Add an SPF record in Cloudflare DNS
Setting up SPF on Cloudflare comes down to one TXT record at the root of your domain, holding a line that starts v=spf1 and names the servers allowed to send mail as you.
Cloudflare is a DNS host, not a mailbox provider, for this record. It answers DNS queries for the domain and holds the TXT record, and it does not run the mailboxes the SPF record has to name.
The order below is the record first and the console second, because the pasting takes two minutes and the thinking is all in the line itself.
What the record says, and why it lives in DNS
SPF (Sender Policy Framework) is a published list of the sources that may send mail for your domain. When a mail server receives a message with your domain in the return path, it fetches the TXT record at that domain, checks whether the connecting server’s address is on the list, and writes the result into the message headers for the filters that run afterwards.
That lookup goes to whichever nameservers the registry lists for your domain, and to nowhere else. This is why the record has to be published where your DNS is answered, and why a record saved in the wrong console is invisible to the world, however correct its text.
Sign in to Cloudflare, open the domain, and check the Overview page: it states whether Cloudflare is active for the domain. Cloudflare only answers DNS for the domain once the nameservers, set at the registrar where the domain was registered, point to the two Cloudflare nameservers shown in the dashboard. If the nameservers still point at the registrar or another host, a record added in Cloudflare does nothing, and the SPF record belongs wherever the DNS lives.
Read the record before you paste it
An SPF record is a single string with three parts: the version tag, a list of mechanisms, and a closing catch-all. The version tag is always v=spf1, and a TXT record without it isn’t SPF at all.
Mechanisms come next, separated by spaces, and each one names a source of legitimate mail. include: pulls in another domain’s SPF record, which is how you authorise a mail provider or a newsletter tool: you name the hostname they publish and their list becomes part of yours. ip4: and ip6: name an address or a range directly, for a server you run yourself. a and mx authorise the hosts behind your domain’s own A and MX records.
redirect= hands the whole evaluation to another domain’s record; it’s a modifier rather than a mechanism, it can’t sit alongside an all, and it matters here because some hosts publish a record in that form on your behalf.
Each mechanism can carry a qualifier in front of it. + means pass and is the default, so include: and +include: are the same thing. - means fail, ~ means soft fail, and ? means neutral. The one qualifier that decides anything is the one on the last term.
That last term is all, and it says what to do about a source the mechanisms didn’t match. -all asks the receiver to fail unlisted mail. ~all asks it to accept the message and mark it suspect. ?all says nothing useful, and +all authorises the entire internet, which is worse than publishing no record.
Start with ~all while you’re confirming that the list is complete, then move to -all once a few weeks of ordinary sending have passed without a legitimate source turning up as unlisted. The choice between the two endings has its own page. Of the 316,600,902 domains carrying a grade in the September 2026 census, 84,153,064 (26.6%) end their record in ~all.
Two limits shape what you can put in the list. A domain gets one SPF record: a receiver that finds two TXT records starting v=spf1 treats the domain as having a permanent error, which counts the same as having none.
A record may also trigger at most ten DNS lookups during evaluation, counting each include:, a, mx and redirect= plus the lookups inside any record they pull in; an eleventh lookup fails the whole record. So a second sender goes into the existing record as another include:, and if the count is getting high, the mail provider’s help page usually says which of its hostnames costs the fewest lookups.
The record for Cloudflare
Cloudflare publishes exactly the TXT record entered on the Add record form and nothing beyond it. It does not generate or supply the SPF value itself; that value comes from the mail provider’s own documentation, such as include:_spf.google.com for Google Workspace or include:spf.protection.outlook.com for Microsoft 365.
v=spf1 include:_spf.google.com ~all
Compare that block against what your mail provider’s own setup page tells you to publish. The include: hostname has to match theirs character for character, because it’s a DNS name the receiver will resolve, and a typo resolves to nothing. If you send from more than one service, each gets its own include: in the same line, in any order, before the all.
Enter the record, click by click
- Sign in to Cloudflare and select the domain from the dashboard.
- In the left-hand menu, go to
DNS/Records. - Click
Add record. - Set
TypetoTXT. - In the
Namefield, enter@, which means the domain itself, and do not type the full domain name. - In the
Contentfield, enter the SPF text, starting withv=spf1, with no added quote marks. - Leave
TTLonAuto. - Click
Save.
Type takes TXT. Name takes @, which Cloudflare resolves to the domain itself, so the full domain name never goes in this field. Content takes the plain SPF text starting v=spf1, pasted without surrounding quote marks, since Cloudflare adds its own quoting and hand-added marks can produce a malformed record. TTL stays on Auto.
Before you save, read the value back once. The whole string sits on one line with single spaces between terms, no line break, no trailing space, and nothing after the all. If the form already shows a TXT record beginning v=spf1, you’re editing that one, and the only reason to be adding a new row is that no such record exists.
Some of the mistakes on this form belong to this console rather than to SPF.
- Cloudflare allows only one SPF record per domain. Two
v=spf1TXT records is a broken state for receiving mail providers, so an existing record gets edited to add the new service rather than a second one being created. - The proxy status, shown as an orange or grey cloud on other record types, does not apply here. TXT records are never proxied and carry no cloud toggle.
- Do not add quote marks around the Content value. Cloudflare handles the quoting itself, and a hand-added
"can produce a malformed record. - Cloudflare Email Routing adds its own DNS records for inbound forwarding and does not send outbound mail as the domain, so the SPF record must still name whatever service the domain sends from.
Wait, then check it
A saved record takes a few minutes to a couple of hours to spread. The saved record appears under DNS / Records in the Cloudflare dashboard for the domain.
Once the change has had time to spread, run the free scan against your domain. It fetches the record the way a receiving mail server would, reports whether one is present, whether it parses, how many lookups it costs and which qualifier it ends on, and it checks DKIM and DMARC in the same pass.
If the scan reports no record when you’ve saved one, the order of suspects is the nameserver check at the top of this page, then a second v=spf1 row that has been sitting in the zone for years, then a value the form mangled on the way in. A record that’s present but reports as invalid is nearly always one of the limits above: two records, too many lookups, a redirect= with an all after it, or quote marks that ended up doubled.
What a missing or broken record costs you
Of the 316,600,902 domains carrying a grade in the September 2026 census, 166,497,698 (52.6%) publish no SPF record at all. For any of them, a receiver has no list to check the sending server against, so mail from the real mailboxes and mail from a forger’s server arrive carrying the same amount of evidence. Your genuine messages then depend on the receiver’s other filters, and a share of them land in spam for no reason you can see.
A malformed record ends in the same place. A permanent error reads as no policy, and some receivers treat a domain that publishes a broken policy less kindly than one that publishes nothing.
What SPF does on its own is narrower than people expect. It authorises the server, judged against the envelope sender that the recipient never sees, and the result is a header line rather than an instruction. A forged message that fails SPF can still be delivered, because until a DMARC policy tells receivers what to do with a failure, they’re under no obligation to do anything with it.
SPF is the list, DMARC is what turns the list into an instruction receivers act on, and the SPF guide covers how the two fit together and how to read the result headers on a message you’ve sent yourself.
Figures as of 5 September 2026, from the September 2026 edition of the defaults.exposed census. Census numbers move every month; the current values are on the census data page.