ChatGPT Ads Opens Multi-Image Ads and Widens the Desktop Card

Since 26 October 2026 every ChatGPT Ads advertiser can build a standard ad that carries four to six square images, in open beta. Since 30 October the standard card uses more of the conversation width on desktop. The images share one title, description and landing URL, Learn more is the only button, and an existing single-image ad cannot be converted. The country list did not move: 89 countries, the same as on 21 October.

What OpenAI shipped

OpenAI's product email of 30 October announced both changes. "Multi-image ads are now available in open beta to all advertisers," it said: you select Multiple images under Ad format when creating an ad in Ads Manager, then upload 4 to 6 square (1:1) images. On the card itself: "We're rolling out a new, wider standard ad card that uses more of the conversation width on desktop. This ad format uses existing campaigns and creative, so no advertiser action is needed."

OpenAI's help article on creating ads dates the multi-image beta to Monday 26 October and sets the rules:

• Both formats work for standard campaigns with the Reach, Clicks or Conversions objective, with the same headline and description requirements.

• Each ad has one destination URL. Every image is clickable and links to it, and Learn more is currently the only call to action.

• The images are square, 1:1, four to six of them.

Its FAQ for ChatGPT users now describes one ad unit below a response "that may contain one or multiple images".

What the API does

The developer changelog of 26 October lists the same format: an ad takes a list of four to six image file IDs in creative.file_ids and returns them, in order, in creative.images. The developer docs add two conditions the email leaves out. The account must be enabled for ads with multiple images, and the ad group must use fixed bidding unless the account is also enabled for multiple images with automatic bidding. A multi-image ad also takes no image crop, price or business agent.

We tested it on our own account on 2 November. A paused ad group on fixed bidding accepted a four-image ad, and the API returned the four images in order with a call-to-action label of Learn more, a field the documentation does not list. The ad went straight into review, and we archived it before it could serve.

To move an existing ad to multiple images you build a new one: OpenAI's documentation says an edit cannot turn a single-image ad into a multi-image ad. Any creative change also sends an ad back to review.

What changes for your copy

The title and description limits are the same, and one title and one description now have to carry four to six pictures. Plan the set as one message: the same product from four angles, four products from one range, or four steps of one service. Four images that each tell a different story under one title read as a mismatch.

OpenAI describes the wider card as a desktop change. The phone, where the card is narrowest, is still where a long title gets cut, so keep writing to the phone. The creative specs guide has the full limits.

What OpenAI has not said

• Whether creative enhancements apply to multi-image ads. Enhancements switched on by default for self-serve campaigns on 30 October and can rewrite titles, translate copy, resize images and change the landing URL. OpenAI's pages on the two features do not say how they combine. An opt-out you saved still holds.

• How results break down by image. OpenAI has not described per-image reporting.

Shopify, now outside the US

The same email says the Shopify app now connects stores in every country where ChatGPT Ads are available, syncing the product catalogue and commerce events into Ads Manager. OpenAI's Shopify help article, last updated in mid-October, still describes US availability with other markets to follow "shortly". Check the app in your own store before you plan around it.

The image-generation ads

The visual ad OpenAI announced on 5 October for image generation is a separate placement. Its help centre now describes that test as running for US users: an ad with several images from one advertiser shows while ChatGPT makes an image and disappears when the image is done. There is no self-serve access to it. Our post of 6 October covers it.