Fix it now
The certificate the mail server presented did not validate, and Outlook stopped before anything else could happen. 0x80072F0D says the certificate authority behind it is unfamiliar; 0x800B0109 says the chain terminated in a root this machine does not trust. Clicking past the warning is not a fix.
w32tm /query /status
certutil -verify -urlfetch server.cer
- Open the failing server name in a browser on the same machine and read the certificate: issuer, expiry, and the names listed under Subject Alternative Name.
- Check the clock. A machine outside the certificate’s validity window fails validation with nothing else wrong.
- Confirm the name Outlook is asking for actually appears on the certificate. Both the mail host name and the autodiscover host name normally need to be there.
- Read the chain and revocation results at the end of the certutil output.
- If an intermediate is missing, install it on the server rather than on every client. Clients build the chain from what the server sends plus what they can fetch.
If a corporate inspection appliance is re-signing this traffic, its root has to be trusted on the client. That is a separate problem, and it presents here first.
If a machine that has never connected before now connects cleanly, you are done. The next section covers which of the three checks failed.
Why it happens
TLS validation asks three questions in order. Does this certificate chain from the server’s own certificate to a root that this machine trusts. Does the name being requested appear in the certificate. Is the certificate inside its validity window and not revoked. One no is fatal, and the client stops before authentication, before the mailbox, before anything.
0x80072F0D is the transport layer’s answer to the first question: WinINet 12045, the function is unfamiliar with the certificate authority that generated the server’s certificate. 0x800B0109 is the chain engine being precise about the same thing: CERT_E_UNTRUSTEDROOT, a certificate chain processed but terminated in a root certificate which is not trusted by the trust provider. Both lead to the same three answers – install the missing intermediate, distribute the root properly, or replace the certificate with one from a publicly trusted authority.
0x80072F06 is the name check rather than the trust check: WinINet 12038, the SSL certificate common name is incorrect. The chain is sound and the dates are fine, but the host name being requested is not among the names the certificate covers. This is very common straight after a migration, when clients begin using an autodiscover host name that was never added to the certificate.
Expiry is worth separating out because it has its own code and its own predictability. Microsoft publishes CERT_E_EXPIRED, 0x800B0101, as a required certificate not being within its validity period when checked against the system clock or the time stamp in the signed file. If everything worked until one particular day and nobody changed anything, check dates on the server certificate and on every certificate in its chain before you look anywhere else.
The chain is incomplete because an intermediate is not being sent
You have this one if The certificate validates on some machines and not others, and the chain check fails to reach a trusted root.
- Obtain the intermediate bundle from the issuing authority.
- Install it into the server’s Intermediate Certification Authorities store.
- On Exchange, confirm the certificate is bound to the right services and rebind it if needed, then restart the web service.
- Re-run the chain check from a client that was failing.
Some clients cache an intermediate they fetched earlier, which is why the same certificate looks perfect on the administrator’s machine and broken on a new one. Test from a machine that has never connected.
The host name Outlook wants is not on the certificate
You have this one if 0x80072F06, and the warning names a host the certificate does not cover.
- List the Subject Alternative Name entries on the current certificate and compare them against the names clients actually use.
- Reissue the certificate with the missing names added, then rebind it to the services.
- Alternatively, change the published URLs so clients only ever ask for names the certificate does cover.
- Retest from a client, not from the server itself.
The certificate or an intermediate has expired
You have this one if Everything worked until one particular day and no change was made on your side.
- Check the validity dates on the server certificate and on every certificate in its chain.
- Replace whichever has expired and rebind services to the new certificate.
- Confirm no service is still bound to the old thumbprint before removing it.
- Put a renewal reminder in place; this is the one cause that is entirely predictable.
An internal certificate authority root is not deployed to the client
You have this one if Domain-joined machines connect without complaint and everything else fails.
- Distribute the internal root through Group Policy or your device management platform.
- For a handful of machines, import the root into the local machine’s Trusted Root Certification Authorities store by hand.
- For anything reached from outside the network, use a publicly issued certificate instead; internal roots do not scale to devices you do not manage.
Full reference
Who is affected tells you where to look
| Who fails | Most likely cause |
|---|---|
| Every client at once | The certificate expired, was replaced, or is missing an intermediate |
| Only unmanaged or personal devices | An internal root is not distributed to them |
| Only one machine | Its clock, its trust store, or local software intercepting TLS |
| The warning names a host you did not type | A name missing from the certificate, or a redirect |
| It started on a specific date with no change from you | A certificate or an intermediate reached expiry |
Which check each code represents
| Code | Constant | Published text |
|---|---|---|
0x80072F0D |
ERROR_INTERNET_INVALID_CA | The function is unfamiliar with the Certificate Authority that generated the server’s certificate |
0x80072F06 |
ERROR_INTERNET_SEC_CERT_CN_INVALID | SSL certificate common name (host name field) is incorrect |
0x800B0109 |
CERT_E_UNTRUSTEDROOT | A certificate chain processed, but terminated in a root certificate which is not trusted by the trust provider |
0x800B0101 |
CERT_E_EXPIRED | A required certificate is not within its validity period |
0x800B0114 appears in the wild alongside these and is widely quoted as a name mismatch, but no Microsoft page publishes a meaning for the value. Diagnose a name mismatch from 0x80072F06, which is documented, and from the names on the certificate itself.
What not to do
Do not fix a certificate error by turning validation off, and do not install anything into Trusted Root Certification Authorities unless you know exactly who issued it and why it is there. The first turns a visible warning into an invisible exposure. The second gives whoever controls that root the ability to impersonate any site to this machine, permanently, to every application on it and not just to Outlook.
Checking a chain properly
- Export the server certificate from the browser to a file.
- Run the chain check with revocation fetching enabled from an elevated prompt.
- Read the chain section: every certificate from the leaf to the root should be listed, each one issued by the next.
- Read the revocation section: a failure to reach the revocation endpoints is itself a network problem worth fixing, because it will bite intermittently.
- Repeat from a machine that has never connected to the server before, because a cached intermediate hides exactly this fault.
When the certificate is provably fine
- Check whether an endpoint security product on the failing machine is intercepting HTTPS. Its re-signed certificate will not be trusted unless its root has been deployed.
- Check the machine’s own clock, and check it in UTC rather than local time.
- Check whether the machine has an out-of-date root certificate store; a machine that has not had updates for a long time will not know newer roots.
- Check whether the failing name is being reached through a redirect you did not expect, in which case the certificate you are inspecting is not the one being validated.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
0x80072F0D |
WinINet 12045 ERROR_INTERNET_INVALID_CA: the function is unfamiliar with the certificate authority that generated the server’s certificate | Microsoft Learn |
0x80072F06 |
WinINet 12038 ERROR_INTERNET_SEC_CERT_CN_INVALID: the SSL certificate common name is incorrect | Microsoft Learn |
0x800B0109 |
CERT_E_UNTRUSTEDROOT: a certificate chain processed, but terminated in a root certificate which is not trusted by the trust provider | Microsoft Learn |
0x800B0114 |
Seen alongside certificate name failures; Microsoft publishes no meaning for this value | not published by the vendor |
Confirm the fix worked
- Outlook connects with no certificate prompt on a machine that was previously showing one.
- The browser on that machine shows the expected issuer, expiry and host name.
- The chain check reports a complete chain and a successful revocation check.
- A machine that has never connected before, ideally off the corporate network, also connects cleanly.
- The renewal date is recorded somewhere a person will see it before it arrives.
Questions people ask about this
Do I have to buy a certificate to fix this?
For a mail server reached from the internet, a publicly trusted certificate is the practical answer, because you cannot distribute an internal root to devices you do not control. For a purely internal service with managed devices only, an internal certificate authority is legitimate and costs nothing beyond running it.
Can I just tick the option to stop asking?
Some clients let you, and it is a poor trade. Once the client and the user are trained to ignore certificate warnings, you have removed the only signal that separates a misconfiguration from someone intercepting the connection.
Do I need a separate certificate for each host name?
No. One certificate can carry several names in its Subject Alternative Name list, and a wildcard covers a level of subdomains. Listing the mail host and the autodiscover host on one certificate is normal.
Why does it work for me and not for a new starter?
Almost always a cached intermediate. Your machine fetched one at some point and now completes the chain without help. A machine that has never connected has to build the chain from what the server sends, which is why the fix belongs on the server.
Will an Office or Windows licence change fix this?
No. This is entirely a server certificate and client trust problem. No licence, edition or subscription level affects certificate validation.
