How many of the most visited domains on the internet tell mail servers to refuse forged mail in their name? Who runs their mail? How many publish no way to receive it at all? On November 16, 2026 we read the email records of the top 10,000 domains on the Tranco research ranking from public DNS, using the record parsers behind our free DMARC checker and MX lookup. This page reports what we found, in totals only, with the method and its limits.
The short version
- 67.8% publish a usable DMARC record, and 51.0% ask receivers to send mail that fails it to spam or refuse it (p=quarantine or p=reject). Among domains whose MX records lead to a mail host, the share asking for that is 61.8%.
- 75.6% publish an SPF record. 51 publish more than one and 31 publish one with a syntax error. Receivers treat both as a permanent error, so SPF cannot pass for those domains.
- Of the 7,205 domains whose MX records lead to a mail host with a public address, 34.5% point at Google Workspace or Gmail and 15.9% at Microsoft 365.
- Only 2.9% publish a valid MTA-STS record, which tells senders that the domain has a policy for encrypted delivery (whether that policy requires encryption is in a file we did not fetch), and 3.3% a valid TLS reporting record.
- 27.1% publish no MX record naming a mail host. A popular domain is not always a mail domain, and an address at one with nowhere to deliver cannot receive mail at all.
Each share is of the domains where that record could be read: of 9,999 domains looked up, 9,954 were measured in full, 34 had at least one lookup that still failed after retries (that record is left out for that domain, never counted as missing), and 11 do not exist in DNS. We publish totals only, not results for any named domain.
DMARC: what the domains ask receivers to do
DMARC is a short text record at _dmarc. plus the domain. It tells a receiving mail server what to do with a message that uses the domain in its From address but passes neither the domain's SPF check nor its DKIM check: deliver it anyway and only report it (p=none), put it in spam (p=quarantine), or refuse it (p=reject). It can also name an address for summary reports (rua=), which is how a domain owner finds out who sends mail in its name.
| What the domain publishes | Domains |
|---|---|
| No DMARC record | 32.0%3,191 |
| A record receivers discard (a wrong version tag, or more than one record) | 0.2%18 |
Monitor only: p=none, or no p= tag, which reads as none | 16.8%1,671 |
An invalid policy value (p=, sp= or np=), or a tag written twice | under 0.1%4 |
Send failing mail to spam: p=quarantine | 19.5%1,942 |
Refuse failing mail: p=reject | 31.5%3,146 |
A published policy is not always applied in full. Of the 5,088 records asking for quarantine or reject, 125 apply it to only part of the failing mail (a pct= below 100) and 1 is in testing mode (t=y), which lowers the policy one step. Of the 6,763 records with the right version tag, 5,827 name a valid address for reports. 14 domains with no usable record of their own are covered by a usable record higher up the domain tree, found by the DMARC checker's tree walk (for example a record published for a whole government suffix).
Our DMARC checker gives each domain one of three verdicts, and we ran its verdict code on every record read here: Looks good for 49.7% (a quarantine or reject policy applied to all failing mail), Worth a look for 18.2% (most often a policy of none), and Needs fixing for 32.1% (most often no record at all).
SPF: who may send, and what happens to the rest
SPF is a text record at the domain itself that lists the servers allowed to send its mail. It ends with an all term that says what to do with mail from any other server: -all fails it, ~all soft-fails it (accept, but treat with suspicion), ?all gives no opinion, and +all passes every server on the internet. A domain must publish exactly one SPF record; two or more is an error, not a backup.
| What the domain publishes | Domains |
|---|---|
| No SPF record | 24.4%2,435 |
| More than one SPF record (an error to receivers) | 0.5%51 |
| One record with a syntax error (an error to receivers) | 0.3%31 |
Ends in -all: other servers fail | 38.3%3,818 |
Ends in ~all: other servers soft-fail | 34.6%3,454 |
Ends in ?all: no opinion on other servers | 1.7%167 |
Ends in +all: every server passes | under 0.1%2 |
No closing all we could read (none, or a redirect that is broken or uses macros) | 0.2%17 |
215 records hand their whole policy to another domain with redirect=; for those we followed the redirect and read the all term there. We did not follow include: terms, so this does not count the DNS lookups a record costs (a record may cost at most 10) or check the records it includes. A record shown here as ending in -all can still be broken for one of those reasons, and our checker reads the whole chain when you check one domain.
Who receives the mail
A domain's MX records name the servers that receive its mail. We read them with the code behind our free MX lookup, which also looks up each mail host's address, so a record pointing at a name that does not exist is not counted as a mail server.
| What the MX records say | Domains |
|---|---|
| MX records lead to at least one mail host with a public address | 72.3%7,205 |
| MX records, but no host a sender can reach | 0.6%59 |
| A null MX: an explicit “this domain accepts no mail” | 0.7%74 |
| No MX records, so senders would try the domain’s own address (its A or AAAA record) instead | 14.5%1,444 |
| No MX records and no usable address: nowhere to deliver | 11.9%1,188 |
Of the 7,205 domains with a reachable mail host, 156 have a problem the MX lookup flags, such as a host name that does not resolve or an MX record that names an alias instead of a host. Hosted mail services name their mail hosts after themselves, so the MX records also show who receives the mail. The lookup recognizes a fixed list of large providers; behind a filtering gateway the mailbox provider is not visible in DNS, so a gateway is counted as itself.
| Provider of the first mail host | Domains |
|---|---|
| Google Workspace or Gmail | 34.5%2,486 |
| Microsoft 365 | 15.9%1,147 |
| Proofpoint (filtering gateway) | 7.0%504 |
| Mimecast (filtering gateway) | 2.0%143 |
| Cloudflare Email Routing (forwarding) | 1.6%113 |
| Amazon WorkMail or SES | 1.5%109 |
| 10 other providers the lookup recognizes | 4.3%309 |
| Not one the lookup recognizes: the domain’s own servers, another provider, or a mix | 33.2%2,394 |
MTA-STS and TLS reporting
Mail between servers can be encrypted, but a sender falls back to plain text when the receiving server does not offer encryption, and an attacker in the path can remove that offer. MTA-STS lets a domain publish a policy that, in its enforce mode, says “only deliver to me over a verified, encrypted connection”; TLS reporting (TLS-RPT) asks senders to report when that fails. Both start with a text record, at _mta-sts. and _smtp._tls. plus the domain. For MTA-STS, that record only announces that a policy exists; the policy itself, and whether it is in enforce mode, sits in a file on a web server.
| Record | All domains | With an MX mail host |
|---|---|---|
| A usable DMARC record | 67.8%6,761 of 9,972 | 83.4%6,009 of 7,204 |
| DMARC asks for quarantine or reject | 51.0%5,088 of 9,972 | 61.8%4,451 of 7,204 |
| One SPF record ending in ~all or -all | 72.9%7,272 of 9,975 | 90.2%6,496 of 7,205 |
| A valid MTA-STS record | 2.9%292 of 9,967 | 4.0%288 of 7,198 |
| A valid TLS reporting record | 3.3%325 of 9,961 | 4.5%323 of 7,193 |
We read the MTA-STS text record only and did not fetch the policy file, so these counts are of domains that announce a policy, not of domains that require encryption. A published record is not proof of a working policy.
What this means if you send email
Never send from a domain you do not control
Receivers check the domain in your From address against that domain's DMARC policy. Here, 51.0% of the domains ask for failing mail to be sent to spam or refused. If you send through an email platform with someone else's domain in the From line, a free mailbox address for example, your mail fails DMARC, and receivers that follow the domain's policy send it to spam or refuse it.
Publish your own records, and check them
One SPF record ending in ~all or -all, DKIM signing from every service that sends for you, and a DMARC record with a reporting address, starting at p=none and moving to quarantine or reject once the reports show only your own services. Gmail's sender guidelines require all three from anyone sending in bulk to Gmail. The free DMARC checker reads yours in a few seconds, including the include chain this study did not follow.
Not every domain can receive mail
27.1% of the domains here publish no MX record naming a mail host, and 74 say outright that they accept no mail. An address at a domain with nowhere to deliver bounces. The free email checker runs a real verification of one address, which includes this DNS check, and the MX lookup shows a domain's mail servers and who runs them.
How we measured
- Which domains. The top 10,000 entries of the Tranco list JZWZY, generated on November 15, 2026. Tranco is a research ranking of popular domains that combines the Chrome UX Report, Farsight, Majestic, Cloudflare Radar and Cisco Umbrella from October 17, 2026 to November 15, 2026, with one entry per registrable domain. 1 entry was not a valid domain name and was left out, leaving 9,999.
- What we asked. For each domain: its MX records and the addresses of up to ten of its mail hosts, or of the domain itself when it has no MX (as the MX lookup does), its text records (for SPF), and the text records at
_dmarc,_mta-stsand_smtp._tls. Where a domain had no usable DMARC record, we walked up the domain tree as the DMARC checker does; where an SPF record ends inredirect=, we followed it. - How we asked. One public resolver, Cloudflare DNS over HTTPS (cloudflare-dns.com), at no more than 40 queries a second. Each attempt timed out after 5 seconds, with up to 3 attempts per question. An answer was reused within the run, so a name many domains share was asked once if it answered. In all, the run sent 67,443 requests between 12:44 and 13:12 UTC on November 16, 2026.
- How we read the answers. With the code of our free tools: the MX lookup's own verdicts and provider list, the DMARC checker's record parser and verdict, and its SPF record parser. A domain whose name does not exist was counted apart.
- Failed lookups. A question that still failed after its retries left that record unmeasured for that domain. Every domain with a record unmeasured was looked up once more; what was still unmeasured after that is left out of that record's shares and never counted as missing.
What this does not show
- DKIM. DNS cannot list the names (selectors) a domain's DKIM keys sit under, so a survey can only guess them, and a guess that finds nothing proves nothing. We did not measure DKIM.
- Popular is not the same as a sender. Tranco ranks domains by popularity (visits, links and DNS lookups), so the list includes content delivery networks, APIs and other infrastructure that may not handle mail at all. Domains whose MX records lead to a mail host are shown apart above for that reason. A domain with no MX record can still take mail at its own address if a mail server answers there, which DNS cannot show, so those domains are not in that group.
- One reading, from one place. DNS changes, and some answers depend on where you ask from. This is what one resolver returned on one day.
- Records, not behavior. A record says what a domain asks for. We did not send mail, connect to any mail server, follow SPF includes, or fetch MTA-STS policy files, so we cannot say how receivers treat any domain's mail in practice.
- Providers by name only. A provider is recognized from its mail host names, for a fixed list of large providers. A domain using its own names for a provider's servers is counted as not recognized.
Citing this research
You are welcome to quote these figures with a link to this page. Please name the date and the list: measured on November 16, 2026 from Tranco list JZWZY. The ranking itself is described in Le Pochat and others, “Tranco: A Research-Oriented Top Sites Ranking Hardened Against Manipulation”, NDSS 2019. We do not publish per-domain results and will not grade or name individual domains.



