DCA memo (source for the story)
