What does "554 5.7.5 permanent error evaluating DMARC policy" mean?
The bounce "554 5.7.5 Permanent error evaluating DMARC policy" means the receiving server tried to check the sender's DMARC record, could not get a usable result, and refused the message. In practice the cause is almost always on the sender's side: a malformed, duplicated or misplaced _dmarc TXT record. Fixing that DNS record stops the bounces.
This guide covers what the code means, which record errors produce it, how to confirm the cause with dig, nslookup and Resolve-DnsName, and who fixes what. For background, start with the DMARC guide.
Key takeaways
- The bounce is a permanent rejection (5xx). The receiver will not retry, so the sender must fix the problem and resend.
- No major mailbox provider documents this exact text. It appears to come from receivers whose local policy rejects mail when DMARC evaluation returns
permerror. - The usual cause is the sender's DMARC record:
v=DMARC1not first, two DMARC records, smart quotes, stray characters, or a record published at the wrong name. - A DNS timeout or SERVFAIL should produce a temporary error, not a permanent one. If you see a 554 for a DNS fault, the receiver is classifying it strictly.
- MSPs can check every client's
_dmarcrecord in one pass and catch these errors before a client's customers start reporting bounces.
What do 554 and 5.7.5 mean in the code?
The code has two parts. 554 is the basic SMTP reply defined in RFC 5321: "Transaction failed", a permanent negative reply. 5.7.5 is an enhanced status code from the system defined in RFC 3463, where the first digit 5 means permanent failure and the middle digit 7 means a security or policy problem.
The 5.7.5 part is less precise than it looks. In the IANA SMTP Enhanced Status Codes registry, X.7.5 is "Cryptographic failure": a system "was unable to do so because necessary information such as key was not available or such information was invalid." That definition predates DMARC. The registry has no DMARC-specific code at all. RFC 7372 later added codes for DKIM, SPF and reverse DNS, such as X.7.20 ("No passing DKIM signature found") and X.7.24 ("SPF validation error"), but none for DMARC.
So the receiver has borrowed a generic code. The text after it is what tells you the DMARC evaluation itself failed.
Which receivers send this exact bounce?
We could not find a mailbox provider or mail filter vendor that documents this exact string. The large providers use their own codes: Gmail uses 5.7.26 for DMARC failures, Outlook.com uses 5.7.515, and Microsoft 365 uses 550 5.7.1 when a p=reject policy is enforced. The open-source OpenDMARC filter also uses different wording: its source code rejects with 550 5.7.1 and the text "rejected by DMARC policy for" the domain.
The safest reading is generic. Some receiving systems, often gateways or self-managed mail servers, are configured to reject when DMARC evaluation ends in an error. Cisco's Secure Email gateway, for example, documents permerror as a distinct DMARC check result in its X-Ironport-Dmarc-Check-Result header, with "a syntactically incorrect DMARC record" as its example, and lets administrators choose the action for it.
What is a DMARC permerror, and how is it different from fail?
A DMARC permerror means evaluation broke down, not that the message failed authentication. RFC 9989 registers the result values used in the Authentication-Results header (whose format is defined in RFC 8601), and Section 9.2 defines them:
| Result | RFC 9989 meaning (Section 9.2) | Typical receiver action |
|---|---|---|
pass |
A DMARC record exists and an aligned SPF or DKIM identifier exists | Deliver |
fail |
A DMARC record exists, but no aligned identifier exists | Apply the domain's p= policy |
none |
No DMARC record exists for the Author Domain | No DMARC action |
temperror |
An error "that is likely transient in nature"; a later attempt might produce a final result | Defer with a 4xx reply, or accept |
permerror |
An error "that is unrecoverable"; a later attempt is unlikely to produce a final result | Local policy: accept, quarantine or reject |
The important detail is what RFC 9989 tells receivers to do with a broken record. Section 4.7 says that if v=DMARC1 is not the first tag, "the entire record MUST be ignored". Section 4.10 says that when multiple records are returned "they are all discarded". Section 4.8 says other syntax errors "MUST be discarded in favor of default values (if any) or ignored outright." In other words, the standard treats a broken record much like a missing one. Rejecting the message instead is the receiver's own, stricter choice.
That does not make the bounce the receiver's fault. The sender's record is still broken, and other receivers are ignoring it, so the domain has no working DMARC protection.
What causes a permanent error evaluating DMARC policy?
Most cases trace back to the sender's _dmarc TXT record. Work through the table below in order; the first rows are the most common.
| Cause | What the DNS answer looks like | Side to fix |
|---|---|---|
v=DMARC1 missing or not first |
"p=reject; v=DMARC1; rua=..." or "v=dmarc1; p=reject" |
Sender |
| Two DMARC records | Two separate TXT answers at _dmarc.example.com |
Sender |
| Smart quotes or literal quote marks | "“v=DMARC1; p=reject”" or ""v=DMARC1; p=reject"" |
Sender |
| Wrong separators | v=DMARC1, p=reject, rua=... (commas instead of semicolons) |
Sender |
| Invalid policy value | p=rejected, p=reject. or p=block |
Sender |
| Record split into two TXT records | One record ends mid-value, a second record holds the rest | Sender |
| Published at the wrong name | Record found at _dmarc.example.com.example.com |
Sender |
Broken or looping CNAME at _dmarc |
CNAME to a host that does not exist, or points back to itself | Sender |
| DNS timeout or SERVFAIL | No answer, status: SERVFAIL |
Sender's DNS host (should be temporary) |
| Strict receiver configuration | Record is valid but the receiver still rejects | Receiver |
Why does the order of tags matter?
RFC 9989 Section 4.7 says the v tag "MUST be the first tag in the list" and its value is case-sensitive: DMARC1, not dmarc1. A record that starts with p=reject or uses lowercase dmarc1 is not a DMARC record at all to a conforming receiver. A strict receiver may report that as a permanent error rather than silently ignoring it.
Why do two records break DMARC?
Only one DMARC record is allowed per name. A second record usually appears when a client adds a new one from a vendor's setup wizard without deleting the old one, or when a DNS migration copies records twice. RFC 9989 tells receivers to discard all of them, so the domain is left with no policy. Common DMARC errors covers merging them.
How can a long record be split badly?
A single TXT record can hold several strings of up to 255 characters, and RFC 9989 Section 4.5 says receivers "MUST concatenate these strings by joining together the objects in order." That is fine. The problem starts when a DNS control panel saves the pieces as two separate TXT records instead of one record with two strings. Now there are two records, and the second one does not start with v=DMARC1.
What about DNS timeouts?
A timeout or SERVFAIL is a transient fault, and it maps naturally to temperror. RFC 9989 Section 4.10.1 leaves DNS error handling "to the discretion of the Mail Receiver", which may deliver the message ("fail open") or return a temporary 4yx reply ("fail closed"). A permanent 554 for a pure DNS outage is unusual. If the record is clean and the bounces are intermittent, check the DNS hosting anyway, and see tempfail and 4xx deferrals for the temporary side.
How do I diagnose the error with dig, nslookup or Resolve-DnsName?
Query the TXT record at _dmarc. plus the domain in the From address of the bounced message, then compare the answer with the causes above. Run the query against a public resolver and against the domain's authoritative nameservers, because a stale or split answer on one server explains bounces that only some recipients see.
Using dig (Linux, macOS)
dig +short TXT _dmarc.example.com
A healthy answer is exactly one line:
"v=DMARC1; p=quarantine; rua=mailto:dmarc@example.com"
A broken answer often looks like this, with two records returned:
"v=DMARC1; p=none; rua=mailto:dmarc@example.com"
"v=DMARC1; p=reject; rua=mailto:dmarc@example.com"
To see the response status and whether a CNAME is involved, drop +short:
dig TXT _dmarc.example.com
dig TXT _dmarc.example.com @ns1.example-dns.net
dig +trace TXT _dmarc.example.com
Look for status: NOERROR in the header. status: SERVFAIL points at the DNS host or a broken CNAME chain. status: NXDOMAIN means nothing exists at that name.
Using nslookup (Windows, macOS)
nslookup -type=TXT _dmarc.example.com 8.8.8.8
Non-authoritative answer:
_dmarc.example.com text = "v=DMARC1; p=quarantine; rua=mailto:dmarc@example.com"
If nslookup shows a canonical name = line, the _dmarc name is a CNAME. Follow it and confirm the target returns one valid record.
Using Resolve-DnsName (PowerShell)
Resolve-DnsName -Name _dmarc.example.com -Type TXT -Server 1.1.1.1
Name Type TTL Section Strings
---- ---- --- ------- -------
_dmarc.example.com TXT 3600 Answer {v=DMARC1; p=quarantine; rua=mailto:dmarc@example.com}
Two rows of type TXT mean two records. One row with two items inside the braces is one record with two strings, which is valid. For more on resolvers and authoritative servers, see DNS lookup fundamentals for MSPs.
Is there a faster check?
Yes. Paste the domain into the DMARC checker, which flags a missing v=DMARC1, duplicate records and invalid tag values. If a test message arrived somewhere else, paste its headers into the email header analyzer and look for dmarc=permerror in the Authentication-Results header.
What does a fixed DMARC record look like?
A fixed record is one TXT record at _dmarc.example.com, starting with v=DMARC1, with semicolons between tags and plain ASCII characters only. Here are three common fixes.
Tag order and case corrected:
Before: p=reject; v=dmarc1; rua=mailto:dmarc@example.com
After: v=DMARC1; p=reject; rua=mailto:dmarc@example.com
Separators and policy value corrected:
Before: v=DMARC1, p=quarantine., rua=mailto:dmarc@example.com
After: v=DMARC1; p=quarantine; rua=mailto:dmarc@example.com
Two records merged into one, keeping the policy the client actually wants:
Before: v=DMARC1; p=none; rua=mailto:dmarc@example.com
v=DMARC1; p=reject; rua=mailto:reports@example.net
After: v=DMARC1; p=none; rua=mailto:dmarc@example.com,mailto:reports@example.net
If the second address is on another domain, that domain must authorise it with a _report._dmarc record; see authorising external report destinations. The DMARC generator builds a clean record you can paste straight into DNS, and DMARC record examples for common setups has templates per scenario.
Should I remove old pct, rf and ri tags?
Yes. Removing old pct, rf and ri tags is good hygiene, although they rarely cause this bounce. RFC 9989 removed pct, rf and ri, lists them as historic in its tag registry (Section 9.3), and says in Section 4.7 that "unknown tags MUST be ignored." A receiver following RFC 9989 should therefore skip them rather than fail. Remove them anyway so the record says what it does, and stage enforcement with p=none, then p=quarantine, then p=reject, optionally using t=y to test a step. What changed in RFC 9989 explains the new tags.
Who has to fix a 554 5.7.5 bounce: the sender or the receiver?
The sender fixes it in nearly every case, because the bounce is about the sender's DMARC record. The person who received the bounce (the sender) should forward it to whoever manages DNS for the domain in the From address. That is often the MSP, sometimes a web agency or registrar account the client controls.
| Situation | Who acts | What they do |
|---|---|---|
_dmarc record malformed, duplicated or misplaced |
Sender's DNS administrator | Correct the record, wait for the TTL, resend |
CNAME at _dmarc points to a dead hosted record |
Sender's DNS administrator | Repoint or replace with a TXT record |
| DNS host times out or returns SERVFAIL | Sender's DNS provider | Fix the nameservers; check every authoritative server |
| Record is valid everywhere, one receiver still rejects | Receiver's mail administrator | Check their DMARC error handling and resolver |
| Mail is forwarded and the From header was altered | Forwarder or list operator | See why forwarded mail fails DMARC |
Contact the receiver only after the record checks clean from several resolvers. Then include the full bounce, the time, and your dig output, so their administrator can see the record is valid and look at their own resolver or rule.
How should an MSP triage this error across client domains?
Treat the first 554 5.7.5 report as a reason to check the whole portfolio. Broken DMARC records are silent: most receivers ignore them, so clients only find out when a strict receiver bounces mail.
- Confirm the domain. Take the From domain from the bounce, not the recipient's domain or the envelope sender.
- Query the record. Run
dig +short TXT _dmarc.<domain>against a public resolver and each authoritative nameserver. Exactly one answer starting withv=DMARC1is the target. - Fix under change control. Log the before and after record with client approval, as described in DMARC change management for MSP portfolios.
- Sweep every client. Script the same query across all client domains, or use the domain DMARC scanner, and flag any domain returning zero, two or malformed records.
- Look for a shared cause. If several clients broke at once, suspect a common DNS panel, a template, or a vendor wizard you used during onboarding.
- Watch the reports. A record that receivers ignore (wrong first tag, or duplicates) also stops aggregate (RUA) reports from conforming receivers, so a sudden stop in reports is an early warning.
A multi-tenant platform such as DMARC AI for MSPs processes RUA reports per client domain, which makes the "reports stopped" pattern visible across the portfolio. For UK clients, the NCSC's email security and anti-spoofing guidance is a useful reference when explaining why a working DMARC record matters.
FAQ
Is 554 5.7.5 a soft bounce or a hard bounce?
It is a hard bounce. The 5 in both 554 and 5.7.5 marks a permanent failure, so the sending server will not retry. Fix the sender's DMARC record first, confirm it with dig or the DMARC checker, and then resend the message manually or ask the platform to resend it.
Can a valid DMARC record still cause this bounce?
Rarely. If one clean TXT record starting with v=DMARC1 resolves from public resolvers and every authoritative nameserver, check for a CNAME at _dmarc, intermittent SERVFAIL responses, and SPF or DKIM errors. If all of those are clean, the receiver's configuration or resolver is the likely cause, so contact their administrator with your evidence.
Does p=none avoid the 554 5.7.5 error?
No. The bounce is about a record that cannot be evaluated, not about the strictness of the policy. A malformed record with p=none can trigger it just as easily as one with p=reject. Publishing a correct record fixes it, whatever policy the record contains.
How long after fixing the record will the bounces stop?
Once the TTL (time to live) of the old record expires at the receiver's resolver. That is often an hour or less, but some records carry TTLs of a day. Check the TTL in your dig output, wait at least that long, and then send a fresh test message.
Next steps
Check the domain named in the bounce with the DMARC checker, rebuild the record with the DMARC generator if it is broken, and then sweep the rest of your client domains. If the record is fine and messages still fail, why your DMARC record is not working and the anatomy of a DMARC record cover the next layer. For bounces from the big providers, see Gmail 5.7.26 and Microsoft 550 5.7.515.