AI UGC for Advertising Platforms: What to Make for Each Channel, and the Rules That Apply
AI UGC works on most advertising platforms, but not in the same form. On Meta and TikTok it is a person with your product in a vertical frame; on Google Performance Max it is a set of clean landscape, square and portrait images with no burned-in text; on Reddit and X it is a believable photo under a post that has to stand up to public replies. This guide maps what to make for each channel and the rule that matters most there. The rule that applies everywhere: the product and the claims are real, and an AI person is never presented as a customer.
4 min read

Most guides to multi-platform creative say “make it native.” That is true but not useful. What actually differs between platforms is the frame you generate, the text the image has to sit beside, how the platform treats AI-made media, and who gets to comment. The table below is the short version; the sections after it cover the channels where the details matter.
The quick reference
| Platform | What to generate | Rule that matters | Go deeper |
|---|---|---|---|
| Meta (Facebook, Instagram) | 4:5 feed, 9:16 Stories/Reels, 1:1 carousels; person with product | Meta may show an "AI info" label on ads it detects were made with generative AI | Facebook ads |
| TikTok | 9:16 video and photo carousels, selfie-angle | Realistic AI-generated content must be labeled; Spark Ads boost a real post | TikTok Shop |
| Google Performance Max | 1.91:1, 1:1 and 4:5 images without text overlays | Up to 20 images per asset group; Google writes the text | Demand Gen guide |
| YouTube | Shorts at 9:16; thumbnails at 16:9 | Creators must disclose realistic altered or synthetic content on uploads | YouTube |
| 4:3 product scenes that look like a community post | Comments are optional per ad and public when on | Reddit ads | |
| X | 1:1 or 1.91:1 images that prove the post's claim | Replies and Community Notes can sit on ads | X ads |
| 2:3 vertical styled scenes | Pins live for months; make them evergreen | Pinterest Shopping | |
| Snapchat | 9:16 full-screen stills and short video | Ads run full-screen between friends' content; show the product in the first second | Snapchat |
| 1:1 and 1.91:1 work scenes with your real product or screen | Shows a Content Credentials (CR) icon on media signed with C2PA | ||
| Retail media (Walmart, Amazon) | Product-forward lifestyle images matched to the listing | Ad must match the product page exactly | Walmart Connect |
Sources for the rules above: Meta's help page on AI labels in ads, YouTube's announcement of altered or synthetic content disclosure, LinkedIn's help page on Content Credentials, Google's asset group specs, and Reddit's ad specs. For Meta's Advantage+ campaigns specifically, see the Advantage+ guide and ppl.studio for Meta Ads.
Google Performance Max: clean images in three ratios
Performance Max builds ads from your assets across Search, Shopping, YouTube, Display, Discover and Gmail. Google's asset group table lists:
| Image type | Recommended size | Minimum size |
|---|---|---|
| Landscape 1.91:1 | 1200x628 | 600x314 |
| Square 1:1 | 1200x1200 | 300x300 |
| Portrait 4:5 | 960x1200 | 480x600 |
Files are JPG or PNG up to 5,120 KB, with up to 20 images per asset group. Three things follow for AI generation. Generate each ratio as its own composition instead of cropping one image three ways. Keep images free of burned-in text, because Google assembles headlines and descriptions itself and will place them next to or over your image. And split asset groups by buyer situation (the gym bag, the desk, the kitchen) so each group's lifestyle images match the audience signal you gave it. Keep your clean product-on-white shots in every group for Shopping placements. For Gmail, Discover and YouTube feed placements, the Demand Gen guide covers the overlap; for shopping feeds, see Google Shopping.
Programmatic and display banners
Banner sizes are small and fixed: 300x250, 728x90, 160x600, 320x50 and their relatives. At 300x250 a wide lifestyle scene turns into a smudge. For display, generate tighter compositions where the product fills a third of the frame, leave a clear area for the price or button, and make the tall 160x600 and wide 728x90 as separate compositions, not crops. Responsive display ads take landscape and square images and assemble the rest, so the same no-text rule as Performance Max applies. Native ad placements on publisher sites want editorial-looking images instead; see the native advertising guide.
Spark Ads and creator whitelisting: what AI can and cannot stand in for
TikTok Spark Ads promote an existing TikTok post rather than an uploaded ad. That post can come from your own account, or from a creator who turns on ad authorization and sends you a code; TikTok lets the creator choose 7, 30, 60 or 365 days, and the caption cannot be edited once authorized (see TikTok's Spark Ads guide). Meta's partnership ads work the same way through a creator's account.
What makes whitelisted ads work is the real creator: their handle, their audience, their reputation. An AI persona has none of those, so AI UGC cannot be whitelisted in that sense. What you can do honestly:
- Post AI UGC on your own brand account, labeled as AI-generated where the platform requires it, and Spark the posts that perform organically.
- Use AI scenes for the product shots in a real creator's video, with their agreement, when they cannot shoot every angle.
- Do not run an account that presents an AI person as a real individual who happens to love your product. That is the fake-endorsement problem with extra steps.
Events, trade shows and webinars
Event marketing needs visuals before the event exists. AI imagery helps with the pre-event layer: invitation emails, LinkedIn promotion and booth screens showing your product in the settings customers use it. Three limits keep it honest:
- Speakers are real. A webinar promo should show the person who will actually be speaking, from a real photo. Generate the supporting scenes, not the presenter.
- The booth is real. Do not post generated “crowds at our booth” or “live from the show floor” images. Pre-make product content for the social calendar, and post real photos for anything that claims to be from the event.
- Attendees are not testimonials. AI “attendees” nodding at a slide are fine as illustration; captioned quotes from them are not.
For short spoken teasers, a real speaker can record a few lines; if you use Animate on an AI Expert instead, present it as your brand's host, not as the speaker.
Press and media kits
Journalists use media-kit images as photographs of what you sell, so a press kit is the one place where generated lifestyle images need the most care. Put real product photography first. If you include AI-generated scenes, put them in a clearly named folder (“illustrative lifestyle images, AI-generated”) so an editor does not caption them as real. Founder and team headshots should be real photos of real people. And never send AI renders of an unfinished product as product photos; if the shipped item differs, the coverage is wrong.
Email and SMS
Lifecycle email and MMS are good homes for AI UGC: a new scene of the exact product someone browsed, a seasonal version of a best seller, a person using a product a customer just bought. Generate landscape or square hero images and keep them consistent with the ads people clicked. The SMS and MMS visuals guide covers sizes and timing.
Where we would not lead with AI UGC
- Connected TV. A living-room screen shows every flaw in a generated face and hand. Use real footage for TV spots; AI stills can work as product B-roll.
- Regulated categories. Health claims, financial results, and anything where an image could be read as evidence of an outcome.
- Anything that needs a specific real person. Founders, staff, named customers, public figures.
One production session, many placements
In ppl.studio, upload your product once, choose an AI Expert, and generate the ratios each channel needs from the same scenes. The Product UGC Photos workflow covers stills; TikTok + Reels Short-Form Video and Creator UGC Video cover vertical video. The app is $1.99 per week (pricing), with images billed by Google on your own Gemini API key. Channel-specific posts: TikTok Shop photos, carousel ads, mobile app marketing and organic social media. Platform hubs: Instagram, Pinterest, Snapchat.
Generate every placement from one product upload
Make feed, vertical, square and landscape scenes of your real product with a consistent AI Expert. $1.99 per week, 10 free photos to start.
Start free with ppl.studio10 free photos · no credit card required
Founder of ppl.studio. Building AI tools for product marketing teams who need visual content at scale without the production overhead.