ChatGPT Ads explained: how the auction, pricing and feed ads actually work

What a ChatGPT ad actually is

A ChatGPT ad is a sponsored placement that appears below a relevant conversation, visually and functionally separated from ChatGPT's own answer. The unit itself is simple: an advertiser name, a favicon, a title, a short line of copy, a landing page link and an image. OpenAI decides which ad to show based on the conversation's context and intent, not a keyword match in the traditional sense.

Only Free and Go tier users see ads. Plus, Pro, Business, Enterprise and Edu accounts stay ad-free, and OpenAI excludes anyone it predicts to be under 18. That matters for targeting: you're reaching people who haven't paid for ChatGPT, which skews the audience differently than a channel like LinkedIn, where the paid, professional tier is exactly who you'd want to reach.

The two ad formats

This is the bit most explainers skip. ChatGPT Ads isn't one format, it's two, and which one applies depends on what you're selling:

• Standard placement ads. No product feed required. Built around "context hints", descriptions of the buyer situations you want to appear in, closer to a creative brief than a keyword list. This is the format for local businesses, lead generation and service brands, a wedding venue, a solicitor, a SaaS product, anything without a catalogue.
• Product feed ads. The Google Shopping equivalent. Your product catalogue is converted into OpenAI's own feed format, which has its own attribute set, several with no Google equivalent, and shown when someone's actively comparing or ready to buy. This is ecommerce and retail only.

Most of the coverage of ChatGPT Ads online talks about one or the other and implies it's the whole picture. It isn't. If you don't run a product catalogue, you're not excluded from the channel, you're just running the other format.

How the auction and pricing work

Ads Manager runs a relevance-weighted, second-price auction. In plain terms: you set a maximum bid, but you pay one increment above whatever the next most relevant competing bid was, not your own maximum. A more relevant, better-built ad can win the placement at a lower effective cost than a less relevant ad with a higher bid.

Those figures are OpenAI's own recommended starting points, not a price floor. Early advertiser-reported ranges circulating elsewhere put effective CPCs and CPMs well above those starting figures once accounts have been live a while, though there's no established, audited benchmark yet, this channel is still under a year old. Treat any specific number you see, ours included, as directional rather than gospel until you have your own account's data.

Which markets are live

Availability has moved fast. The original launch covered nine markets: the US, UK, Canada, Australia, New Zealand, Japan, South Korea, Brazil and Mexico. From 31 August 2026, OpenAI opened self-serve access to a further 31 European markets and 12 across the Middle East and North Africa, taking the total to 52 countries by early September 2026.

Regulated categories, financial services, healthcare and legal, sit behind manual, case-by-case approval outside the US, and are largely restricted outside America entirely. Worth checking before you plan a launch date around this channel if you're in one of those sectors.

How it stacks up against Google and Meta

Early, publicly circulating comparisons put ChatGPT Ads somewhere between search and social on cost, but the framing that matters more than any single number is funnel stage. Google Search captures intent that already exists. Meta builds awareness and interest. ChatGPT sits at awareness and consideration, someone mid-conversation, working through a decision, often before they've typed anything into a search bar at all.

The practical read: it's not a replacement for Search or Meta, it's a third funnel stage most accounts aren't currently paying for at all.

Is it worth running yet

For some businesses, not yet, and we'd rather say that here than on a call. It's worth running if you already have proven paid demand elsewhere, margin that tolerates a $3 to $5 CPC while the auction is still early, and conversion tracking that actually works today. It's not worth it yet if you're chasing cheap clicks, if your buyers mostly sit on ChatGPT Plus or Business plans and never see an ad at all, or if you can't yet attribute a sale or a lead back to the channel.

The upside of moving early is real: the auction is under-priced relative to where it'll likely land once more advertisers show up, and that's a temporary window, not a permanent condition. But an under-priced auction with no measurement behind it just produces clicks you can't account for.

FAQ