Build the records that stop mail being sent as you DMARC
SPF, DKIM and DMARC decide whether the world trusts mail claiming to be from you. They are hand-written records where one slip lets a spoofer through. This builds them right and checks the ones you have.
Your domain
SPF: who may send your mail
DMARC: what to do with mail that fails
DKIM: sign your mail
security.txt: how to report a problem to you
Check a record you already publish
Paste an SPF, DMARC or security.txt record and it is read back for the faults that matter.
Questions people ask about DMARC
How does DMARC alignment work?
DMARC passes when SPF or DKIM passes and the domain that check vouched for matches the domain in the visible From line. SPF vouches for the envelope sender, the address bounces go to; DKIM vouches for the d= domain in the signature. A message can pass SPF for a mail provider’s own domain and still fail DMARC, because that domain is not yours.
Relaxed alignment, the default, accepts the same organisational domain, so mail.example.com aligns with example.com. Strict demands an exact match. The builder writes adkim=s or aspf=s only when you choose strict, and leaves them out otherwise.
What does p=none mean, and when should I move to reject?
p=none asks receivers to report on mail that fails and do nothing else, so spoofed mail is still delivered. It is where to start, with a rua address so the aggregate reports reach you. Once those reports show every service that sends as you passing, move to p=quarantine, then p=reject.
The checker marks none and quarantine as notices and reject as good, warns when there is no rua, and points out a pct below 100, which applies the policy to only part of the failing mail and is meant as a step on the way.
Why does my SPF record fail with too many DNS lookups?
SPF allows at most 10 mechanisms that need a DNS lookup: include, a, mx, ptr, exists and redirect. Past 10 the receiver returns a permanent error, and SPF then fails for all your mail. ip4, ip6 and all cost nothing.
The checker counts the lookups written in the record you paste and warns from 8. It cannot see inside an include, because it does not query DNS, and each one usually brings lookups of its own. A record that counts 6 here can still be over the limit once its includes are followed.
What is the difference between -all and ~all?
They say what happens to a sender you have not listed. -all is a fail, ~all a soft fail, which most receivers accept and mark, ?all is neutral and protects nothing, and +all authorises everyone on the internet to send as you.
For DMARC only an SPF pass counts, so a soft fail and a fail both leave DMARC to depend on DKIM. The builder offers -all, ~all and ?all; the checker calls +all a fault, and a record ending in ?all or in no all at all a notice.
Why is my DKIM record split into several quoted strings?
One string in a TXT record holds at most 255 characters, and a 2048-bit RSA public key is about 392 characters of base64 on its own. So the record is published as several quoted strings in one TXT record, and receivers join them back together with nothing in between.
Paste the key your mail system generated, with or without its -----BEGIN PUBLIC KEY----- lines, and the builder strips the armour and line breaks, writes v=DKIM1; k=rsa; p=… for selector._domainkey.yourdomain, and splits it at 255. It always writes k=rsa; for an Ed25519 key, change it to k=ed25519. Some DNS hosts take the value unquoted and split it themselves, so check what yours expects.
Can it check whether my records are actually published?
No. It does not query DNS. Copy the record from your DNS host, or look it up with dig TXT _dmarc.example.com, and paste it into the checker: a record starting v=spf1 is read as SPF, one starting v=DMARC1 as DMARC, and text with a Contact: line as security.txt. DKIM records are built here but not checked.
To see what a receiving server actually decided about a message you sent, paste its headers into HEADERPROOF.
What does a security.txt file need to contain?
Under RFC 9116 it lives at /.well-known/security.txt and needs at least one Contact: line, a mailto: or https: address for reporting a vulnerability, and one Expires: date. Encryption, Policy, Preferred-Languages and Canonical are optional, and the builder writes each one you fill in.
The checker flags a missing Contact or Expires, a date that does not parse, and a date that has passed. The builder fills Expires eleven months from today, inside the year the RFC recommends, and the checker notes a date more than a year away. WELLKNOWN reads an existing file line by line.
Doing this often? The OBSCURA Sovereign Suite keeps checks like this as evidence against the controls of ISO 27001, SOC 2 and Cyber Essentials. On its own, US$2 a month; with everything else, US$5 a month or US$48 a year.