OpenAI tests image-based ads in ChatGPT for 1.2 billion weekly users

OpenAI will begin testing image-based advertisements inside ChatGPT later this month, targeting a user base the company now puts at 1.2 billion people every week. The figure, shared in an OpenAI blog post on Monday, makes ChatGPT the largest AI-native consumer platform in the world, according to the company.

The ad rollout marks the next step in OpenAI's effort to build a revenue stream beyond subscriptions. OpenAI framed the move as part of its core mission. "Advertising helps us bring the benefits of AI to more people," the company wrote. "That broad access is central to our mission, and a strength of our business."

The post did not specify which month the company means by "later this month," or name the initial group of US advertisers.

What does the new ad format look like?

The new format appears during image generation inside ChatGPT. It displays product imagery meant to help users "imagine how products and services could fit into their lives," OpenAI said.

Three creative categories anchor the unit:

• Product inspiration
• Product usage scenarios
• Experiences the product enables

Ads display separately from the image a user is creating. OpenAI said every unit carries a clear label. "Advertising does not influence the answers ChatGPT provides," the company added.

Testing begins later this month in the United States. OpenAI will run the format with an initial group of advertisers before any wider release.

Why is measurement a focal point?

OpenAI is also expanding the tools advertisers use to track campaign performance. New integrations with Hightouch, Tealium, and LiveRamp let brands pipe conversion data from their existing stacks into ChatGPT Ads.

The company also extended its attribution partner list. AppsFlyer, Triple Whale, Adjust, DV Rockerbox, Northbeam, Branch, Singular, Kochava, Airbridge, and Tenjin now support Conversions API, reporting, and click attribution inside ChatGPT Ads.

For incrementality testing, OpenAI is working with Haus, Measured, and WorkMagic on geo-based experiments. The partnerships aim to measure the causal lift of ChatGPT Ads, not just last-click attribution.

OpenAI shared three early performance figures from those partners:

• WeightWatchers saw an attributed cost per acquisition on ChatGPT Ads 15.3% lower than its blended paid-search benchmark, per DV Rockerbox.
• Dose, a wellness brand, recorded a statistically significant incremental lift, with 67% of incremental purchases coming from net-new customers, according to WorkMagic.
• Portland Leather drew 93% of its ChatGPT Ads visitors as new users, per Triple Whale.

Those numbers come from partner reports, not independent audits. OpenAI acknowledged that its incrementality work "is still in its early stages."

How is OpenAI handling brand safety?

Brand suitability sits at the center of OpenAI's pitch to enterprise advertisers. The company relies on placement guardrails that score whether a conversation is appropriate for ads.

The system filters "emotionally vulnerable, sensitive, or otherwise unsuitable contexts," OpenAI said. Automated review, human oversight, and ongoing monitoring form the enforcement layer.

For brand-specific controls, OpenAI now offers Negative Phrases to qualifying advertisers. The tool lets brands block ad placements adjacent to specific terms in the conversation.

OpenAI also announced pilots with two major verification firms:

• DoubleVerify (DV)
• Integral Ad Science (IAS)

Both companies will evaluate how OpenAI applies its brand safety standards in a controlled testing environment. The pilots run "without accessing private user conversations," OpenAI said.

The setup gives independent partners a window into the system without exposing user data. OpenAI's stated goal: greater advertiser confidence without eroding conversation privacy.

What stays off-limits for advertisers?

OpenAI restated three principles it says govern every campaign:

1. ChatGPT's answers stay independent of ad buys.
2. Conversations stay private.
3. Users retain control over their experience.

The principles matter because generative-AI ads raise distinct questions. A sponsored block inside a model output is not equivalent to a banner next to a search result. OpenAI's policy framing attempts to draw that line explicitly.

OpenAI also pointed readers to its advertising policies and guardrails, which the company says are designed to protect users and preserve trust.

How does this fit the broader ad market?

ChatGPT now reaches more people each week than the entire population of any single country. The 1.2 billion figure positions OpenAI to negotiate from strength with advertisers used to the scale of Google and Meta.

The platform's role differs, though. Users come to ChatGPT to ask questions, draft documents, and write code. Ads inside those flows must work alongside reasoning, not interrupt it.

OpenAI is betting that integrated product imagery beats display banners for that setting. The visual format aligns with how ChatGPT already serves image-generation users, and it gives advertisers a way to embed creative directly into a workflow.

The measurement push mirrors moves by Meta, Google, and TikTok. Each platform has invested heavily in attribution and incrementality tooling as advertisers demand proof of return on ad spend. OpenAI is following the same playbook with a younger stack and a narrower set of formats.

What comes next?

The US image-ad test starts later this month. OpenAI has not published a timeline for broader rollout or international expansion. Brands that want early access to Negative Phrases must already meet OpenAI's qualifying criteria.

The pilots with DV and IAS also remain in testing. Advertisers can sign up at ads.openai.com, OpenAI said, and the company will continue adding partners and formats over time.

For now, the company is signaling what an AI-native ad business looks like. The format is visual. The measurement is partner-led. The placement logic stays proprietary. Each piece is still in motion.

Advertisers and privacy researchers will watch how OpenAI threads the needle between ad relevance and user trust. The 1.2 billion weekly users give the company room to experiment. They also raise the cost of getting the model wrong.