When to Disclose AI Assistance in Support Replies

Disclose AI use when customers interact directly with AI, when they ask, or when a relevant rule or customer agreement requires it. For replies that you personally check and send, a clear explanation of your drafting workflow is a sensible default; a disclosure in every message may add little.

That is a recommended support policy, not a universal legal rule. The key questions are what the AI does, what you actually review, and what the customer has been led to expect.

Distinguish AI drafting from AI answering

For a solo developer, these workflows deserve different treatment:

| Workflow | Recommended disclosure approach | |---|---| | AI fixes grammar in a reply you wrote | Explain editing assistance in your general support policy if relevant; avoid a repetitive message label. | | AI drafts a reply that you verify and send | Describe the workflow in an accessible support notice. Add a message-level explanation when it matters to the conversation. | | AI sends answers without your review | Clearly identify the answer as automated and AI-generated. | | A customer chats directly with an AI assistant | Identify the assistant at the start and explain the available route to human support. | | AI helps determine an outcome, such as whether to reject a request | Explain its role where material, provide meaningful human review, and assess any separate legal obligations. |

The amount of text you edit is not a useful disclosure threshold. An unchanged draft can receive careful review. A heavily rewritten draft can still contain an unchecked claim.

SupportMe’s supplied pre-launch product description places it in the drafting category: it prepares replies, and nothing sends without explicit human approval. A disclosure for that workflow should accurately describe both the assistance and the human review.

When to disclose AI in the conversation

When an answer goes out without human review

As a practical policy, label an automated AI answer where the customer will see it before relying on the content. Avoid sending it under a founder’s name in a way that suggests the founder personally investigated the issue.

For a chatbot, place the disclosure at the beginning of the interaction. For an automated email, put it near the opening rather than hiding it in a long footer.

Hypothetical wording:

This is an automated reply from our AI support assistant. A person has not reviewed it.

If human help is available, explain the real process. Only say “reply to reach a person” if someone actually monitors those replies.

When the customer asks

Answer directly and describe the work you performed.

Hypothetical wording, appropriate only if accurate:

Yes, I used AI to help draft this reply. I checked the instructions against our documentation and reviewed your account details before sending it.

Avoid answering only “I’m a real person.” That leaves the customer’s question about AI assistance unresolved.

When your existing promises would make AI use surprising

Review how you describe support on your website and in customer agreements.

“Support from the founder” can describe a workflow in which the founder reviews and sends replies. “Every reply is written entirely by me” makes a different promise.

If you introduce AI drafting, update statements that no longer reflect the process. For B2B customers, check relevant contractual commitments before introducing the tool; a footer does not resolve a conflicting agreement.

When AI influences the outcome

Distinguish help with wording from help making a decision.

Hypothetical scenario: An AI tool drafts the explanation for a refund decision you made after checking the policy. That differs from a system deciding that the customer does not qualify.

For consequential cases, such as account restrictions or disputed charges, the recommended approach is to explain the actual basis for the outcome and provide a way to challenge it. “AI-assisted” alone tells the customer very little.

What the EU transparency rules say

The European Commission states that the AI Act’s Article 50 transparency obligations apply from 2 August 2026. For covered systems that interact directly with people, providers must ensure people know they are interacting with AI, unless that is obvious. Notice must be clear and provided from the start of the first interaction. European Commission: Article 50 FAQs

The Commission distinguishes direct AI interaction from communication through a human intermediary. It says systems operating solely in the background fall outside that particular interaction-disclosure obligation. This supports treating a human-reviewed drafting workflow differently from a customer-facing chatbot; it does not establish an exemption from every other transparency requirement. European Commission: direct-interaction guidance

For a small SaaS team, the practical lesson is to assess the actual workflow and applicable jurisdiction. Neither “we use AI” nor “a human clicks send” settles every compliance question.

Keep privacy information separate from reply labels

An AI disclosure explains how a reply was produced. A privacy notice explains how customer information is processed. One cannot replace the other.

The UK Information Commissioner’s Office says organisations using personal data in AI must explain their processing, including purposes, retention periods, and sharing. When collecting information directly from people, relevant privacy information must be provided at collection. The ICO currently flags this guidance as under review following legislative changes. ICO: Transparency in AI

Before describing your workflow publicly, verify what your tools actually do with support messages. Distinguish generating a draft from retaining conversations or using them to improve a model. Avoid unsupported promises such as “your data never leaves our systems.”

Use a short, accurate support notice

For a team that reviews every reply, an accessible support-page explanation can describe the routine workflow.

Hypothetical policy wording:

We use AI tools to help draft support replies. A member of our team checks each reply before sending it and is responsible for the answer.

A message-level version can be shorter:

Drafted with AI assistance; reviewed and sent by me.

Use either statement only when it matches your process. As an internal review standard, check:

• Facts: Are the instructions, links, and product details correct?
• Customer context: Does the answer address this person’s actual issue?
• Actions: Did you really check the logs, issue the refund, or reproduce the bug?
• Promises: Can you deliver any timeline or commitment in the reply?

Disclosure cannot make an inaccurate statement acceptable. If a draft says “I tested this,” remove it unless you actually did.

References

• European Commission — Transparency obligations under Article 50
• Information Commissioner’s Office — How do we ensure transparency in AI?

Conclusion

Match the disclosure to the work: identify direct AI interaction, explain human-reviewed drafting honestly, and answer questions without evasion. Customers should understand who checked the reply, who stands behind it, and how to get further help.