Canada AI ad disclosure: #MadeWithAI and AI influencer testimonials
Ad Standards' 2025 influencer guidelines say AI-made or significantly altered content should be disclosed. What that means for an avatar ad and its testimonial.
In Canada, the industry body Ad Standards says content generated or significantly altered using AI should be disclosed, with #MadeWithAI or #AIcreated as its examples. That comes from its Influencer Marketing Disclosure Guidelines, released in October 2025. It is self-regulatory guidance rather than a statute, and it speaks to influencer marketing, so an AI avatar presenting your product falls under its AI-influencer lines.
Ad Standards' influencer marketing page lists the guidelines, and I could read only that page, not the PDF text. The wording below comes from Bennett Jones' summary, read on 2026-10-03, so check it against the guidelines before relying on it. Nothing here is legal advice.
What do the guidelines say about AI?
As summarized by Bennett Jones, there are four points a video ad team needs:
- Content generated or significantly altered using AI should be disclosed, for example #MadeWithAI or #AIcreated.
- An AI-generated influencer is not a real person, but that does not remove the duty to disclose a material connection with the brand.
- An AI influencer cannot give a testimonial about a good or service it cannot experience itself; the example given is taste.
- Best-practice labels for a virtual presenter include #virtualinfluencer, #AIinfluencer or #AIcreated.
Where does each disclosure go on an avatar ad?
Hashtags live in the post, not the video file. Sume generates and edits the file; the caption field, the platform's AI toggle and the paid-partnership tag belong to the publishing platform. Sume's avatar video route makes a script-driven talking video from a ready avatar, 4 to 60 seconds long, so the script is where a testimonial problem starts.
| Guidance point | Where it lives | What you can do with Sume |
|---|---|---|
| Disclose AI-generated or significantly altered content | Post caption or platform field, and optionally on screen | Burn a line with video captions cues; put the hashtag in the post yourself |
| Disclose the material connection | Platform paid-partnership tool and the post copy | Nothing in the file; set it at publish |
| No testimonial an AI cannot experience | The avatar's script | Review the script before generating; read the finished transcript with video inspect |
| Label a virtual presenter as virtual | Post copy and, optionally, on screen | Same caption cue route; keep the wording in your brief |
How do I check the script after the render?
A script edit is cheap before a render and a re-render is not, so read the script first. After the render, POST /v1/video-inspect with transcribe: true returns the words the avatar actually says, at $0.01 per audio minute, so you can search it for sensory claims such as taste, smell or feel. Add segmentation: { "mode": "sentence" } and it also returns sentence-level segments[]. If your ad also carries a spoken disclosure, check that it reached the final mix.
How does this compare with other markets?
Rules differ by country and by what they cover. India's advertising council asks three questions on when an AI ad needs a label, and the UK takes a misleading-or-not approach in its own AI-in-ads guidance. For the platform side of the label, see how to label a TikTok ad as AI-generated.
Sources
Related posts
More in Use cases
- ChatGPT Try On won't promise your size: what sellers publish
ChatGPT Try On is reported to carry no size or fit guarantee. What apparel sellers should still publish, and where Sume's fitting preview helps and stops.
- Course trailer from three lesson clips: one audio track in Timeline
Cut a 30-second course trailer from three lesson recordings: detach audio once, join slices with audio.parts and source_in, then lay the video slots over it.
- Turn a creator's raw clip into a brand-avatar ad with face swap
A workflow for re-casting your own or a licensed creator clip onto a brand avatar: trim to 4-15 s, swap on Sume, caption. Python sample and costs by tier.
- Demand Gen A/B experiment: 50 conversions per arm, then what?
Google asks for at least 50 conversions per arm in Demand Gen conversion experiments. What that means for how many AI video variants you can afford to test.
Written by Sume