Security

DMARC setup for Google Workspace

DMARC setup for Google Workspace requires a valid SPF record, DKIM signing enabled per domain in the admin console, and a DMARC TXT record at _dmarc that starts at p=none while you read reports. Moving to p=reject before reviewing those reports is what breaks legitimate mail.

Records you need
SPFOne TXT record, include Google, under 10 DNS lookups
DKIMEnabled per domain in Admin console, 2048-bit key
DMARCTXT at _dmarc.yourdomain, rua address set
Start policyp=none with reporting for 2–4 weeks
End policyp=reject with pct=100 once senders are clean
Typical rollout6–8 weeks from p=none to p=reject

Get SPF right before you touch DMARC

SPF has a hard limit of ten DNS lookups, and every include statement consumes lookups. Organisations that have accumulated a CRM, a mail marketing platform, an invoicing tool and a helpdesk over several years frequently exceed it without knowing, at which point SPF evaluation fails permanently and silently.

Only one SPF record may exist per domain. Two records is a configuration error, not a redundancy, and it causes a permanent error rather than a fallback. Auditing and flattening the existing record is the first task, before DKIM or DMARC enters the conversation.

  • Count your DNS lookups; ten is a hard ceiling
  • One SPF record per domain, never two
  • Enable DKIM for every domain and alias in the admin console, not just the primary
  • Use a 2048-bit DKIM key where your DNS provider supports the record length
  • Publish DMARC at _dmarc with a reporting address you actually monitor

Read the reports before enforcing

A DMARC record with p=none changes nothing about mail delivery. What it does is instruct receiving providers to send you aggregate reports about every source sending as your domain. Those reports are the inventory you never had, and they routinely surface three to six legitimate senders nobody remembered.

Two to four weeks of collection is enough for most organisations. Seasonal senders such as annual invoicing or recruitment campaigns argue for longer. Enforcement before that inventory exists is exactly how a company blocks its own payroll notifications.

Moving to enforcement safely

The path is p=none, then p=quarantine with a percentage, then p=reject. The percentage tag matters: quarantining ten per cent of failing mail tells you what enforcement will do while limiting the damage if you have missed a sender. Increase in stages, watching reports at each step.

Alignment is the subtlety that catches people. A sender can pass SPF and still fail DMARC because the domain that passed is not the domain in the visible From header. Third-party platforms usually need a custom sending subdomain and their own DKIM key to align properly, which is a configuration change on their side, not yours.

Rollout stages
Weeks 1–3p=none, collect and inventory every sender
Weeks 4–5Fix alignment on third-party senders
Week 6p=quarantine; pct=25
Week 7p=quarantine; pct=100
Week 8p=reject

Why this matters commercially, not just technically

Enforcement stops other people sending invoices in your name, which is the mechanism behind most supplier payment fraud aimed at your customers. That is the security argument. The deliverability argument is separate and increasingly forceful: major mailbox providers now apply authentication requirements to bulk senders, and unauthenticated mail is treated with growing suspicion.

For organisations that send transactional mail, missing authentication shows up as quiet revenue loss long before anyone reports it as a security problem. The mail does not bounce; it lands in spam, and nobody tells you.

What we see that others don't say

The commonest cause of a failed enforcement rollout is not the mail platform at all: it is a marketing, invoicing or recruitment tool sending as your domain without DKIM, which only becomes visible in DMARC aggregate reports after two to three weeks of collection.

What this doesn't cover

  • We do not manage DNS on your behalf unless you ask us to. We supply exact records; the change is made by whoever controls your DNS.
  • We cannot fix alignment on a third-party platform that does not support custom sending domains. Where that happens, the honest answer is to change platform or accept the gap.
  • DMARC does not stop lookalike domain attacks. It only protects the exact domains you publish records for; defensive registrations are a separate exercise.
  • This does not cover inbound phishing protection. Filtering inbound mail is a different control set from authenticating your outbound.

Questions we get asked

What DMARC record should Google Workspace start with?
Start with v=DMARC1; p=none; rua=mailto:your-reporting-address. That changes no delivery behaviour and begins collecting aggregate reports so you can inventory every source sending as your domain.
How long before we can move to p=reject?
Six to eight weeks is typical: two to four weeks collecting reports, one to two weeks fixing third-party alignment, then staged quarantine before reject. Seasonal senders justify a longer collection window.
Why does mail pass SPF but fail DMARC?
Because of alignment. DMARC requires the authenticated domain to match the domain in the visible From header. Third-party senders usually need a custom sending subdomain and their own DKIM key to align.
Do we need DMARC if we only send internal mail?
Yes. DMARC protects your domain from being used by others to send outbound mail to anyone, including your customers and suppliers. Your own sending volume is not the exposure being managed.

How this page is verified

Reviewed by Pearl Lemon Cloud security desk, Workspace security and compliance reviewers. Last checked .

  • SPF ten-lookup limit and permanent error behaviour per RFC 7208 section 4.6.4.
  • DMARC policy tags, alignment and reporting behaviour per RFC 7489.
  • DKIM key configuration steps reflect Google Workspace Admin Help as of 2026-08-06.

Sources you can check

Related pages

Get to p=reject safely

We audit your senders, fix alignment on third-party platforms, and stage enforcement over six to eight weeks so nothing legitimate gets blocked.