Should I tell email customers they're talking to an AI bot?

Yes. Saying up front that a reply was AI-drafted avoids backlash if the customer finds out later, and some laws require it. AskVault adds no disclosure by default, so add one to your email signature in the Email channel's settings.

Do regulations require disclosing that an email reply is AI-generated?

Some jurisdictions do: notably California's Bot Disclosure Act and the EU AI Act require disclosure in consumer-facing AI interactions. Confirm with your legal team. AskVault does not add a disclosure for you; put one in your email signature if you need it.

Email autoresponder etiquette

Written by Joseph Russo, Principal Architect, Networks and Tech, AskVault · Reviewed by Allison Chan

Last updated: June 18, 2026 · 3 min read

Rule 1: identify as bot

Customers trust bots that disclose. Avoid lying.

Good footer:

Sent by your AI assistant. Reply for human help.

Bad footer:

[no indicator; impersonating a human]

Disclosure avoids backlash if the customer realises later, and some jurisdictions require it.

Rule 2: respond fast

Email is high-tolerance channel; customers expect under 4 hours normally. Bot beats that:

Fast first reply. AskVault checks the inbox every 60 seconds and answers on the next check.
Compare: human auto-reply within 4 hours; full human reply within 24 hours.

A quick first answer sets the tone for the rest of the thread.

Rule 3: keep replies concise

Email allows long replies, but concise wins:

Under 300 words for routine answers.
3 to 5 sentences is the sweet spot.
Lead with the answer, then context.
End with a clear next step ("Need more help? Reply or visit our help center.").

Long bot replies feel auto-generated. Concise feels human.

Standard footer:

[Bot reply]

AskVault Support [email protected] | yoursite.co Sent by AI assistant. Reply for human help.

Set the signature in the Email channel's settings.

Rule 5: offer human escape

Every bot reply should make escalation easy:

Was this helpful? Reply YES, or reply with "agent" to talk to a human.

Some customers will always prefer a person. Make that path frictionless.

Avoiding spam triggers

Bot emails sometimes land in spam. Reduce risk:

SPF/DKIM/DMARC configured on your domain.
Avoid spam-triggering words in subjects (free, urgent, win, $$$).
Don't send to invalid addresses (bounce rate impacts sender reputation).
Vary message-IDs and timestamps (built-in; just ensure no rapid-fire bot sending).
Warm up new domains by sending low volume first.

Standard email deliverability practices.

Subject-line conventions

For bot replies:

Prefix with "Re:" automatically.
Match original subject if exists.
Don't include "Auto-reply" explicitly (some filters demote).

For proactive bot emails (rare):

Clear subject like "Order #12345 shipped".
Avoid clickbait that triggers spam.

Threading

The bot replies to the original message:

In-Reply-To and References headers set correctly.
Threaded reply that joins the customer's existing thread.
Single thread in customer's inbox.

Useful for keeping conversation coherent.

When the bot shouldn't auto-reply

Five cases:

Customer messages from a clearly-personal address about a sensitive issue.
Subject indicates urgency ("URGENT", "SOS", "DOWN").
Customer marked VIP in CRM. Escalate to a human immediately.
Auto-responder loop detected (their out-of-office plus your auto-reply ping-pong).
Subject contains "harassment", "abuse", "legal" etc.

Configure exclusions under automation rules.

Draft-only mode

For sensitive workflows:

Rolling out Review-before-send drafts are rolling out. Today every reply is sent automatically, so for sensitive inboxes:

Tell the agent what to hand off in its system prompt, for example "For refunds or legal questions, say a person will reply."
Add escalation keywords in the Email channel's settings so matching emails go to your team instead of an AI reply.

Measuring success

Track:

First-response time. One inbox check, plus the time to write the answer.
Resolution rate. % bot-resolved vs human-needed.
CSAT. AskVault doesn't send surveys. If you run your own, record ratings with the CSAT API (see resolve a conversation).
Spam rate. Replies landing in spam (monitor sender reputation).

Visible under Analytics > Channel: Email.

Common pitfalls

Bot too verbose. Trim response-length setting.

Generic auto-reply feel. Customize the system prompt for tone.

Customers don't know it's a bot. Add the disclosure footer.

Replies break threading. OAuth handles correctly; IMAP can drift. Use OAuth when possible.

FAQ

Should I tell customers they're talking to a bot?

Yes. Trust improves; backlash drops. Disclosure is best practice.

Can I have a different tone per topic?

Not per topic. The Email channel has one tone setting (formal, friendly or casual) and an optional system prompt for email. Use the prompt to describe how to handle different topics.

Do regulatory rules require disclosure?

Some (California's bot disclosure law, the EU AI Act). Confirm with legal. AskVault does not add a disclosure by default; add one to your signature.

Was this page helpful?