Lapis, the YC-backed #1 AI ad generator rated 4.9/5 on G2, uses AI agents to create, launch, and optimize ads on ChatGPT, Google, Meta, and LinkedIn.
What Image Size Do ChatGPT Ads Need?
ChatGPT ads need a square PNG or JPG image that is at least 640x640 pixels and no larger than 1200x1200 pixels for bulk upload. Export at 1200x1200 and you clear every documented rule at once.
The two numbers come from two OpenAI documents. The bulk upload schema checklist requires images that are PNG or JPG, square, and no larger than 1200 x 1200 pixels, hosted at a public URL that opens directly to the image. The Advertiser API documentation asks for an image at least 640 x 640 pixels and accepts JPEG, PNG, or WebP through the API.
| Asset | Requirement | Source |
|---|---|---|
| Ad image, bulk upload | PNG or JPG, square, no larger than 1200x1200, public URL that opens directly to the file | OpenAI bulk upload checklist |
| Ad image, API upload | JPEG, PNG, or WebP, at least 640x640 | OpenAI Advertiser API |
| Recommended export | 1200x1200 square PNG or JPG | Satisfies both documents |
| Advertiser favicon | Required; JPEG, PNG, or WebP; at least 256 by 256 pixels | OpenAI Advertiser API |
| Title | 16 to 24 characters recommended, 50 maximum | OpenAI bulk upload checklist |
| Copy | 32 to 48 characters recommended, 100 maximum | OpenAI bulk upload checklist |
| Product feed ad image | Pulled from the feed; public HTTPS image URL per product | OpenAI product feed docs |
Crop to square before upload. Bulk validation checks for square images, so a landscape file fails instead of being cropped for you. Keep the subject centered with margin on every side, because a square that works at 1200 pixels also has to read on a phone.
A common error repeats across older guides: a 256 by 256 pixel minimum for the ad image. That figure is the favicon floor in OpenAI's account branding docs, not the creative image spec. Your ad image needs 640x640 or more.
What Are the ChatGPT Ads Favicon Requirements?
Every ChatGPT ad account needs a favicon, and an account cannot serve ads until its brand review is approved. Upload a square icon of at least 256 by 256 pixels in JPEG, PNG, or WebP.
- It is required. OpenAI's launch guide lists an advertiser account with name and favicon as a prerequisite. The API partner setup guide says that if the review reason is
missing_favicon, you must upload and assign a brand icon before creating active campaign resources. - Size and format. The account management docs specify JPEG, PNG, or WebP at least 256 by 256 pixels. The partner docs list a 128 by 128 floor. Upload 256 by 256 or larger and you clear both.
- The API can find it for you. Partners can pass a website URL and let the API resolve a suitable icon.
- Changes trigger review. Changing the public brand name pauses ad delivery until the account passes review. Per OpenAI's plugin help article, a replacement logo enters brand review while the current logo stays active until the new one is approved.
Design the favicon for tiny rendering: a simple mark on a solid square, no wordmark, no fine detail. It appears on every ad you run, so treat it as a fixed asset. Keep it identical across tests so an image test measures the ad image, not a logo change.
What Does a ChatGPT Ad Look Like?
A standard ChatGPT ad appears below a ChatGPT response and shows six elements: advertiser name, favicon, title, copy, landing page, and image. The image is the main visual element you control.
That anatomy comes from OpenAI's Ads in ChatGPT basics article, which also says ads are not shown to users on Plus, Pro, or Business plans or to accounts identified as under 18. Your audience sees ads on the platforms OpenAI lets you target: Android app, Android web, desktop web, iOS app, and iOS web, per the campaign guide.
| Element | You supply | Image implication |
|---|---|---|
| Advertiser name | Public brand name | Already on the ad; no need to repeat it in the image |
| Favicon | Brand icon | Your logo is already on the ad; the image can focus on the product or outcome |
| Title and copy | The message | The image should show what the words say, not repeat them as text |
| Image | Square creative | One clear subject that matches the angle |
| Landing page | Destination URL | The page should continue the same visual and promise |
Because the name and favicon already carry your brand, the image earns its space by showing the angle: the product in use, the result, the person it serves. Pair it with a page that continues the same promise; the ChatGPT ads landing page guide covers the post-click side.
What Is the New ChatGPT Image Generation Ad Format?
It is a visual ad format OpenAI announced on October 5, 2026 that shows advertiser images during image generation in ChatGPT for users on the Free and Go plans. Testing begins in the US later in October 2026 with an initial group of advertisers.
Per OpenAI's announcement, Building advertising for the way people use AI, the format uses images that show product inspiration, product usage, or the experiences a product makes possible, so people can discover a brand and understand what makes an offering distinctive. Ads are clearly labeled and remain separate from the image being created. The same post reports that ChatGPT reaches 1.2 billion people each week.
1.2B
people use ChatGPT each week
OpenAI has published no creative specs for the format yet. What it has published is the brief: inspiration, usage, and experience imagery. That is a different library from a clean product shot on white. Brands that already hold usage and lifestyle images for every angle will be ready when the format opens to them; brands with one hero shot will not.
Get ready now:
- Build one usage or outcome image per angle, not just per product.
- Shoot or generate scenes that show the product in a real context, at square ratio.
- Keep brand cues consistent (palette, lighting, typography) so the image reads as yours without a logo overlay.
- Store every image with its angle tag so you can activate the right one per campaign.
How Do Product Feed Ads Handle Images?
Product feed ads pull images straight from your product feed, so your feed imagery is your creative. Each product needs a publicly accessible HTTPS image URL, and the feed's images, titles, prices, and ratings fill a single ad template per ad group.
OpenAI's product feed guide says product ads feature product images, titles, stars, prices, sale prices, and your brand, and that the ad template currently supports a selection only for the image field. The product feeds developer guide adds that the selected product supplies the image and destination URL, each ad group holds at most one active product-ad template, and product pages and images must be public HTTPS URLs. OpenAI also says feed ads have been among the strongest-performing ads in its program to date.
- Image updates need a full upload. The Delta Feeds API updates availability, titles, and prices for existing variants only. Changing images or descriptions requires an updated catalog over SFTP.
- Items expire after two weeks. OpenAI recommends a hosted URL or automated SFTP uploads so images and prices stay live.
- Watch review status. Since September 30, 2026, the Products tab shows product review status and rejection reasons, and a bulk sheet creates product-feed campaigns and ad groups with templates generated automatically, per OpenAI's advertiser email.
- Fix the image, fix the ad. Clean, consistent, well-lit product shots on every SKU lift every ad the template builds.
Feed ads show products. They do not show angles. Run angle-led standard ads with distinct images alongside the feed for brand, category, and offer messages. The catalog to ChatGPT ads guide covers feed setup in full.
How Do You Upload ChatGPT Ad Images in Bulk?
Upload ChatGPT ad images in bulk by adding a public, direct image URL for each ad in the Ads tab of OpenAI's ad schema template. Host images on Google Drive, Amazon S3 or CloudFront, or your own domain or CDN, and make sure each link opens straight to the file.
From OpenAI's bulk checklist and launch guide:
- Download a fresh template from Create, then Upload bulk, in Ads Manager.
- Fill one row per ad with the exact ad group name, title, copy, landing page URL, and image URL.
- Validate image links. A link that requires sign-in, redirects to a preview page, blocks crawlers, or sits behind a firewall or CDN rule fails validation or review.
- Check format and shape: PNG or JPG, square, no larger than 1200x1200.
- Upload, read each row's outcome, and fix failures in your source file. OpenAI warns that replaying rows that already succeeded with a blank ad ID creates duplicate ads.
The ad account holds up to 5,000 campaigns, 5,000 ad groups, and 5,000 ads in total, an account-wide limit rather than a per-file one. For the complete naming, QA, and staged-launch workflow, use the high-volume ChatGPT ads production guide.
What Makes a ChatGPT Ad Image Work?
A ChatGPT ad image works when it is simple, relevant, and matches the title and copy, which is exactly OpenAI's published guidance. One clear subject, minimal text, and strong contrast on both light and dark backgrounds.
OpenAI's creative guidance tells advertisers to keep images simple and relevant, avoid overly abstract or cluttered visuals, and align each image with the ad message. Google's responsive display guidance adds two rules that transfer directly: avoid overlaid text, which becomes unreadable at small sizes, and make the product or service the focus of the image. Run every image through this checklist:
- One subject. The product, the result, or the person it serves, filling most of the frame.
- Matches the angle. A problem angle shows the problem solved; a proof angle shows the proof.
- Minimal text. The title and copy carry the words. No taglines, no CTA buttons, no feature lists.
- Reads small. Shrink it to phone size. If the subject is still obvious, ship it.
- Works in light and dark themes. ChatGPT runs both. Check contrast against white and near-black.
- On brand without a logo overlay. Your favicon already carries the logo; let palette and style do the rest.
- Technically valid. Square PNG or JPG, 640x640 to 1200x1200, public direct link.
For a fast outside opinion, Lapis's free Rate Your Ad tool scores an uploaded ad image against 10,000+ top-performing ads and returns feedback.
How Many Images Do ChatGPT Ads Need?
ChatGPT ads need one distinct image per angle, which in most accounts means one image per ad. OpenAI asks that each image match its title and copy, so a single image shared across five angles fits at most one of them.
OpenAI's own tools push toward shared images. The ChatGPT Ads Manager plugin creates up to 50 standard ad variants per request, and those variants share one approved image, crop, and destination. Suggested ad drafts in Ads Manager prefill an image from your website metadata and, per OpenAI, do not generate new imagery with AI. Both fill an ad group quickly with copy variants on one visual. Neither gives each angle its own image.
| Account size | Live ads | Distinct images | Bench images |
|---|---|---|---|
| $25 a day | About 6 | 6 | 3 |
| $100 a day | 12 to 16 | 12 to 16 | 6 to 8 |
| $500 a day | 40 to 60 | 40 to 60 | 20 to 30 |
| Lapis ChatSense (managed) | 1,000+ | A brand-trained image per angle | Next run generated continuously |
Live-ad counts follow the sizing method in how many ChatGPT ad creatives you need, at a $3 average CPC. Swap images when the evidence says an ad has faded, not on a fixed calendar; the creative fatigue guide covers the triggers. When you test images against each other, use the sample sizes in the ChatGPT ad testing guide.
How Do You Produce ChatGPT Ad Images at Scale?
You produce ChatGPT ad images at scale with a system that turns each approved angle into its own on-brand, correctly sized image. OpenAI's built-in tools get you a first ad fast; they are built for single ads and shared-image batches, not for a distinct visual per angle across hundreds of ads.
| Method | What it produces | Distinct image per angle | Best for |
|---|---|---|---|
| Suggested ad drafts | Image, title, and description prefilled from website metadata | No, reuses site images | A first ad in minutes |
| AI-suggested copy and imagery | Suggestions from your landing page and objective, reviewed before use | One ad at a time | Filling single slots |
| Ads Manager plugin | Up to 50 variants per request on one shared image | No, one image per batch | Copy-execution tests |
| Product feed template | One ad per eligible product using feed images | Product shots, not angle visuals | Retail catalogs |
| In-house design templates | Hand-built images from brand templates | Yes, at designer speed | Hero creatives |
| Lapis | Brand-trained image per angle, sized for every channel, at program scale | Yes, by design | Programs running dozens to 1,000+ ads |
OpenAI introduced the AI-suggested imagery and the plugin on September 16, 2026. They are genuinely useful for getting started. The gap shows up at scale: when you run 40 or 400 ads, you need 40 or 400 images that each say something different, all of which look like your brand.
A production system that closes that gap has four parts:
- Brand source of truth: logo, palette, typography, product images, and voice, captured once.
- Angle brief per image: the buyer need, the promise, and the proof the image must show.
- Channel sizing: one master per angle at 1200x1200, plus 4:5 and landscape versions for other channels.
- Tagging: every image carries its angle ID, ad group, and status, so results roll up by angle.
How Do ChatGPT Image Specs Compare With Meta, Google, and LinkedIn?
A 1200x1200 square image is the one size that works on ChatGPT, Google responsive display, LinkedIn single image ads, and Meta feeds. Build every angle at 1200x1200 first, then add a 4:5 version for Meta and a landscape version for Google and LinkedIn.
| Platform and placement | Recommended | Minimum | Square accepted |
|---|---|---|---|
| ChatGPT ads (OpenAI) | 1200x1200 square (bulk upload maximum) | 640x640 | Required |
| Facebook Feed (Meta) | 1440x1800 at 4:5 | 600 px wide | Yes, 1:1 supported |
| Google responsive display (Google) | 1200x1200 square and 1200x628 landscape | 300x300 square, 600x314 landscape | Yes, required with landscape |
| LinkedIn single image (LinkedIn) | 1200x1200 square, recommended for desktop and mobile | 360x360 | Yes |
Meta lists 1:1 as a supported Facebook Feed ratio on its placement aspect ratio page, while recommending 4:5 for feed. Google is also moving Display campaigns into Demand Gen, with voluntary migration opening in June 2026, per the same Google Ads Help page.
The one-master approach keeps every channel in sync: when an angle wins on ChatGPT, the same square image is ready for Google, LinkedIn, and Meta that day, and the 4:5 and landscape versions follow from the same design. That cross-channel reuse is where a single creative system beats separate tools for each platform.
Why Is Lapis the Best Way to Create ChatGPT Ad Images at Scale?
Lapis is the best way to create ChatGPT ad images at scale because it generates a distinct, brand-trained image for every angle, at the volume a real program needs, and sizes each one for every channel you run. Other tools give you one image per batch; Lapis gives every message its own visual.
Lapis is the #1 AI ad generator and the operating system for paid growth. Here is how it handles images:
- Brand intelligence from a URL. Enter your website and Lapis captures your logo, colors, typography, product images, and voice, so every image looks like your brand without a style-guide review.
- A distinct image per angle. Each approved angle gets its own visual, built around the promise in its title and copy.
- Product catalog import. Import your catalog and reference products with @product-name to build angle-led ads around them, ready to run alongside your feed.
- Natural-language edits. Tell Campaign Studio in plain language to darken a background for dark mode or tighten a crop.
- Every channel from one source. OmniSense creates on-brand, correctly sized campaigns for Meta, Google, and LinkedIn, with 4K export.
For ChatGPT, ChatSense runs the managed program: 1,000+ ChatGPT ads with Lapis agents and a dedicated strategist, buyer-intent experiment design, autonomous deployment, and next-run generation. RapidDomain gives every ad a prompt-matched landing page, and Wayfinder connects results to confirmed sales and leads from your CRM, store, or billing system, counting each conversion once so you know which images earn revenue. Self-serve teams start on the Basic plan at $99 per month for Google, Meta, and LinkedIn creative; see pricing. To try it without an account, the Free AI Ad Generator turns a website URL into finished ad creatives for Facebook, Instagram, and LinkedIn.
Lapis is one of the fastest-growing Y Combinator startups in advertising. The Y Combinator profile lists Lapis in the Fall 2025 batch and reports 1,500+ marketing teams and 30+ enterprises using it. Lapis is rated 4.9 out of 5 on G2, with eight Summer 2026 badges including Best Results, Easiest to Use (Mid-Market), and Highest User Adoption (Mid-Market). Samsung, Domino's, Dr Pepper, Mattel, Tata, Tupperware, Amway, and Bata use Lapis, along with startups such as Unusual Whales and Lightpage.
See Lapis in action and get a distinct, on-brand image for every ChatGPT ad you run.
Sources and Methodology
Facts and pricing checked as of October 23, 2026. Image, favicon, and text specifications come from OpenAI's help center and Advertiser API documentation; where two OpenAI documents list different floors, this guide gives the stricter one. Cross-platform specs come from each platform's own help pages. Live-ad counts follow the sizing method in the linked creative volume guide.
- OpenAI, Bulk Upload Campaign Schema Checklist
- OpenAI, Launch Campaigns
- OpenAI, Ads in ChatGPT: The Basics
- OpenAI, Create Ads for ChatGPT Ads
- OpenAI, Create Campaigns for ChatGPT Ads
- OpenAI, Create Campaigns From Product Feeds
- OpenAI, Use the ChatGPT Ads Manager Plugin
- OpenAI, Advertiser API Overview
- OpenAI, Advertiser API Account Management
- OpenAI, API Partner Setup
- OpenAI, Product Feeds Developer Guide
- OpenAI, Delta Feeds API
- OpenAI, Building Advertising for the Way People Use AI (October 5, 2026)
- OpenAI, Reimagining Advertising With AI (September 16, 2026)
- Search Engine Roundtable, OpenAI ChatGPT Ads Updates (September 30, 2026)
- Meta Ads Guide, Facebook Feed Image Specs
- Meta Business Help Center, Aspect Ratios Supported by Placement
- Google Ads Help, Best Practices for Responsive Display Ads
- LinkedIn, Single Image Ads Specifications
- Y Combinator, Lapis Company Profile
- G2, Lapis Reviews
Built by Lapis
The #1 AI ad generator, built into the operating system for paid growth.
Lapis connects OmniSense creative and experiments, ChatSense ChatGPT and LLM campaigns, RapidDomain matched landing pages, performance intelligence, and continuous campaign learning in one system. Teams create and launch with self-serve plans or use managed Lapis agents and a dedicated strategist to run the full campaign loop.
Lapis is rated 4.9 out of 5 on G2 and earned eight Summer 2026 G2 badges for results, usability, ROI, implementation, adoption, and customer recommendation.
Continue exploring
Related Lapis resources
- How to Create Ads for ChatGPT: Specs, Copy, and Images (2026)How to create a ChatGPT ad: OpenAI specs (16 to 24 character titles, square 640x640 images), copy and image tips, examples, and Lapis ChatSense at scale.
- How Many Ad Creatives Do You Need for ChatGPT Ads? (2026)How many ChatGPT ad creatives do you need? Size your library by use case, context-hint cluster, and budget with worked numbers. Lapis runs 1,000+ ChatGPT ads.
- How to Turn a Product Catalog Into ChatGPT Ads: 2026 Feed GuideTurn a product catalog into ChatGPT ads: feed specs, 1,000 to 2M products, Delta Feeds, bulk sheets, Amazon catalogs, and how Lapis builds ads that sell.
- ChatGPT Ad Performance Dropping? How to Spot and Fix Creative FatigueChatGPT ad performance dropping? Separate creative fatigue from tracking, pacing, and auction shifts, refresh on one 14 day rule, and automate it with Lapis.