Short answer: deploy ARC where mail is actually being forwarded or transformed.
That usually means the forwarder, mailing list, or gateway should do the sealing. The final receiver should do validation and trust decisions. A domain owner that only sends direct mail usually does not get much benefit from sealing outbound mail just for the sake of "having ARC."
If ARC still feels abstract, start with What is ARC in email and how does it work? (Authenticated Received Chain explained) and then come back here. This post is the deployment decision guide.
ARC has two operational roles:
RFC 8617 is explicit that Internet mail handlers can be both validators and sealers, and that sealing can be done by intermediaries that relay or modify messages: https://www.rfc-editor.org/rfc/rfc8617.
In real operations, think in terms of where authentication context is lost:
That failure pattern is covered in DMARC forwarding and mailing lists: why SPF and DKIM fail (and how ARC helps).
Deploy ARC as a domain owner when the domain owner's infrastructure also acts as an intermediary, for example:
In those cases, the infrastructure is not just "sending." It is taking custody and passing mail along, so sealing is meaningful.
When the domain owner is only originating outbound mail directly to recipients, ARC sealing is usually low priority compared with:
This is the highest-value ARC deployment target.
Forwarders, mailing lists, and inbound gateways are exactly where original authentication results get separated from final delivery context. Google explicitly asks forwarding services (including mailing lists and gateways) to add ARC headers: https://support.google.com/a/answer/81126.
Yahoo guidance also calls out ARC for forwarding use cases: https://senders.yahooinc.com/best-practices/.
So if the question is "who should seal first?" the default answer is: the intermediary that forwards or modifies the message should seal.
"Both" is correct when both sides have meaningful ARC responsibilities:
Microsoft 365 makes this concrete with trusted ARC sealers. Administrators can designate specific sealer domains and allow preserved results to influence handling when direct SPF/DKIM signals fail after transit changes: https://learn.microsoft.com/en-us/defender-office-365/email-authentication-arc-configure.
That is the key nuance: ARC value appears when sealing and trust policy meet. Sealing alone is not enough, and blind trust is worse than no trust.
Use this quick model:
Example flow:
alerts@example.com -> list.example.net -> mailbox.example.org
At list.example.net:
- evaluate SPF/DKIM/DMARC as received
- add ARC set (AAR/AMS/AS)
At mailbox.example.org:
- validate ARC chain
- apply local trust policy for list.example.net sealer
- combine with current-hop SPF/DKIM/DMARC verdictsIf incident handling is still noisy, use a header-first workflow from DMARC troubleshooting: read Authentication-Results headers and fix alignment failures.
If only one system gets ARC work first, make it the forwarder or intermediary that changes message path/content.
Then make sure receiving policy can validate and selectively trust those seals. That pairing is what turns ARC from "extra headers" into a practical reduction in false DMARC failures.