CloudSpex

Domain health guide

What is a DKIM selector?

A DKIM selector identifies one DKIM key record for a sending system. It lets a domain rotate or use more than one signing key without replacing every record at once.

Use the DKIM Checker

Türkçe

How a selector fits into DKIM

DKIM uses a selector together with the domain to locate a public DNS record. A provider might use a selector such as google or selector1, producing a lookup name like selector1._domainkey.example.com.

The selector is chosen by the sender or email provider. It is not a value to guess from a domain name, and different providers can use different selectors for the same domain.

Where to find a known selector

Use the email provider's setup documentation or the DNS configuration created during provider setup. For mail you are authorized to inspect, a sent message's DKIM-Signature or Authentication-Results headers can also identify the selector.

Do not upload email messages or private headers to a public checker. The CloudSpex tool accepts a domain and one selector, then reads the relevant public DNS record.

Validate one selector at a time

A selector-specific check can confirm whether the public record is available and safely interpretable. It cannot prove that every message from the domain is signed correctly or that every mail provider is configured as intended.

Check the selector used by each sending system, especially after an email-provider migration, DNS change, or key rotation.

  1. Identify the sender or provider you are checking.
  2. Find its documented selector.
  3. Check that one selector against the sending domain.
  4. Review provider setup when the record is missing, invalid, or revoked.
  5. Recheck after a DNS or key-rotation change.

Avoid selector guesswork

Trying common selectors across a domain is not a reliable configuration review. It can miss the actual key and creates unnecessary lookup attempts.

CloudSpex validates one selector per request and does not perform automatic selector discovery or brute-force lookup.

Where the selector appears

Every DKIM-signed message carries a DKIM-Signature header. Two tags in that header tell a receiver where to find the public key: d is the signing domain and s is the selector. The receiver joins them into a DNS name, selector, then _domainkey, then the domain, and looks up a TXT record there.

Nothing else is involved in finding the key. If the record is missing at that exact name, verification fails, no matter how many other DKIM records the domain publishes.

A signature header and the DNS record it points to
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed;
  d=example.com; s=s1; h=from:to:subject:date;
  bh=...; b=...

s1._domainkey.example.com.  TXT  "v=DKIM1; k=rsa; p=MIIBIjANBgkqh..."

Selectors used by common providers

Selectors are chosen by whoever generates the key, so they vary. A few providers use fixed, documented names, which is useful when you need to check a domain without access to its mail. Google Workspace uses google unless the administrator chose another prefix. Microsoft 365 uses selector1 and selector2, published as CNAME records that point into Microsoft's own domain so that Microsoft can rotate the keys.

Many other services generate a selector per customer or per key, often a short code or a date. For those, the only reliable sources are the service's DNS setup page and the header of a real message.

Microsoft 365 publishes its selectors as CNAME records
selector1._domainkey.example.com.  CNAME  selector1-example-com._domainkey.example.onmicrosoft.com.
selector2._domainkey.example.com.  CNAME  selector2-example-com._domainkey.example.onmicrosoft.com.

TXT or CNAME

A DKIM key can be published directly as a TXT record or indirectly as a CNAME that points to a TXT record held by the provider. Both work. The difference is who can change the key.

With a TXT record, the key is yours to manage: rotating it means generating a new pair and editing DNS. With a CNAME, the provider can replace the key behind the name without asking you to touch DNS again. That is convenient, and it also means the record keeps resolving only as long as the provider keeps its side in place. When a service is cancelled, the CNAME is left pointing at nothing, and should be removed.

Why one domain has several selectors

Each system that signs mail for a domain needs its own key, and each key needs its own selector. A typical business domain ends up with one selector for the mailbox provider, one for a marketing platform, one for a help desk and one for transactional mail from the website. This is normal and desirable.

Separate selectors keep the services independent. A key can be rotated or revoked for one service without affecting the others, and a selector that appears in a DMARC report tells you which system sent the message.

Rotating a key without losing mail

Selectors exist to make rotation safe. Because the selector is part of the lookup name, an old key and a new key can be published at the same time under different selectors.

  1. Generate a new key pair at the sending system and choose a new selector for it.
  2. Publish the new public key in DNS under the new selector. Leave the old record in place.
  3. Wait until the new record is visible from public DNS, then switch the sending system to sign with the new key.
  4. Keep the old record published for at least a week. Messages signed with the old key may still be in transit or be re-verified when forwarded.
  5. Remove the old record, or replace its key with an empty p= value, which explicitly marks the key as revoked.
A revoked key: the record stays, the key is empty
s1._domainkey.example.com.  TXT  "v=DKIM1; p="

Key size and long records

The length of a DKIM record depends on the key. A 1024-bit RSA key fits in a single DNS string. A 2048-bit key, which is the current recommendation, does not: its record exceeds 255 characters and must be published as two or more quoted strings inside one TXT record. Receivers join the strings before reading the key.

This is the most common reason a correct key fails verification. A DNS panel that silently truncates the value, or a copy that picks up a line break or a space in the middle of the key, produces a record that looks plausible and does not verify. After publishing, compare the key that public DNS returns with the one the provider gave you, or use a checker that parses it.

Ed25519 keys are much shorter and are supported by DKIM, but not every receiver verifies them yet. Where a system offers both, sign with RSA as well, under a second selector.

Reading the result of a check

A selector check answers a narrow question: is there a usable public key at this name. Four outcomes are possible. The record exists and contains a valid key. The record exists but the key is empty, meaning it was revoked. The record exists but cannot be parsed, usually because of truncation or stray characters. Or nothing is published at that name, which either means the selector is wrong or the record was never created.

None of these says whether mail is actually signed. A perfect record does nothing if the sending system is not configured to sign, or signs with a different selector. That part is only visible in a received message, where the Authentication-Results header shows dkim=pass together with the domain and selector that were used.

The line to look for in a received message
Authentication-Results: mx.example.net; dkim=pass header.d=example.com header.s=s1

Selectors and DMARC alignment

The selector itself plays no part in DMARC, but the domain beside it does. DMARC accepts a DKIM pass only when the signing domain in the d tag matches the domain in the visible From address. A service that signs with its own domain passes DKIM and contributes nothing to DMARC.

When a provider asks you to publish selector records under your own domain, it is setting up aligned signing: the d tag will carry your domain. If a service never asked for DNS records, it is almost certainly signing with its own domain, and its mail depends on SPF alignment alone.

Common mistakes

Most DKIM problems come from a short list of errors, and each has a recognisable symptom.

  1. Publishing the record at the wrong name, for example with the domain appended twice because the DNS panel adds it automatically.
  2. Pasting the key with line breaks or spaces inside it, so it no longer decodes.
  3. Leaving the provider's signing switch off after publishing the record. The record alone does not start signing.
  4. Removing the old selector at the same moment as switching to the new one.
  5. Leaving selectors for cancelled services in DNS, especially CNAME records that now point at names someone else could register.

Frequently asked questions

Is a DKIM selector secret?

No. The selector and public key record are published in DNS, but private signing keys must remain private.

Can a domain use more than one selector?

Yes. Multiple senders or key-rotation stages can use different selectors.

Does a present DKIM record prove mail delivery?

No. It confirms only the relevant public record can be assessed; mail flow also depends on signing and receiver-side evaluation.

How do I find the selector of a message I received?

Open the message source and find the DKIM-Signature header. The s= tag is the selector and the d= tag is the signing domain.

Does every domain have a selector called default?

No. There is no standard selector name. Some systems happen to use default, but most use a provider-specific or generated name.

How often should DKIM keys be rotated?

There is no fixed rule. Rotating once or twice a year is a common practice, and a key should always be rotated when someone with access to the private key leaves or a system holding it is compromised.

Use public checks as a starting point

CloudSpex reads limited public-facing signals. It does not change DNS, website, certificate, or email settings. Confirm the responsible provider and test relevant changes before applying a remediation.

Editorial details

Author and reviewer: CloudSpex. Sources: public technical standards and provider documentation relevant to the topic. Last reviewed: November 10, 2026. We update guides when the product behavior or the underlying guidance changes.