ChatGPT Ads or the Answer Itself: Where to Spend This Holiday Season

OpenAI spent late October emailing businesses with an offer: sign up for ChatGPT Ads Manager before the holidays and get ad credit on your first campaign. For a retailer, an agency, or a software company, the email raises a fair question. If shoppers are asking ChatGPT what to buy, is the quickest way in to pay for it?

Sometimes. But a ChatGPT ad and a ChatGPT answer are two different products, sold by two different systems, reaching two different groups of people. Knowing where one ends and the other begins decides where the money should go.

The ad sits under the answer, never inside it

That separation is the most important fact for anyone deciding between the two. Buying an ad does not move you into the answer. SE Ranking ran 50,006 commercial prompts through ChatGPT in the US on 23 July 2026 and published the results: an ad appeared on 25.94% of them, always as a single sponsored card below the answer. In only 3.63% of those placements did the answer above also cite the advertiser's domain. The exact URL being advertised was cited 0.09% of the time.

So the ad and the answer are two separate slots. You can pay for the one underneath. You cannot pay for the one people read first.

What an ad buys you

Speed and control. A self-serve campaign starts once it is approved. You choose the message, the landing page, and the budget, and you can pay per thousand impressions or per click. A product launching in November has no history for ChatGPT to draw on, and an ad is the only way it appears next to a relevant question this season.

A guaranteed spot on queries you would otherwise miss. If a competitor owns the organic answer for "best running shoes for flat feet", an ad is the one way to sit on the same screen.

Holiday timing. A promotion that runs for ten days has no time to earn its way into answers. Paid placement can start and stop with the sale.

The limits are just as concrete. Ads reach only Free and Go users, so the paying subscribers who use ChatGPT most never see them. Click-through is low: Similarweb measured 0.50% for the ChatGPT ad unit in June 2026, against roughly 6.4% for Google search ads, and notes the two are not directly comparable because 65.3% of ChatGPT sessions carry on after an ad appears. SE Ranking's own test campaigns earned a 1.30% click-through rate across 97,000 impressions and produced, in its words, very few sign-ups. Relevance is uneven too: SE Ranking judged 14.35% of the ads it saw to be no closer to the question than a random pairing.

And the spend stops working the day the budget does.

What the answer buys you

Everyone sees it. The answer reaches every plan, including the paying users ads never touch. It is also what people act on. A recommendation in the body of the reply carries the weight of the reply; a labelled card underneath it carries the weight of an ad.

It compounds. A clean product feed, a page an agent can read, and facts other sites can cite keep working after the campaign ends, and they work in every assistant, not only ChatGPT.

It is increasingly decided by data, not copy. As we covered in ChatGPT Shops From Product Feeds Now, Profound's tracking found the share of ChatGPT shopping picks from merchants with a direct product feed jump from 8.26% to 61.54% in a single day in July. The way into the answer is to be the source an agent can confirm.

The costs are real as well. Organic visibility is slower, you cannot choose the query, and you cannot guarantee the result. OpenAI does not publish how it picks products or sources, and a single unannounced change in July cut visibility by a third or more for 450 of the 687 retailers Profound tracks.

Side by side

What we recommend

Fix the foundation before you buy anything. Both routes run on the same data. OpenAI's product feed specification uses one feed for organic discovery, ads, and checkout, with an is_ads_eligible flag for products you want to advertise. Every ad click lands on your site, and OpenAI checks advertiser landing pages with its own crawler, OAI-AdsBot, which we covered in Ads Arrive in ChatGPT. A feed with stale prices, or a landing page an agent cannot read, wastes ad money and organic chances alike. Run the page you would advertise through the Agent Analyzer first.

Then buy ads for the gaps, not the wins. Good uses: a new product, a time-limited promotion, or a category where a competitor holds the answer. Poor use: paying for a query where you are already the recommendation, because the ad and the answer will sit on the same screen and you pay for the click you would have had anyway. Ask ChatGPT your customers' questions before you bid, and note where you appear.

Make the promotion readable, not just visible. A holiday sale that lives in a banner image is invisible to an agent. Put the sale price and its end date in the feed, mark it up in the page's Offer structured data, and state it in text. An agent quoting last week's price sends a shopper to a page that disagrees with it, and the shopper blames you, not the agent.

If you sell a service, the same split applies. There is no feed for a consultancy or a clinic, so the answer depends on pages: what you do, where, for whom, at what price, and who else says you do it well. Ads can cover the gap while those pages earn their place. The Agent Visualizer shows where an agent reading your service page stops before it finds those facts.

Measure them separately. Ads Manager reports on the ads. Nothing reports on the answers yet, so keep a fixed list of customer questions and check them monthly. If the ads are working and the answers are not, the ads are renting attention the foundation should be earning.

Our view is plain: for most businesses this season, the organic work is the better investment and the ad is the tactical layer on top. An ad buys a place under the answer. Only the data buys a place in it.

What this does not tell you

The ad figures come from third parties with their own samples. SE Ranking used one set of US prompts on one day; Similarweb does not disclose its sample size. Other trackers report very different ad rates, and OpenAI has published no benchmarks. OpenAI's statements about answer independence describe its own systems and have not been independently audited. The ad product is less than a year old and has changed every few months, so any number here may be out of date by the time you read it.