Ads in ChatGPT are clearly labeled paid placements that can appear below an organic ChatGPT response. They are selected using the context and intent of a conversation, but OpenAI says advertisers cannot influence the answer, access individual conversations, or buy placement inside the model’s response. For marketers, this creates a new channel between search and social advertising: the user is not scrolling a feed or typing a short query. They are describing a problem, comparing alternatives, and moving toward a decision.
OpenAI began testing ads with logged-in adult users on the Free and Go plans in the United States on February 9, 2026. By August 31, the company reported availability in more than 40 countries, tens of thousands of participating advertisers, and a $1 billion annualized advertising revenue run rate. That is not an experiment marketers can dismiss as a novelty. It is the beginning of a new intent market.
How ads in ChatGPT work
Placement is separate from the answer
An ad may appear after a ChatGPT response when OpenAI’s advertising system determines that a relevant sponsored product or service is available. The placement is labeled as sponsored and visually separated from the answer.
According to OpenAI’s advertising principles, advertising does not change the model’s answer. Advertisers cannot pay to be named, ranked, or recommended inside the organic response.
That distinction matters. There are now two different ways a company can appear in ChatGPT:
- Organic visibility: The model cites, mentions, or recommends the company because its sources and retrieval systems consider the information useful.
- Paid visibility: A sponsored placement appears below the answer through the advertising platform.
Buying the second does not secure the first. Brands still need authoritative content, third-party mentions, structured information, and strong digital entities if they want organic visibility in AI-generated answers.
Delivery uses conversational intent
Traditional paid search begins with a compact query such as “best project management software.” A ChatGPT conversation can contain the user’s team size, current tools, budget, technical constraints, previous frustrations, and required integrations.
OpenAI says its delivery system can consider:
- The context and intent of the current conversation
- The ad’s landing page, title, description, and image
- Context hints supplied by the advertiser
- Selected signals from the user’s broader ChatGPT experience when personalization is enabled
- Expected relevance and campaign outcomes
Context hints describe conversations, topics, or keywords where an offer may be useful. They guide matching, but they do not operate as guaranteed exact-match keywords.
That changes campaign architecture. A company selling accounting software should not build one generic “best accounting platform” ad. It should develop separate creative for freelancers managing quarterly taxes, agencies billing retainers, retailers tracking inventory, and growing companies replacing spreadsheets. Each situation represents different intent, objections, and proof requirements.
Privacy and eligibility have explicit boundaries
Ads can appear for eligible Free and Go users. Plus, Pro, Business, Enterprise, and Edu accounts remain ad-free. OpenAI also says ads are not shown to accounts identified as belonging to users under 18.
Advertisers do not receive individual conversations, chat histories, memories, or personal details. They receive aggregated campaign data such as views and clicks. Users can manage personalization, hide or report ads, inspect why an ad appeared, and delete advertising data. Deleted advertising data may be retained for up to 30 days before removal.
OpenAI also restricts placement near sensitive or regulated conversations, including areas such as health, mental health, and politics. The platform’s policies and controls will continue to evolve, so campaign approval should be treated as an ongoing compliance process rather than a one-time checkbox.
What advertisers can buy, measure, and optimize
OpenAI Ads Manager supports objectives designed for reach, traffic, and conversions. Advertisers set budgets at the campaign level and maximum bids at the ad-group level.
Cost and objective breakdown
| Objective | Billing model | Primary use | Published cost guidance |
|---|---|---|---|
| Reach | CPM | Awareness and message distribution | Advertiser sets a maximum bid per 1,000 impressions |
| Clicks | CPC | Qualified site traffic | OpenAI recommends an initial maximum bid of $3–$5 per click |
| Conversions | oCPC | Downstream actions after a click | Delivery optimizes toward the configured conversion event |
There is no universal price for a campaign. Actual costs depend on bid competition, delivery eligibility, expected outcomes, and relevance. OpenAI describes its selection process as a relevance-weighted, second-price auction, meaning the largest bid does not automatically win every eligible placement.
A $5 click from a user comparing three vendors can be more valuable than a $1 click from a vague search query. But that advantage must be demonstrated with conversion data. Conversational context is a signal of potential intent, not proof of revenue.
The ad unit
The initial format contains familiar components:
- Advertiser name
- Favicon or logo
- Headline
- Description
- Landing-page URL
- Image creative
The format looks simple. The matching environment is not.
OpenAI’s creative guidance for ChatGPT Ads recommends specific, benefit-focused messaging and multiple distinct creative variations. Repeating one slogan across ten ads gives the system little useful information. A better library changes the audience situation, benefit, proof point, or use case in each variation.
Landing pages also matter to delivery. They must be reachable by OpenAI’s advertising and search crawlers, accurately reflect the offer, and align with the ad’s language. A technically blocked or semantically weak landing page can undermine an otherwise strong campaign.
Measurement
Ads Manager Beta reports:
- Impressions
- Clicks
- Spend
- Click-through rate
- Average CPC
- Average CPM
- Conversions
Advertisers can configure conversion measurement and attach static UTM parameters to destination URLs. Those parameters persist after the click, allowing traffic to be analyzed in existing analytics platforms.
That is the minimum measurement layer. Serious advertisers should also track qualified leads, opportunities, acquisition cost, revenue, payback period, and conversion lag. Optimizing for clicks alone invites the same failure mode marketers have spent years learning on other platforms: cheaper traffic that produces weaker customers.
Why conversational advertising is different from paid search
ChatGPT Ads share mechanics with paid search, but the user experience is closer to guided decision support. The system has more context, the user is often deeper into a problem, and the ad arrives after an answer rather than above a list of links.
| Dimension | Paid search | Paid social | ChatGPT Ads |
|---|---|---|---|
| Primary signal | Search query and keywords | Audience profile and behavior | Conversation context and intent |
| User state | Actively searching | Browsing or consuming content | Exploring, comparing, or deciding |
| Placement | Search-results page | Feed, story, or video | Below a generated answer |
| Creative strategy | Query-to-ad relevance | Attention and audience fit | Situation-to-solution relevance |
| Optimization risk | Overfitting to keywords | Chasing low-cost engagement | Mistaking rich context for purchase readiness |
| Organic relationship | SEO and ads coexist | Organic and paid feeds coexist | Organic AI visibility and sponsored placement remain separate |
The practical shift is from keyword lists to intent models.
A keyword tells you what someone typed. A conversation can reveal what they are trying to accomplish, why previous solutions failed, what constraints matter, and what outcome they value. That creates the possibility of more relevant advertising, but it also punishes generic positioning.
“Powerful software for modern teams” says almost nothing. “Replace manual weekly reporting across five client accounts without rebuilding your analytics stack” describes a problem, an audience, a moment, and a result.
This is where agentic marketing becomes useful. A campaign system can continuously classify intent, produce controlled creative variations, monitor downstream conversions, flag policy risk, and shift budgets. The human still defines positioning, acceptable claims, financial guardrails, and the final standard. Agents handle the repetitive analysis and execution loop.
How to build a campaign system that can learn
1. Map situations, not just keywords
Start with the conversations customers have before buying:
- What problem triggered the search?
- What have they already tried?
- Which alternatives are they comparing?
- What constraint could block the purchase?
- What evidence would reduce risk?
- What action represents real commercial intent?
Turn those answers into a structured intent map. A B2B software company might identify migration, integration, security review, pricing comparison, executive approval, and vendor-replacement conversations. Each deserves different creative and a matching destination page.
2. Build a controlled creative library
Create multiple ads for each important use case. Vary one meaningful element at a time: audience, problem, benefit, proof, objection, or offer.
Every claim should be traceable. Every image should match the landing page. Every landing page should answer the same problem described in the ad. This gives the delivery system useful alternatives while preserving enough structure to identify why one message performs better.
OpenAI’s advertiser site currently highlights companies including Newegg, Best Buy, Lowe’s, and VistaPrint as early participants. Their presence is evidence that established performance teams are testing the channel. It is not evidence that every advertiser should copy their creative or economics.
3. Connect the complete measurement chain
Track the path from impression to business outcome:
impression → click → engaged visit → conversion → qualified lead → opportunity → revenue
If reporting stops at the click, the campaign cannot learn which conversations produce valuable customers. Use consistent UTMs, define conversion events before launch, pass lead-source data into the CRM, and compare platform reporting with first-party analytics.
Measure incrementality as the account matures. Some users exposed to an ad may have converted through another channel anyway. A platform-reported conversion and a genuinely incremental customer are not always the same thing.
4. Use agents for operations, not judgment theater
At BattleBridge, we operate 10 deployed AI agents across three servers with 46 registered skills. Those systems support real production assets, including a senior-living directory covering 977 cities, 51 states, and 4,757 communities, plus a CRM containing 8,442 contacts.
The lesson is not that “AI makes marketing faster.” Speed without controls produces mistakes faster.
A production advertising machine needs specialized responsibilities: research, intent classification, creative generation, policy checking, landing-page analysis, bid monitoring, attribution, and exception handling. Our multi-agent architecture separates those jobs so one model is not pretending to be strategist, copywriter, analyst, compliance reviewer, and media buyer simultaneously.
5. Establish financial and policy guardrails
Before spending, define:
- Maximum daily and campaign budgets
- Maximum acceptable CPC and customer-acquisition cost
- Minimum conversion volume before making decisions
- Claims that require human approval
- Restricted products, audiences, and conversation contexts
- Conditions that pause an ad or campaign
- The human responsible for final accountability
Automation should shorten the gap between signal and action. It should never eliminate ownership.
Frequently asked questions
Does ChatGPT have advertisements?
Yes. OpenAI launched its initial United States test on February 9, 2026, then expanded into additional international markets. Ads may appear below responses for eligible users on the Free and Go tiers.
How much do ads in ChatGPT cost?
OpenAI supports CPM, CPC, and conversion-optimized campaigns. Its published guidance recommends beginning standard CPC campaigns with a maximum bid of $3–$5 per click, but actual prices vary by auction conditions and relevance.
Who can advertise in ChatGPT?
Eligible businesses in supported countries can use OpenAI Ads Manager or approved partner channels. Every advertiser, ad, destination, and offer remains subject to OpenAI’s policies and review process.
Does OpenAI give advertisers access to conversations?
No. OpenAI says advertisers cannot see individual chats, chat histories, memories, or personal details. Campaign reporting provides aggregated performance data rather than users’ conversation content.
How do I create an ad for ChatGPT?
Open an Ads Manager account, select a campaign objective, establish the budget and bids, create ad groups, upload creative, configure measurement, and submit the campaign. Build the account around customer situations and outcomes instead of copying a paid-search keyword structure.
The window is open, but the winning strategy is not “run another campaign.” Build a system that can understand intent, generate controlled creative, measure revenue, and improve without losing human accountability.
Build My AI-Native Advertising System
Start with a defined budget and measurable pilot. No black-box spending, no long-term commitment, and no pretending clicks are revenue.
BattleBridge applies the same operating model behind 10 deployed agents, 46 registered skills, and production systems managing thousands of pages, communities, and contacts.