Using Composer responsibly (AI disclosure)

The Federal Trade Commission (FTC) treats AI-generated marketing content under its rules against deceptive advertising. If AI created or substantially modified your advertising content (copy, images, testimonials, or endorsements) in a way that could mislead recipients into thinking it is human-created or depicts real people or real experiences, you may need to clearly disclose that AI was involved. This applies even when the underlying information — such as actual customer feedback — is real, but the wording or presentation was generated by AI.

A few practical takeaways:

• AI-generated testimonials and endorsements carry the highest risk. Do not present AI-written content as a real customer review or a personal endorsement of an experience that didn't happen. This includes AI-generated personas that act like influencers — these must be clearly identified as non-human, not presented as real people with real opinions.
• The safest approach is to disclose clearly and conspicuously whenever AI meaningfully shaped what the recipient sees, rather than trying to judge case by case.
• If your content is both sponsored and AI-generated, disclose both. These are separate obligations — a sponsorship disclosure alone is not enough if AI also created or substantially shaped the content.
• Disclosures should be easy to notice at normal viewing size, not buried in fine print.

As a practical rule, disclosure may be needed when omitting the AI involvement would be likely to mislead a reasonable consumer, especially for synthetic people, endorsements, testimonials, product demonstrations, or content that implies real experiences.

New York requires a conspicuous disclosure in certain advertisements that include a “synthetic performer.” A synthetic performer is a human likeness created by AI that is not based on a real, identifiable person, for example an AI-generated model in a product email or promotional campaign. If your campaigns include AI-generated human figures and may reach recipients in New York, plan to include a clear disclosure.

There is also a broader New York rule that requires disclosure whenever AI substantially generated your advertising content, not just when it includes AI-generated people. Both requirements are now in effect.

The EU AI Act introduced transparency requirements that apply to marketing content sent to EU audiences. In practice, this means two things:

• Your recipients must be told when content was artificially created or manipulated by AI. A clear, visible disclosure in the message itself satisfies this.
• AI-generated images, video, and audio must also carry embedded technical identification — markers built into the file itself that identify it as AI-generated.

Klaviyo handles this technical marking for images created through Composer. If you bring in AI-generated assets from other tools, check that the identification data is still intact — avoid running those files through additional processing or conversion steps before importing them, as that can strip the embedded markers out.

Note: Across all markets, the safest default is to disclose clearly and consistently whenever AI meaningfully shaped what your recipients see. That approach holds up well regardless of which jurisdiction's rules apply.

Beyond disclosure, a few rules apply regardless of where you are sending.

Don't create fabricated testimonials

Do not send AI-generated content that reads as a customer review, testimonial, or personal endorsement of an experience that didn't actually happen. A common example to avoid is a synthetic, AI-generated person appearing to vouch for a product ("I love these track pants").

This also means: do not present AI-generated content as a real customer review even when the underlying sentiment is real. If AI wrote the words, say so. This is the highest-risk area across all major markets, so keep testimonials and endorsements grounded in real customers and real experiences.

Likeness, copyright, and IP

Using AI to create content does not exempt you from copyright, trademark, or right-of-publicity law.

• Do not use AI to depict real, identifiable people without their permission. For example, you cannot put a recognizable public figure into a promotional image.
• Do not use AI to generate content that copies protected or branded material you don't have rights to use.
• Be aware that AI-generated content may not be protected by copyright the way traditionally created content is. If owning your campaign assets matters to your business — for example, if you want to reuse or license them — talk to your legal counsel about what level of human creative input is needed to establish ownership.

When in doubt, treat AI-generated images and copy the same way you would treat any other asset: make sure you have the rights to use what's in them.

Keep AI provenance metadata intact

In a growing number of markets, preserving embedded AI identification information in your images and other media is not just good practice — it's becoming a legal requirement. When you use Klaviyo's AI-generated images, we handle the technical side of this. Some AI-generated images carry hidden, machine-readable information that identifies them as AI-generated. When you use AI images in your messages, avoid running them through tools that re-process or re-encode the image in a way that strips out this information before you send.

In short: don't remove the AI provenance data that comes with an AI-generated image.