Back to Blog

dns-persist-01: Which CAs Accept It, Which Record to Publish, and What to Read First

dns-persist-01 replaces per-renewal DNS writes with one standing TXT record. Google Trust Services accepts it today, Let's Encrypt has no date, and two record formats are in circulation. The current state, kept up to date.

CertGuard TeamSSL Security Experts7 min read
Dark metal signpost seen from below against a pale evening sky, its arms pointing in different directions and reading North Pole 2517 Miles, New York 3418 Miles, South Pole 9919 Miles and Sydney

dns-persist-01 is an ACME challenge type that lets you prove control of a domain once, with a TXT record that stays put. Your CA checks that record at every issuance. Your renewal job no longer needs write access to DNS.

It is still an IETF draft. One CA runs it in production, the biggest one doesn't, and the record format changed underneath the early adopters in September. This page tracks where things stand. We update it when something moves, and the two deep dives it links to explain the details.

Where it stands on 16 November 2026

The specification is draft-ietf-acme-dns-persist-03, published 15 November. The CA/Browser Forum already allows the method through ballot SC-088v3, which passed in October 2025.

WhoStatusRecord format
Google Trust ServicesIn production since 28 September 2026draft -01, plaintext account URI
Let's EncryptNot deployed, no announced datewill follow the final draft
lego--dns-persist since v5.0.0, GTS workaround in v5.6.0draft -01
Posh-ACMEFull support since 4.35.0, 11 Octoberdraft -02
cert-manager, CertbotNo release with support

Let's Encrypt's last public word on timing is Aaron Gable's, on 25 June: no deployment until the issue about client-computed information in the record was resolved. It was resolved with draft -02 in September. Nobody from Let's Encrypt has given a production date since.

How it works in one minute

With dns-01, the CA hands you a fresh token for every order. Your client writes it to _acme-challenge.example.com, waits for DNS to propagate, and the CA reads it back. Every renewal needs a DNS API key, so that key ends up in cron jobs, CI pipelines and Kubernetes secrets.

With dns-persist-01 you publish one record at _validation-persist.example.com. It names the CA and your ACME account. As long as the record is there and matches, the CA treats the domain as validated for that account. Add policy=wildcard and it covers subdomains and wildcard certificates too. Add persistUntil with a Unix timestamp and it stops counting after that moment.

Deleting the record stops new validations against it. A validation the CA already did can still be reused until that authorization expires, so if you no longer trust the account, deactivate it. A deactivated account can't place new orders. Certificates you already have stay valid until they expire.

Two record formats are in circulation

This is the part that trips people up.

Draft -01, from March 2026, put your account URI in the record as plain text. That's what Google Trust Services accepts:

_validation-persist.example.com. IN TXT "pki.goog; accounturi=https://dv.acme-v02.api.pki.goog/account/12345678; policy=wildcard"

Draft -02, merged on 20 September, replaced that with a hash. The CA publishes an accountHashPrefix in its directory, and your client computes a SHA-256 over the domain, your account key's JWK thumbprint and your account URL. The example from draft -03:

_validation-persist.example.com. IN TXT ("authority.example;"
  " accounturi=https://ca.example/account-hash/"
  "sha-256/5SQm7n6tPh2-PlLbCKGnViTXX5z19SCN4cPGQHSk-kw")

The hash closes a real attack. With a plaintext URI, a fake ACME server could trick you into publishing someone else's account URI, and that account could then get certificates for your domain. The cost is that your account key becomes part of the record. After a rotation the old record keeps validating, but the draft wants you to republish it with the new key. The hashed account URI post walks through the attack and what it breaks in your tooling.

Google has said it "will not make changes unless required by a final IETF standard." Read that as: your GTS record is good until the RFC lands, and then expect to replace it.

Which CA, which client

If you want dns-persist-01 in production today, Google Trust Services is the only option we know of. Use lego 5.6.0 or newer. Google's challenge objects leave out the accounturi field the draft describes, and older lego versions stop with "missing accounturi in challenge object". Version 5.6.0 falls back to the account URI it registered.

Posh-ACME computes the draft -02 hash, which makes it the client to watch once a CA ships the hashed format. Its 4.35.0 release notes say it still works with draft -01 servers, which is what GTS runs, as long as the record already exists or you publish it with the Advanced parameter set of Publish-DnsPersistChallenge.

If you're on cert-manager or Certbot, there is nothing to configure yet. Keep dns-01 running.

Should you switch now?

For most teams, no. Prepare instead: find your ACME account URL and the key behind it, and decide which domains would get a standing record and whether they need policy=wildcard. Our post on the ACME account key covers the first part.

There's one case where we'd use it today. If you issue certificates for customer domains and you're asking each customer to give you a DNS API key, a single record on GTS is a better deal for them. Budget for one round of record changes when the RFC is final.

Don't build anything on the draft -01 format with Let's Encrypt in mind. Whatever Let's Encrypt ships will be key-bound.

DNS-Persist-01 Is Almost Here. Here's What It Actually Changes. is the long explanation: the credential problem it solves, the record fields, wildcards, expiry, and where it does and doesn't fit.

The dns-persist-01 Record Just Changed Shape. covers the September change to a hashed account URI, the attack behind it, and why your account key now matters at DNS time.

DNS vs HTTP domain validation explains how the existing challenge types behave in practice. And because the CA still has to read your record from several networks, multi-perspective validation applies to dns-persist-01 exactly as it does to dns-01.

Frequently asked questions

Does Let's Encrypt support dns-persist-01? Not as of 16 November 2026. Let's Encrypt has not deployed it to production and has not announced a date. Its last stated blocker, binding the record to the account key, was resolved in draft -02 in September.

Which CA supports dns-persist-01 in production? Google Trust Services, since 28 September 2026. It implements draft -01 of the specification, with the account URI in plain text and pki.goog as the issuer domain.

Will my Google Trust Services record still work after the specification is final? Probably not unchanged. Google says it will only change its implementation when a final IETF standard requires it, and the current draft uses a hashed account URI the GTS format doesn't have. Plan to republish the record once the RFC is published.

Do I have to update the record when I rotate my ACME account key? No, not to keep it working. From draft -02 on, the CA has to keep the thumbprint of every key your account has used and accept a record computed with any of them, so an existing record still validates after a rotation. The draft does say you SHOULD republish with the new key soon after. With the draft -01 format Google Trust Services uses, the key isn't in the record at all.

Which ACME clients support dns-persist-01? lego has supported it since v5.0.0 and needs v5.6.0 or newer for Google Trust Services. Posh-ACME supports the hashed draft -02 format since 4.35.0. We know of no released cert-manager or Certbot version that supports it.

Related posts

Never miss an expiring certificate again

CertGuard watches your SSL certificates and domains, and alerts you long before anything expires. Free for your first five domains.