Inside ChatGPT Ads: Formats, Targeting and Costs Explained

ChatGPT is officially entering its ad era. In February 2026, OpenAI began testing ads in ChatGPT in the United States for logged-in adult users on the Free plan and the Go plan. Ads do not influence ChatGPT’s answers; they appear separately and are clearly labeled as sponsored. 

For marketers, that unlocks a new channel that sits somewhere between search and social: a conversational interface where people are actively researching, comparing options and making decisions. For brands, it raises a new set of questions about formats, targeting and costs, plus what “good” looks like for ChatGPT marketing campaigns in a world of AI advertising.

Here’s what we know right now, what is still evolving and how to think about an advertising strategy that actually fits the moment.

What are ChatGPT ads?

ChatGPT ads are paid placements that can appear below the end of a ChatGPT response. They are visually separated from the organic answer and labeled as sponsored.

OpenAI has been direct about two non-negotiables:

• Answer independence: advertisers cannot shape, rank or alter responses
• Conversation privacy: advertisers do not get access to your chats, chat history or memories, and OpenAI says it does not sell your data to advertisers

That matters because this channel will live or die on trust. If users feel like paid placements are steering answers, performance will not save it.

Ad formats: what shows up today

OpenAI is starting small. During the current test, users may see a single ad unit below a response when there is a relevant match to the conversation. That unit can feature one or more items from an advertiser. 

From OpenAI’s own examples, the early format looks like a sponsored product or service card that includes key details like the item name and price with a clear sponsored label.

A few placement rules also matter for planning:

• Ads appear below responses, not inside the answer itself 
• Ads do not appear in Temporary Chats, when you are logged out, after you generate an image or in the ChatGPT Atlas browser 
• Ads are not eligible to appear near sensitive or regulated topics like health, mental health or politics, and some sensitive ad verticals are excluded at this time 

For a new ad platform, this is a pretty deliberate first step: one placement, clear labeling, limited surfaces.

Targeting: how ads get matched to a user

ChatGPT ad targeting is not keyword bidding in the traditional sense. Think “context matching” first, then personalization if the user allows it.

OpenAI says it starts by matching ads to what’s being discussed in your current chat thread. If a user opts into personalized ads, OpenAI may use additional signals such as past chats, memory and how you’ve interacted with ads to improve relevance over time. 

OpenAI also notes that “basic context” like general location or language may be used for relevance, depending on settings. 

A few important boundaries are worth calling out:

• Advertisers only receive aggregated, non-identifying reporting, such as total views or clicks
• OpenAI states advertisers never receive chats, chat history, memories, name, email, precise location or sensitive information 
• Users can turn off ad personalization, clear ad data, hide or report ads and see why they are being shown an ad 

From a marketer’s perspective, this is “conversational AI marketing” with a new targeting layer: you are buying proximity to an intent-rich moment inside a chat thread.

Costs: what ChatGPT ads are priced at right now

OpenAI has not published a public rate card on its advertiser interest page, and access is still gated through a pilot and partner agencies. 

That said, reputable industry reporting has started to outline early pricing for the beta:

• eMarketer summarized the beta as impression-based buying at around $60 CPM with minimum commitments in the ~$200,000 range, citing Adweek and other reporting. 

Two implications for ad budgets:

1. This is currently an enterprise channel, not a self-serve button like Meta or Google Ads
2. The measurement package is still limited compared with mature platforms, since OpenAI says advertisers get aggregated views and clicks today 

In other words, expect media buying and campaign management to look more like the early days of a premium beta than a performance marketer’s dream dashboard.

What does “performance” mean for ChatGPT ads?

If you treat this like standard paid search, you may get frustrated fast. ChatGPT ads are closer to “high intent discovery” than “click, land, convert” for many categories.

Here’s how we would frame early ChatGPT ad performance:

Primary metrics

• Views and clicks that OpenAI shares in aggregate today
• Click quality on your side: engaged sessions, scroll depth, time on site, add-to-cart, lead start

Downstream metrics

• Assisted conversions in analytics
• Brand lift signals: branded search growth, direct traffic lift, returning users
• Conversion rate by landing page and offer

Because chat is a research environment, attribution is going to be messy. Plan for that upfront, especially if you are comparing this to channels with clearer last-click behavior.

Creative: How to build ads that fit the interface

This is where ad creative and brand messaging do the heavy lifting. A sponsored card below an answer has to feel useful or it gets ignored.

What tends to work best in conversational environments:

• Keep ad copy outcome-oriented: what problem does this solve right now
• Match the user’s language, not your internal product taxonomy
• Use specifics that reduce decision friction: price, timing, guarantee, delivery window, location availability
• Make the landing page feel like the natural continuation of the chat, not a hard pivot

If you are using AI content generation to draft variations, treat it as a starting point. Your differentiator still needs a human brain: category insight, customer nuance and a voice that feels like you.

Prompts for ads: using ChatGPT as an internal production tool

Even if you never buy ChatGPT ads, most teams are already using ChatGPT for business workflows to move faster. That includes ad ideation, testing angles and building creative matrices for campaigns.

A simple prompting framework we like for AI-powered ads:

• “Write three headline options for [audience] who want [outcome] but worry about [objection]. Keep it under [character count].”
• “Give me five variations that sound premium, practical and bold. No hype.”
• “Turn this offer into benefit-led copy for a sponsored product card. Include one specific proof point.”

Then validate against reality: your actual reviews, your actual differentiation and your actual conversion data.

How to get started with ChatGPT ads

Right now, the most accurate “how to start” is also the simplest: OpenAI is collecting interest through its advertiser page while it learns from the early test. 

If you are planning, focus on the fundamentals that will carry over regardless of format changes:

• Clarify your highest intent use cases and the conversations that signal them
• Tighten landing pages so they convert intent into action fast
• Build a testing plan that connects creative themes to business outcomes
• Prepare for limited platform metrics and stronger reliance on your own analytics

The bigger takeaway for 2026

ChatGPT ads are not just a new placement. They are a signal that artificial intelligence marketing is moving from “tooling” to “media.” The channel is young, the formats are intentionally constrained and the costs are currently premium, but the direction is clear: conversational interfaces are becoming buying moments.

The brands that win will be the ones that show up usefully, protect trust and treat this like a new kind of performance problem to solve, not a copy-paste extension of their existing playbook.