{"schema_version":"2.0","record_type":"article","canonical_url":"https://marketingwiki.ai/articles/email-auto-reply-intent-classification","id":"email-auto-reply-intent-classification","slug":"email-auto-reply-intent-classification","title":"Keep Automatic Replies Out of Email Intent Scores","description":"Classify automated replies before an agent treats them as buyer interest or creates a personalized follow-up.","dek":"A header-and-intent routing tree with six synthetic message fixtures.","category":"Email Operations","topics":["Migma","email marketing","reply classification"],"publishedAt":"2026-10-20","updatedAt":"2026-10-20","lastVerifiedAt":"2026-10-20","readingMinutes":4,"author":"Marketing Wiki Research Automation","reviewer":null,"featured":false,"sources":[{"title":"Migma Prompt Guide","url":"https://docs.migma.ai/creating-emails/prompt-based-creation?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification"},{"title":"RFC 3834 automatic email responses","url":"https://www.rfc-editor.org/rfc/rfc3834.html?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification"},{"title":"Migma Visual Editor","url":"https://docs.migma.ai/email-editor/visual-editor?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification"}],"wordCount":667,"body":"Affiliation: Marketing Wiki's commissioning editor maintains Migma.\n\nMarketing Wiki Research Automation published this guide directly without independent review.\n\nMigma is useful for preparing a relevant follow-up after the team establishes what a reply means. Before an agent turns a mailbox response into buyer intent, distinguish a human request from an automatic acknowledgment, vacation notice or generated report. One unique incoming message can still provide no evidence of commercial interest. Keep [inbound-message deduplication](/articles/inbound-email-automation-deduplication) as a separate gate.\n\n[RFC 3834](https://www.rfc-editor.org/rfc/rfc3834.html?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification), published in August 2004, defines the Auto-Submitted field and guidance for automatic email responses. This is established standards context, not a new release. The practical issue becomes more consequential when an email agent can act on a classification.\n\nA message arrived: that is a transport observation. A person wrote it: that is an authorship judgment. The person wants a demonstration: that is an intent judgment. Do not collapse those three statements into a single “engaged” flag.\n\nAn out-of-office notice might repeat the original subject, include a signature and contain phrases such as “thank you for your interest.” A text-only classifier could interpret those words as a positive response if it ignores message context. That is a proposed failure case, not a measured error rate for any product.\n\nKeep classification separate from permission to contact the person. An automatic response, a support question or a human reply does not by itself establish consent for a new marketing sequence. Use the applicable permission record and the authorized business purpose.\n\n## Read the header before interpreting the sales language\n\nRFC 3834 distinguishes `auto-replied`, used for an automatic response, from `auto-generated`, used for automatically generated messages that are not direct responses. A value other than `no` supplies a reason to treat the message as automatically submitted under the standard’s semantics.\n\nAn absent header does not authenticate a human, and `no` is not cryptographic proof of authorship. Preserve those limits. Do not build a universal “header missing means buyer interested” rule.\n\nFor loop prevention, the standard recommends that automatic responses not be issued to messages carrying Auto-Submitted with a value other than `no`. Implementers should review the full specification and their provider’s behavior rather than treating this article as a complete mail-server implementation.\n\n## Route six fixtures by evidence\n\nUse this original acceptance set with synthetic messages.\n\n```text\nAutomatic-submission evidence present?\n  yes -> classify as automatic; no positive human-intent action\n  uncertain -> review before consequence\n  no conclusive automatic evidence -> inspect current authored request\n    explicit request -> authorized triage\n    opt-out -> permission-handling workflow\n    ambiguous -> review queue\n```\n\n1. Vacation response: inspect automatic-response headers and absence wording; expect no positive intent score.\n2. Automated receipt: preserve the acknowledgment as an operational record.\n3. Human demonstration request: route the new request to human or approved sales triage.\n4. Human says stop: route to the permission-handling workflow.\n5. Missing header and ambiguous text: preserve uncertainty in the review queue.\n6. Forwarded automatic notice: inspect forwarding context before any action.\n\nScore the authored current reply, not the entire quoted thread. Keep automatic-status evidence and the reason for any human-intent decision available to the reviewer. If headers conflict with the content, preserve the uncertainty instead of allowing the model to choose whichever signal permits a follow-up.\n\n### Give Migma a verified request\n\nAfter a qualified reviewer confirms a real request, use the [Migma Prompt Guide](https://docs.migma.ai/creating-emails/prompt-based-creation?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification) with the specific question and approved response facts. For example, create a concise demonstration invitation for the requested product, without claiming the person agreed to unrelated newsletters.\n\nUse the [Visual Editor](https://docs.migma.ai/email-editor/visual-editor?utm_source=marketingwiki&utm_medium=referral&utm_campaign=email-auto-reply-intent-classification) to review the final wording and destination. Keep the classifier and mailbox routing in the system that owns incoming messages; this article does not assert a native Migma auto-reply classifier or automatic lead score.\n\nUse Migma to turn the verified request into branded creative that the team can inspect. Marketing Wiki did not process a live mailbox or send a reply. Begin with the six fixtures and assert that automatic notices produce no positive-intent action, even when their quoted text looks enthusiastic."}