ChatGPT Ads compete for the next question, not the first one

ChatGPT has already answered the shopper's first question by the time an ad appears under the reply. So what is the ad for?

Take foundation. An ad that could appear in any conversation about makeup is a broad bet. A shopper who has just told ChatGPT she has oily, sensitive skin, needs ten hours of wear and wants a natural semi-matte finish has said far more about what she needs. But matching that context isn't enough on its own. If the ad that wins that conversation still says "Bestselling foundation, 20% off", the precision is wasted. At that moment the shopper is more likely wondering why a foundation that matched in the morning looks two shades darker by afternoon, or how to control oil without drying out. The ad's real job is to answer her next question.

How ChatGPT picks an ad

OpenAI's own help pages describe the system:

• The system "selects and delivers ads based on expected relevance and outcomes." The signals it weighs include "the context and intent of the current conversation, the ad's landing page, title, copy, advertiser-provided context hints, and, when ads personalization is enabled, select signals from a user's broader ChatGPT experience."
• At the ad group level, advertisers can provide context hints that describe "the conversations, topics, or keywords where their products or services may be relevant." These hints "help guide ad matching, but they are not exact-match keywords and do not guarantee delivery in specific conversations."

(Source: OpenAI, Ads in ChatGPT: the basics.)

Read together, that is a different job from search ads. Hints are not exact-match keywords, and the ad's own copy and landing page are among the signals. The unit to plan around is a type of conversation.

What that means for a D2C ad group

1. Build ad groups around a situation, not a category. "Foundation for oily, sensitive skin that lasts a workday" is an ad group. "Foundation" is not.
2. Write the hint the way the situation sounds. Plain language, the way the conversation would describe it.
3. Write the copy for the next question. Answer what the shopper will ask after ChatGPT's reply ("why does it oxidise?", "will it dry me out?"), then send her to the page that answers it.
4. Test answers, not just discounts. In the same conversation context, two ads that answer the next question differently are a more useful test than two discount levels.

How Citable plans this

Citable is the campaign management platform for D2C brands winning AI-assisted shoppers. For ChatGPT Ads, that means:

• Plan around the buying questions we track in ChatGPT. Each brand has a tracked set of questions, grouped by persona and buying stage. We see which ones ChatGPT answers without naming the brand.
• Measure. Paid impressions, clicks and spend sit next to the brand's AI visibility.

The merchant creates and runs the ads in their own ChatGPT Ads account. Citable plans and measures; it doesn't buy or change ads.

New to the channel? Start with our guide to ChatGPT ads in 2026.