How Your Campaign Is Built

A ChatGPT Ads campaign has fewer moving parts than a search campaign, so each one matters more. This page explains what we build from your brief and why.

The structure

A campaign holds one or more ad groups, and each ad group holds the conversations it should appear in and the ads to show there. Your first campaign usually has one ad group per service you asked us to promote, so each ad can speak to exactly that need.

The conversations: context hints

Instead of keywords, each ad group carries a list of context hints: plain descriptions of the conversations where your ad is useful. A plumber's ad group for emergency call-outs might carry hints like "someone whose water heater has stopped working and needs a plumber today" or "asking how urgent a burst pipe is and who to call". ChatGPT matches conversations to those hints, so the quality of the list decides who sees the ad.

We write the hints from your brief and your website, and from the questions people actually ask AI assistants about businesses like yours. Each hint describes a real situation: the need, the moment, the comparison being made. Never a bare keyword, never just a location (locations are targeted separately), never your brand name on its own.

The ad card

The ad itself is a card inside the conversation:

• Title: a few words that say what you offer.
• Text: up to a sentence that says why you, written for the result you told us matters.
• Image: one square image, at least 640 by 640 pixels. We use your logo or a strong image from your website, shown whole on a matching background, never cut across.
• Destination: the page the click goes to, from your preferred landing page in the brief or the best page on your site for that service.

Every ad is reviewed by OpenAI before it can show. We submit it, follow the review and, if OpenAI asks for a change, rewrite the card and send it back the same day.

Your own images (optional)

If you would rather we used your own photos or logo, add them on the Brand Assets tab of your project. Each image is checked when you upload it: its size and shape, whether it is a photo, a logo or mostly text, and whether anything in it would be refused by ad policies. You see the result next to the file, in plain words ("Good for: chat card", "Too small for a logo"). Images that pass are used in your next campaign draft; if your campaign already exists and is still paused for your review, the card is updated straight away, and if it is live the change is offered to our team first so a running campaign is not disturbed. Nothing here is required: without uploads we work from your website.

Budget and bidding

Your monthly budget from the brief becomes a daily budget on the campaign (or a lifetime budget for a fixed-length push). Bidding starts on clicks, so the campaign gathers data quickly, and moves to conversion bidding once the measurement pixel on your site is reporting results. See Conversion Tracking.

Where the ads show

Location targeting comes from the "where should the ads show" line of your brief. Beyond that, placement is decided by the conversation: the ad appears where ChatGPT shows ads and the conversation matches your hints. Your dashboard's Analytics view shows delivery by ChatGPT surface, country and device.

A few rules we always follow

• Only claims your site or brief supports. No invented prices, guarantees or awards.
• Your name and your voice. The ads never mention us or any agency.
• Built paused. Nothing runs until you have reviewed it and said go.

After launch

Every week the hints that bring the wrong conversations are removed and new ones added from what is converting, the card is tested with fresh titles, text and images, and spend is paced against your budget. Each change is proposed and reviewed by a specialist before it is applied, and listed in the Activity tab of your project. Want something specific changed? Ask in your project chat.