A secure domain does not make every message safe
SPF, DKIM and DMARC check domain authorisation and its relationship to a message. They can make direct domain forgery harder. They do not check whether an invoice is authorised, an attachment contains malware or a real account is now being used by an attacker. Transport encryption protects a connection and makes no such content decision. RFC 9989, sections 1 and 2.2
Start with a specific question: are messages missing? Is your domain being forged? Does a newsletter service need to send correctly? Is the concern confidential content or a suspicious account? The answer determines the checking route. One overall score cannot replace these questions.
The route an email takes
- Compose: A person or application creates a message with a visible
From, recipients, subject and content. - Submit: The mail client or sending service hands it to a submission server. Defined routes include port 587 with STARTTLS and port 465 with implicit TLS; webmail or an API may use another internal entry point.
- Prepare for sending: The system sets the SMTP reverse path and may sign the message with DKIM. Changes after signing can cause verification to fail.
- Find the receiving target: The sending mail server obtains the recipient domain's MX records and attempts SMTP delivery, usually on port 25. Forwarding and gateways can add hops.
- Check and decide: The receiving system may evaluate SPF, DKIM, DMARC, reputation, content and local rules. It decides on acceptance, rejection and further treatment.
- Make available: An accepted message may reach a mailbox, spam or another processing stage. Acceptance proves neither inbox placement nor reading. Users access messages through webmail or IMAP, for example.
Sources: RFC 5598, Internet Mail Architecture, RFC 5321, RFC 6409, section 3.1, RFC 8314, section 3.3.
Schematic explanation · not a measurement
Mail route, DNS queries and protection boundaries
Alice at example.com writes to Bob at example.net. Account protection and operational processes concern the participating accounts throughout the flow.
Transfer the message
- Client or application
Alice creates the message.
-
Submission
Submission and sending systemThe sender queries MX and transport rules in DNS.
-
Server transport; TLS applies per connection
Receiving MXOptional forwarding or a gateway adds another hop and possible changes.
Receive and decide
- Authentication and filtering
The recipient checks SPF and DKIM against the connection and message, plus DMARC alignment and policy. This also requires DNS queries.
-
Evaluate results and local rules
Rejection or acceptanceAcceptance may be followed by further filtering and processing.
-
After further processing
Possible availability in the mailboxSMTP acceptance is not yet evidence of inbox placement.
TLS ends at each intermediary. This does not provide continuous TLS protection from author to reader or an overall security seal.
Complete description of the flow
Alice at example.com creates a message for Bob at example.net. Client or application → submission/sending system → receiving MX → recipient checks → mailbox form the main route. DNS queries branch from the sender to MX and transport rules, and from the recipient to SPF, DKIM and DMARC. An optional “forwarding/gateway” intermediary shows another transport hop and possible changes. TLS applies to each connection; there is no continuous TLS protection bar from author to reader. At the recipient, decisions branch into rejection, acceptance with further filtering and possible availability. An “account protection and operational processes” frame surrounds the participating accounts; it is not a green overall security seal.
The visible address and technical identity are different
On narrow screens, scroll the table sideways.
| Detail | Purpose | Common misunderstanding |
|---|---|---|
| Display name, such as “Accounts” | Human-readable label | May be freely chosen and does not prove a person's identity |
Domain in From | Visible sender domain; DMARC's reference point | Need not be identical to the SMTP reverse path |
MAIL FROM / reverse path | SMTP identity for delivery feedback and the usual SPF path | Often visible as Return-Path after delivery; not a reliable author name |
DKIM d= | Domain that signed | A pass for an arbitrary service provider does not automatically align with From |
| Connecting IP address | Server currently delivering the message | May be a forwarder rather than the original sending system |
One successful aligned SPF or DKIM path suffices for DMARC: its authenticated domain must match the From domain under the chosen alignment mode. A bare spf=pass or dkim=pass without this relationship is insufficient. RFC 9989, section 4.4
For diagnosis, trust only the Authentication-Results from your trusted receiving system. An attacker can supply similarly named fields; position alone does not prove their origin. Interpreting header findings explains the distinction. RFC 8601, section 5
Which measure answers which question?
On narrow screens, scroll the table sideways.
| Measure | What it does | What it does not do |
|---|---|---|
| MX infrastructure | Publishes receiving targets and preferences | Proves neither sending authorisation nor resilience to failure |
| SPF | Checks whether an IP address is authorised for the SMTP domain used | Does not automatically authenticate the visible From domain or content |
| DKIM | Associates signed message parts with a signing domain | Does not prove truthful statements or harmless attachments |
| DMARC | Connects SPF/DKIM with From alignment, policy and reports | Does not prevent lookalike domains, display name deception or misused accounts |
| SMTP TLS | Encrypts one transport hop | Is not end-to-end content encryption |
| MTA-STS / DANE | Can require supporting senders to use TLS and appropriate server authentication | Do not enforce processing by every external sender or guarantee inbox delivery |
| DNSSEC | Authenticates DNS data for validating users | Does not encrypt mail or prevent validly signed misconfigurations |
| ARC | Transmits signed statements from participating intermediaries | Does not create trust by itself or repair a broken original signature |
| BIMI | Enables optional brand display under recipient rules | Is not general evidence of protection or delivery |
| Account protection and receiving filters | Limit access misuse and detect suspicious messages | Configuration alone does not prove complete effectiveness |
| S/MIME or OpenPGP | Can encrypt and sign content | Protection depends on key control, endpoints and actual use, among other factors |
| Organisational approvals | Check matters such as changes to payment details through independent routes | Cannot be replaced by a DNS record |
TLS between two servers may terminate at each intermediary. Plaintext processed there remains accessible to that intermediary. Conversely, encrypted content does not automatically protect metadata and endpoints. Plan according to the actual protection needed, rather than the number of enabled acronyms.
Who is responsible for what?
On narrow screens, scroll the table sideways.
| Role | Typical task | Necessary coordination |
|---|---|---|
| Domain owner or administrator | Registrar access, DNS responsibility, sender inventory and policy | Know all actual sending services and delegations |
| DNS operator | Correct records, signing and availability | Account for the registrar/DS and external MX zones |
| Mail or sending provider | Signing, transport, acceptance and account features | Check custom domain configuration alongside actual sending |
| Receiving administration | Filters, authentication boundary, quarantine and reporting | Integrate gateways and forwarding correctly |
| Users and business teams | Report suspicious activity and independently confirm changes to payment details | An accessible reporting route, suitable authentication and clear approvals |
A DNS provider is not automatically your mail provider. Nor does a mailbox provider necessarily know every newsletter, CRM or form submission service in your organisation. RFC 5598, sections 2 and 4
The order in which to proceed
- Inventory and responsibilities: Record domains, subdomains, receiving systems and every sending application. Include infrequent invoice or alert messages.
- Check reception and accounts: Secure MX, the mailboxes actually used, administrative access, strong authentication and recovery routes. Do not postpone account protection until a DNS project ends.
- Authenticate sending routes: Configure SPF authorisation and DKIM for each service; check your own messages at the recipient for results and alignment.
- Observe DMARC: Maintain the reporting destination, publish a suitable monitoring policy and compare reports with the inventory. The domain check does not receive these reports.
- Choose the policy deliberately: Resolve legitimate failures, assess forwarding and mailing lists, then tighten the policy.
p=rejectis not a mandatory target for every general-purpose mailbox domain. The new DMARC standard, RFC 9989 section 7.4, states explicit interoperability limits for this case. - Harden transport and DNS: Plan MTA-STS/TLS-RPT or DANE according to actual support; coordinate DNSSEC and certificate changes. Errors may delay or prevent delivery.
- Continue operational maintenance: Include new sending services before they start, monitor reports and accounts, remove orphaned DNS references and practise response and recovery procedures.
DMARC enforcement describes policy changes. Alias strategy helps with deliberate allocation and replacement of addresses. Compliance and documentation places the technical contribution in its organisational and legal context.
Domains that should neither send nor receive
Consider the following combination only if no service actually sends with the domain and no reception is needed. This example is invented and has not been published:
example.net. IN MX 0 .
example.net. IN TXT "v=spf1 -all"
_dmarc.example.net. IN TXT "v=DMARC1; p=reject"Null MX declares the absence of reception, SPF -all authorises no sending server, and DMARC requests treatment of failures. These are separate tasks. A custom DMARC record may also affect subdomains; check existing sending there first. Null MX must not coexist with normal MX records. It does not prohibit a web server under the same domain. RFC 7505, sections 3–4, RFC 7208, section 10.1.2
Why sender authentication alone is insufficient
An exact forged domain differs from a similarly spelt domain or a compromised real account. QR codes, calendar invitations and attachments also need content and behavioural checks. The email threats chapter explains these mechanisms, protection limits and useful steps when something is suspicious. The threat reports linked there each describe limited studies; they do not establish an individual attack rate.
The right starting point for your task
On narrow screens, scroll the table sideways.
| Starting question | Where to start |
|---|---|
| We are setting up our own domain. | MX, then SPF/DKIM, DMARC and transport in the sequence above |
| A service does not send reliably. | Interpret header findings, then investigate bounces and the specific service |
| Our domain is being imitated. | Distinguish spoofing, DMARC limits and account protection |
| We are planning a gateway or forwarding. | ARC and gateways, then test the affected transport and sender routes |
| We need sound organisational evidence. | Compliance and dated configuration and operational evidence |
Targeted DKIM setup guides cover the general process, Google Workspace and Microsoft 365. Choose according to the sending route you actually use.
What the Email Security Check actually examines
The Email Security Check examines publicly accessible configuration and limited transport features as a snapshot. It is not an examination of a received message.
On narrow screens, scroll the table sideways.
| Measurement route | Limit of the conclusion |
|---|---|
| SPF and DMARC from DNS | Published authorisation and policy, not SPF/DMARC success for actual messages |
| DKIM selector search | Findings under a limited set of queried names; other freely chosen selectors may be missed; no signature verification |
| SMTP | Up to three equally preferred primary MX targets; EHLO, STARTTLS/TLS and QUIT, no MAIL/RCPT/DATA |
| Measurement location and cache | The primary/fallback probe supplies one selected complete result, not a systematic location comparison; consider measurement time and cache |
| MTA-STS | DNS indication, limited HTTPS retrieval and parser/MX comparison; does not prove every sender cache or actual enforcement |
| TLS-RPT | Published reporting destinations; report generation, receipt, analysis or alerting are not confirmed |
| DNSSEC / DANE | Limited resolver/TLSA findings; no complete independent DNSSEC chain or DANE certificate verification |
| ARC, gateway, BIMI | Unweighted name/DNS indications; no filter effectiveness, chain verification or logo display |
Content, attachments, QR codes, sign-in security, forwarding rules and DMARC XML reports are not analysed here. The check does not continuously monitor your domain. Read results and product weighting in the context of the data collected; a fixed runtime or score is not a security promise. Details: check scope.
Frequently asked questions
Are SPF, DKIM and DMARC enough to stop phishing?
They limit certain forms of domain forgery. Lookalike domains, display names, harmful content and misused real accounts remain separate risks.
Is TLS end-to-end encryption?
No. It protects individual connections. Intermediaries may process messages in plaintext; content encryption has different prerequisites.
Does every domain need a mail server?
No. A domain with no email requirement can declare this through appropriate records. First check form mailers, feedback and subdomains.
Should I set p=reject immediately?
No. First examine sending routes and effects. Quarantine may suit general-purpose mailbox domains; because of indirect routes, the standard advises against a blanket reject target in that case.
Can a DNS check confirm a message's authenticity?
No. That needs the message and trusted verification along its receiving route. Even an authenticated domain relationship does not confirm harmless content.
Does a good result prove compliance?
No. Scope, actual operation, organisation and documentation need separate assessment.
Sources and standards status
As at 27 September 2026: RFC 5321 is the published SMTP standard referenced here; its successor draft rfc5321bis is in the RFC Editor process and is not yet a published replacement. DMARC is specified in RFC 9989, aggregate reporting in RFC 9990 and failure reporting in RFC 9991. ARC under RFC 8617 remains Experimental; the proposal to move it to Historic is a draft. BIMI is also not yet an RFC.
Primary sources: SMTP status, rfc5321bis, DMARC, aggregate reports, failure reports, ARC. The linked technical chapters explain the respective conditions in more detail.
Back to the email security hub; for the next technical starting point, see MX infrastructure.