Introduction
To authorise an external DMARC rua address, the owner of the destination domain publishes a TXT record named <policy-domain>._report._dmarc.<destination-host> with the value v=DMARC1. Receivers look it up before sending aggregate reports to another Organizational Domain, and if it is missing they ignore that rua address.
The rule comes from RFC 9990, Section 4 ("Verifying External Destinations"). This page covers the exact lookup, wildcards, testing with dig, the mistakes that stop reports arriving, and a checklist for moving a client portfolio to a new reporting platform. For the protocol as a whole, start with the DMARC guide.
Key takeaways
- The authorisation record lives in the destination zone (the domain receiving reports), never in the client's zone.
- The name is the policy domain, then
_report._dmarc, then the host part of the rua address. The value must start withv=DMARC1. - A wildcard (
*._report._dmarc.platform.example) authorises reports for every domain at once, which is how reporting platforms usually handle it. - Without a valid record, RFC 9990 says the receiver MUST ignore that rua URI. Nothing bounces and nothing warns you; the reports simply stop.
- Test each destination with
dig TXTbefore you change a client's DMARC record, not after.
When do you need a _report._dmarc record?
You need one whenever a DMARC record sends reports to an address in a different Organizational Domain. RFC 9990 makes the check mandatory when the Organizational Domain of the rua host is not identical to the Organizational Domain where the policy record was found. Same-organisation addresses need nothing extra.
| DMARC record at | rua address | Authorisation needed? |
|---|---|---|
_dmarc.client-a.example |
mailto:dmarc@client-a.example |
No, same domain |
_dmarc.client-a.example |
mailto:dmarc@reports.client-a.example |
No, same Organizational Domain |
_dmarc.client-a.example |
mailto:client-a@rua.platform.example |
Yes, in platform.example's zone |
_dmarc.client-a.example |
mailto:dmarc@msp-brand.example |
Yes, in the MSP's zone |
The same procedure applies to failure report addresses. RFC 9991, Section 5 tells receivers to use the RFC 9990 verification with ruf substituted for rua.
How does a receiver build the lookup name?
The receiver takes the host part of the rua URI, puts _report._dmarc in front of it, then puts the domain from which the policy was retrieved in front of that. It sends a TXT query for the result. In a mailto: address, the host is everything after the @.
For a client record like this:
_dmarc.client-a.example. TXT "v=DMARC1; p=none; rua=mailto:client-a@rua.platform.example"
the receiver queries:
client-a.example._report._dmarc.rua.platform.example. TXT
Two details matter here. First, the destination host is the full host after the @, so rua.platform.example, not platform.example. A record at client-a.example._report._dmarc.platform.example would not match. Second, the leading part is the domain the policy came from. If a subdomain such as mail.client-a.example publishes its own DMARC record with an external rua, the lookup starts with mail.client-a.example. Internationalised domain names are converted to their A-label (punycode) form first.
What value must the record contain?
The TXT record must contain v=DMARC1, and that tag must appear first. A trailing semicolon is optional. The receiver discards any TXT record at the name that fails this test, and the relationship is authorised as long as at least one valid record remains.
client-a.example._report._dmarc.rua.platform.example. 3600 IN TXT "v=DMARC1"
RFC 9990 also lets the authorisation record carry its own rua tag to override the address, but the overriding URI MUST use the same destination host. If it points to a different host, the receiver sends reports to neither address. Most setups never need an override, so a plain v=DMARC1 is the safe choice.
How do wildcard records work for MSPs and report platforms?
A destination that accepts reports for any domain can publish one wildcard TXT record at *._report._dmarc.<destination-host> containing v=DMARC1. RFC 9990 names this option explicitly. It removes per-client DNS work: a new client's reports are authorised the moment their DMARC record goes live.
*._report._dmarc.rua.platform.example. 3600 IN TXT "v=DMARC1"
The trade-off is control. A wildcard means anyone can point their rua at your address and you will receive their reports. Platforms filter unknown domains at ingestion, so this is usually acceptable for them. An MSP using a white-label address under its own domain should decide deliberately:
- Wildcard in your zone: zero per-client DNS changes, but you receive reports for any domain that names your address.
- Per-client records: you authorise exactly the clients you manage, but every onboarding and offboarding needs a DNS change in your zone.
Standard DNS wildcard rules also apply. A wildcard does not answer for a name that exists explicitly in the zone, so if you publish a specific record for one client next to the wildcard, that specific record must itself be valid.
What happens if the record is missing?
The receiver drops that rua URI. RFC 9990 says that where authorisation cannot be confirmed, the URI MUST be ignored by the receiver generating the report. Other valid URIs in the same rua tag still get reports, because receivers attempt delivery to every remaining URI.
That makes failures silent. The client's DMARC record looks fine in every checker, mail flows normally and the only symptom is an empty dashboard. In a portfolio, a domain that normally sends reports and suddenly stops is the signal to check authorisation first. Two related behaviours are worth knowing:
- Temporary DNS errors: a receiver MAY defer and retry the check later rather than drop the URI immediately.
- Caching: after you publish or fix the record, positive caching means the change can take up to the record's TTL to be seen.
How do you test the authorisation with dig?
Build the name yourself and query it as TXT. A correct record returns "v=DMARC1"; an empty answer means receivers will ignore the address. Test from outside your network, or against a public resolver, so you see what a receiver sees.
$ dig +short TXT _dmarc.client-a.example
"v=DMARC1; p=none; rua=mailto:client-a@rua.platform.example"
$ dig +short TXT client-a.example._report._dmarc.rua.platform.example @1.1.1.1
"v=DMARC1"
If the destination uses a wildcard, any client name will answer, so you can confirm the platform's setup before migrating anyone:
$ dig +short TXT test-does-not-exist.example._report._dmarc.rua.platform.example
"v=DMARC1"
Our DNS lookup tool runs the same TXT query in a browser. For a whole portfolio, loop over the client list with dig and flag any empty answer.
How do multiple rua destinations behave?
The rua tag takes a comma-separated list, and RFC 9990 says receivers attempt delivery to every remaining URI, up to their own limits on how many they support. Each external URI is verified separately, so one missing authorisation drops only that address.
v=DMARC1; p=quarantine; rua=mailto:dmarc@client-a.example,mailto:client-a@rua.platform.example
Here the first address needs no authorisation and the second needs a record in the platform's zone. Neither RFC sets a fixed maximum number of URIs, so keep the list short (two is plenty) because receivers may cap it. RFC 7489 allowed a size suffix such as !10m on a URI. RFC 9989 keeps it only as obsolete syntax that reporters should ignore, so leave it out of new records. Commas and exclamation marks inside a URI must be percent-encoded.
What are the most common mistakes?
Most failures come from putting the record in the wrong place or building the name slightly wrong. Check these before anything else:
- Record in the client's zone.
_report._dmarcrecords inclient-a.exampledo nothing. The destination domain owner publishes them. - Wrong destination host. Reports go to
rua.platform.examplebut the record sits underplatform.example. - Missing the policy domain.
_report._dmarc.rua.platform.examplewithout the client name in front matches nothing (unless you meant a wildcard, which needs the*.). - Zone name doubled. In DNS consoles that append the zone, typing the full name creates
client-a.example._report._dmarc.msp-brand.example.msp-brand.example. Enter only the relative part. - Typos in
v=DMARC1.v=DMARC 1,v=dmarc1with other text first, or a different tag first fails parsing. - CNAME conflicts. A CNAME at the authorisation name works if its target holds
v=DMARC1, but a CNAME cannot share a name with a TXT record. Some providers also refuse wildcard CNAMEs. A plain TXT record avoids both problems. - Names too long. If the combined name exceeds DNS length limits, RFC 9990 says authorisation cannot be confirmed. Very long client and destination names can hit this.
What should an MSP check when moving a portfolio to a new reporting platform?
Confirm the new destination is authorised before touching a single client record, then cut over in batches and watch for domains that stop reporting. The order matters because a client record pointing at an unauthorised address loses reports from that day on.
- [ ] Get the platform's exact rua host for each client (some use a per-client local part, some a per-client subdomain).
- [ ] Confirm with
digthat the platform publishes a wildcard or per-client_report._dmarcrecord for that host. - [ ] If you use a white-label address under your domain, publish the wildcard or per-client records in your own zone first.
- [ ] For each client, add the new rua alongside the old one, so both platforms receive reports during overlap.
- [ ] Wait at least one full reporting day and confirm reports arrive on the new platform for each domain.
- [ ] Remove the old rua from each client record, keeping policy (
p=,sp=,np=) unchanged. - [ ] Remove per-client records for departed clients from your zone, and ask the old platform to stop accepting reports.
DMARC AI processes aggregate reports per client domain, and its AI assistant can check a domain's live DNS records, which helps with the per-domain verification step. Multi-tenant DMARC management covers how the wider onboarding and offboarding workflow fits together.
Frequently asked questions
Does the client need to publish anything besides its DMARC record?
No. The client publishes only its normal _dmarc TXT record with the external rua address. The authorisation record belongs to whoever owns the destination domain: the reporting platform, or the MSP if the address sits under the MSP's own domain. That split is the point of the mechanism: the receiving side consents to the reports.
Is a wildcard _report._dmarc record safe?
It is safe in the sense that it exposes nothing and cannot affect mail delivery. Its only effect is that receivers will send you reports for any domain naming your address. Reporting platforms expect that and discard unknown domains. If you want to accept reports only for clients you manage, publish per-client records instead.
Why did reports stop after we moved clients to a new address?
The most likely cause is a missing or misnamed authorisation record for the new destination host. Run dig TXT client.example._report._dmarc.<new-host> for an affected client. An empty answer means receivers are ignoring the address. Fix the record, then allow for TTL caching and one reporting day.
Did DMARCbis change the authorisation rules?
The mechanism itself is the same one RFC 7489 described, now defined in RFC 9990 for aggregate reports and applied to failure reports by RFC 9991. What changed around it is the report size suffix, now obsolete. What changed in RFC 9989, 9990 and 9991 covers the wider updates.
Further reading
- DMARC aggregate reports explained: what arrives once authorisation works.
- RFC 9990 explained: DMARC aggregate reporting for MSPs: the rest of the reporting specification.
- How to generate a compliant DMARC record: the client-side record that names the rua address.
- Multi-tenant DMARC management: why MSPs need it: platform requirements, including report authorisation.
Conclusion
Authorising an external DMARC report destination takes one TXT record, but it has to be in the right zone with exactly the right name: the policy domain, _report._dmarc, then the host after the @. Publish it (or a wildcard) before you point any client at a new address, test it with dig, and treat a domain that suddenly stops reporting as an authorisation problem until proven otherwise.