Write the Ad for the Situation, Not the Product

VistaPrint, one of the pilot advertisers in OpenAI's September 2026 stat pack, shifted its ChatGPT ads "from product-oriented messaging to use-case-driven creative" and reports that click-through rates rose. Its Senior Manager of Paid AI Media, Garima Singh, adds that "the majority of traffic ChatGPT drove was new visitors, consistent with its discovery-oriented nature". We have been building campaigns on that rule since the first one, because the platform's own mechanics demand it. This is why it works and how to apply it.

Why the situation beats the product

A ChatGPT ad renders in one place: below the answer, after the person has read what the assistant said about their question. The reader's state of mind at that moment is the question they asked. Copy that continues the answer ("still comparing physio and osteopathy for back pain? here is how to choose") reads as the next useful line. Copy that announces a product ("Award-winning back pain clinic, book today") reads as an interruption, and the platform notices, because click-through rate feeds the relevance score that decides how often and how cheaply the ad serves.

The three parts the platform scores together are the context hint, the creative and the landing page, and all three should describe the same situation. The hint tells the auction which conversations the ad group belongs in. The title and description answer the situation the hint describes. The landing page opens on it. When those three agree, relevance is high, delivery is cheaper and clicks convert; when the hint says "comparing options" and the creative says "buy now", the auction is being asked to place a mismatch.

What the platform's own data says

VistaPrint's report is one data point, and it lines up with the rest of OpenAI's pack. Every one of the six anonymous Conversion Optimization cases shows click-through rate rising alongside cost per result falling, by up to 180% in the jobs case, without a creative change. That is the bidder finding conversations where the situation fits. Use-case creative does by hand what conversion bidding does with data: it puts the ad in front of the reader whose situation it names. On a client account we run, cost per click fell about 29% over the first weeks as the platform learned which conversations fit the situations we had written, with click-through rate holding between 0.8 and 0.9%.

A worked example

Take a fictional accounting software for freelancers, and one situation: a freelancer in Portugal who has just been told they need to issue invoices with VAT and has no idea where to start. The character limits we measure on live ads are about 50 for the title and 100 for the description, and the first 40 characters of the title are what a phone shows before truncation.

The product-led version is not wrong; it is what the same brand would write for Google, where the reader has typed a product category. Under a ChatGPT answer about VAT obligations for freelancers, only the second version continues the conversation.

How to write a hundred of them

The launch bar on this platform is around a hundred reviewed ads across distinct situations, and use-case creative scales better than product creative because situations are plentiful and product features are few. The method we use: list the claims the brand can already support from its site; map the situations its buyers are in, in their own words; write one ad group per situation with a hint, three to five creatives and a landing page that opens on it; review every claim against the site before upload. The ad generator drafts the first batch against the real character limits, the context hints guide covers the hint format, and the creative specs guide has the measured limits.