OpenAI began testing ads on February 9, 2026. Here’s a plain-language breakdown of the mechanics, the audience, and what it means for your campaigns.

After years of OpenAI saying ads aren’t part of the plan, guess what? 

ChatGPT announced it would start testing sponsored placements, and the first ads appeared on February 9.

 The platform serves 800 million monthly users, so this is a channel we all need to understand.

Below, I’m cutting through the PR language and answering the four questions that actually matter to marketers and advertisers.

How Will ChatGPT Ads Work?

The format is deliberately conservative:

• Ads will appear below the end of a ChatGPT response.
• They will appear after the model finished answering.
• They will always be labeled “Sponsored,” similar to other platforms. 

They are visually separated from the AI’s answer, so while they might count as native, OpenAI says they actually aren’t. They will sit beneath like a footer unit. 

Matching works by topic relevance.  OpenAI’s system looks at what you’re currently discussing in chat and matches that context to advertiser inventory. 

No auction? Seems so. If multiple advertisers are eligible for a given topic, the most relevant one wins the placement. 

During the initial test phase, you’ll typically see a single ad unit beneath a response, which may feature one or more items from a single advertiser. 

Image courtesy of OpenAI

How Many Ad Formats Does ChatGPT Have?

Currently, the only format is the bottom-conversation generated ad. 

However, OpenAI has also hinted at future interactive formats. 

In its launch announcement, the company noted that “soon you might see an ad and be able to directly ask the questions you need to make a purchase decision.”

This signals conversational ad units that let users engage with an advertiser directly in the chat interface. That’s not live yet, but it’s the roadmap. 

It’s like a CRM touchpoint integrated into the moment of intent. Which sounds powerful in theory.

Who Will See ChatGPT Ads?

The short answer: free and Go-tier users in the United States. 

The Go plan launched globally in January 2026 at $8/month and sits between the fully free tier and the $20/month Plus subscription.

Anyone on Plus, Pro, Business, Enterprise, or Education plans will not see ads. 

This is a deliberate choice that protects OpenAI’s paying subscriber base from the friction of advertising and creates a clean upsell motion. 

Pay more, lose the ads. It’s the same playbook Spotify and YouTube have run for years.

The free tier is the audience. And the free tier of ChatGPT is enormous.

— Assad Awan, OpenAI Ads & Monetization Lead, OpenAI Podcast

The addressable pool here is significant. Free and Go users represent the vast majority of ChatGPT’s 800 million monthly active users. 

And unlike a passive scroll on a social feed, these users are in an active decision-making state where they’re researching, comparing, planning.

That’s unusually high intent for an ad impression.

Will Ads Impact ChatGPT Answers?

This is the trust question and OpenAI has been unusually direct about it. 

The answer, per OpenAI’s official communications, is no: ads do not influence what ChatGPT says.

Here’s the structural reason why: the model itself does not know when ads are being shown. 

According to OpenAI’s ads lead Assad Awan on the OpenAI podcast, the ad layer is technically separate from the model’s inference pipeline. 

ChatGPT generates its answer, then the ad system independently decides whether to append a sponsored unit below it. 

The model cannot reference the ad in its response unless a user explicitly asks about it.

Awan also described an internal priority framework: user trust ranks above user value, which ranks above advertiser value, which ranks above revenue. 

Whether that holds as commercial pressure scales is the open question every marketer should be watching. For now, the architecture enforces the separation. 

The ad system and the model are not in a feedback loop.

Will ChatGPT Ads Be Personalized?

Yes, ChatGPT ads will be personalized. But with meaningful guardrails, it’s opt-in beyond the baseline. Here’s how the targeting layers stack:

Default targeting (always on): OpenAI matches ads to the topic of your current chat thread, your general location, and your language. This is pure contextual targeting — no cross-session history involved.

Personalized targeting (opt-in, on by default): If you leave personalization enabled in settings, OpenAI will also draw on your past chat threads, how you’ve interacted with previous ads (what you hid, clicked, or dismissed), and if you enabled memory, your stored memories and recent conversations. 

The company is explicit that this data stays inside ChatGPT and is never shared with advertisers. 

Advertisers only receive aggregated, anonymized performance data: total views, total clicks. No individual user data, no conversation access, no email or IP address.

Can ChatGPT Users Turn Off Personalization?

Yes, users can turn off personalization at any time, delete their ad history and interests independently, hide individual ads, or report ads that feel inappropriate. 

Disabling personalization doesn’t automatically delete stored data, so users need to actively clear the History and Interests tabs if they want a clean slate.

The Marketer’s Takeaway: Things Worth Acting On

Here’s what we guess will be the core takeaways:

• The format is contextual-first. Start building creatives that reads as a helpful next step, not an interruption — think sponsored recommendations, not banner ads.
• The addressable audience is enormous but US-only right now. Plan international budget allocation for when OpenAI expands the test.
• Sensitive verticals (health, politics) are excluded from ad matching — route those budgets elsewhere during the test phase.
• Advertisers get zero access to conversation data. Your measurement strategy will depend on aggregated signals and downstream attribution, not audience profiling.
• Interactive conversational ad units are on the roadmap. Start prototyping what it means for a user to “ask your ad a question” — that’s a genuinely new creative brief.

OpenAI is taking a deliberately slow approach — describing the current phase as a test designed for learning, not revenue maximization.

Early results and user feedback will shape whether this expands to other regions and surfaces. 

As growth marketers, we should treat this as a chance to get in early on a channel with unusually high contextual intent. The playbook is still being written.