Creative for ChatGPT Ads is one of those areas where a thousand small decisions stack up into the difference between a campaign that returns 6x and one that returns 2x. This piece compiles the creative best practices we have observed across the ChatGPT Ads beta cohort through mid-2026.
Image composition
The placement card renders images at small sizes (200 to 320 pixels) inside a busy text response. Images that work at this size are different from images that work as Instagram hero shots or product page primaries. The best-performing image attributes:
- Square aspect ratio (1:1). Non-square images get cropped to square, often badly. Provide square images directly.
- Neutral background. White, light gray, or muted color. Heavy lifestyle backgrounds compete visually with the surrounding response text.
- Product occupies 75 to 85 percent of frame. Smaller product-in-context shots disappear at placement size.
- Single dominant subject. Multi-item bundle shots, complicated scenes, or overlay graphics all underperform.
- High contrast against background. Light products on light backgrounds disappear into the page. Push contrast higher than you would for other channels.
The image style that performs best in ChatGPT Ads looks closer to a clean Amazon listing photo than to an Instagram brand shot. Brands struggling with this often need to reshoot top SKUs specifically for placement use.
Description voice and structure
Covered in depth in our piece on writing ChatGPT Ads copy. The summary: conversational tone, use-case opening, specific differentiator, buyer pattern, soft closer. 40 to 80 words total. No superlatives, no discount openings, no bullet points.
Placement-format fit
ChatGPT Ads has three placement formats (inline cards, comparison panels, annotated recommendations) and your creative needs to work across all three. Some considerations:
For inline cards: the merchant description needs to stand alone. Users see only the card content, not the surrounding response context (other than the question they asked). Make sure the first 12 to 15 words communicate use case and fit clearly.
For comparison panels: structured attributes matter. If you sell a standing desk and the comparison panel lists "height range," "weight capacity," "warranty," and "price," ensure your feed has accurate data for each. Missing attributes show as blank in the comparison panel, which kills conversion.
For annotated recommendations: the product name itself becomes the clickable element. Make sure your product titles read naturally in a sentence ("the Apex Pro keyboard") rather than as SKU strings ("Apex Pro Mechanical Keyboard, RGB, Brown Switches, US Layout").
The testing cadence that works
The ChatGPT Ads dashboard supports A/B testing on merchant descriptions and (less commonly) image variants. Our testing cadence across client accounts:
- Always test in pairs, never solo: always have two creative variants running simultaneously on top SKUs.
- 14-day minimum test window: shorter windows produce noisy data that often reverses on re-run.
- Test one variable at a time: if you change both description and image at the same time, you cannot tell which change drove the difference.
- Move test to production once winner is clear: a winner with 95 percent statistical confidence after 200+ engagements per variant should move to production.
- Re-test top performers quarterly: what worked in Q1 may not work in Q3 as the user base and competitive landscape shifts.
Creative refresh frequency
The right refresh cadence for ChatGPT Ads is meaningfully slower than for Meta Ads. Meta typically rewards weekly creative refreshes. ChatGPT Ads rewards 60 to 90 day refresh cycles, with daily tweaks usually destroying value rather than adding it.
Reason: ChatGPT Ads creative performance is driven by message-market fit more than novelty. A well-fit description keeps converting at a stable rate for months. Constant tweaking adds noise without signal.
Category-specific creative observations
Some patterns we have noticed across categories:
Apparel and fashion: styled product shots outperform flat lays. Photos showing the product on a neutral mannequin or being held in hand outperform pure product cutouts.
Home goods: in-context placement shots ("desk in a home office") perform worse at card size. Pure product shots win.
Beauty: the product image matters less than other categories. Description quality drives the result.
Electronics: images that show scale (product next to a hand or laptop) outperform abstract product shots. Buyers want to know the size.
Food and beverage: packaging shots dramatically outperform prepared food shots. Buyers want to recognize the product on a shelf.
Our team at ScaleWise VA produces and tests ChatGPT Ads creative for Shopify clients across all major ecommerce categories. If you want creative that consistently hits category-leading performance, book a free 30-minute discovery call.