ChatGPT Ads: What Changes When the Search Box Talks Back
So OpenAI is selling ads inside ChatGPT now. And the most interesting part isn't the ads themselves — it's that the targeting signal is the conversation you're having.
Google targets what you type. Meta targets who you are. ChatGPT targets what you're thinking through.
I've been doing GEO work for months and thinking about where paid acquisition fits in the strategy. When I saw the ChatGPT Ads announcement, my first thought was: "I've been building for this surface without knowing paid was coming." My second thought was that the speculation-to-fact ratio online is roughly 5:1. So I spent the week separating what's real from what people made up.
Here's what I found.
What ChatGPT Ads look like right now
OpenAI started testing ads on February 9, 2026. Free and Go ($8/month) users in the U.S. see them. Plus, Pro, Business, Enterprise, and Education subscribers don't.
The format is simple. A sponsored unit shows up at the bottom of ChatGPT's answer. Labeled, visually separated, matched to the topic of your conversation. No interactive elements. No multi-turn dialogue. Just a link below the response.
What OpenAI has put in writing:
- Ads don't touch the answers. Responses optimize for helpfulness, not revenue.
- Advertisers get aggregate data only. Impressions and clicks. Not your chat history.
- Targeting is contextual. Based on conversation topic, past chats, and ad interactions. Research recipes, see meal kit ads.
- Sensitive topics excluded. Health, mental health, politics — no ads.
- Users control it. Dismiss ads, turn off personalization, delete ad data, see why an ad appeared.
Now, the pricing. The Verge reports CPMs around $60 — triple Meta's typical rates — with a $200K minimum commitment. No self-serve ads manager exists yet. This is NFL broadcast pricing, not "start at $10/day" territory.
Scale context: ChatGPT has 800+ million weekly active users. Even limited to free tiers, the audience is massive.
Why conversations change the ad model
ChatGPT Ads match on conversation context — the full thread, not a single query. That's a structural shift from Google, where the targeting signal is one search.
Google Ads get one signal — a query. You type "best project management software" and Google matches ads to those words. One shot.
ChatGPT Ads get the whole thread. Someone asks "What causes low water pressure?" then follows up with "We have galvanized pipes" then asks "What's the typical cost to replace them?" The system watches someone move from curiosity to pricing research in real time. That's a fundamentally different targeting signal than a single search query.
OpenAI's January announcement frames it this way: "People often use ChatGPT when they're actively exploring options, comparing ideas, or working toward a decision."
That's mid-funnel positioning. Google owns high-intent queries. Meta owns visual awareness. OpenAI is going after the research phase — when people are thinking out loud with an AI.
ChatGPT Ads features: confirmed vs. made up
OpenAI has confirmed six features. Third-party marketing sites have invented at least five more and presented them as real. This is where the ecosystem gets messy — and frankly, where I spent most of my time fact-checking. A growing number of sites are publishing breakdowns of "conversational phase targeting," "Dynamic Response Ads," and "expertise inference." These sound like real product features. They're not. Nobody at OpenAI has confirmed any of them.
What OpenAI has confirmed
| Feature | Status | Source |
|---|---|---|
| Sponsored units below answers | Live (testing) | OpenAI, Feb 2026 |
| Contextual matching on conversation topic | Live | OpenAI, Feb 2026 |
| Past chats influence ad selection (opt-in) | Live | OpenAI, Feb 2026 |
| Sensitive topic exclusion | Live | OpenAI, Feb 2026 |
| Answer independence (ads don't affect responses) | Live | OpenAI, Jan 2026 |
| Future interactive formats (hinted) | Not live | OpenAI, Jan 2026 |
What third parties are speculating
| Concept | Source | Reality |
|---|---|---|
| "Conversational phase targeting" (Exploration → Diagnosis → Specification) | chatgptadstrategy.com | Not an OpenAI feature |
| "Dynamic Response Ads" (multi-turn ad dialogues) | chatgptadstrategy.com | Not an OpenAI feature |
| Expertise inference (beginner/advanced targeting) | chatgptadstrategy.com | Not an OpenAI feature |
| Topic cluster seeding (replacing keyword lists) | Various marketing blogs | Not an OpenAI feature |
| Session-depth bidding, conversational phase bidding | chatgptadstrategy.com | Not an OpenAI feature |
I want to be fair — some of these speculative concepts are directionally right. The technical capability exists. But there's a difference between "this could probably happen" and "this is a feature you can use today." The marketing ecosystem around ChatGPT Ads has already outrun the actual product.
Likely future ChatGPT Ads features
Three things I'd bet on, even though none are confirmed.
Conversation-aware targeting gets more granular
The model already understands the difference between "I'm curious about project management" and "I'm comparing tools and asking about pricing." Exposing that as a targeting lever is a product decision, not a technical one. Whether they call it "conversational phase targeting" or something else, the capability is sitting right there.
Ad formats will become interactive
OpenAI said this directly in their January post: "Soon you might see an ad and be able to directly ask the questions you need to make a purchase decision." The current sponsored link format is version one. They've told us there's a version two.
Customer language beats keywords
This one doesn't require any prediction at all. If targeting matches conversation context instead of keyword strings, then the businesses that understand how their customers actually describe their problems will win. That's true today with the current format.
Where ChatGPT Ads fit in the ad stack
| Platform | Best for | Targeting model | Access |
|---|---|---|---|
| Google Ads | High-intent queries, brand defense | Keywords and query matching | Self-serve, any budget |
| Meta Ads | Visual discovery, broad awareness | Audience graphs and behavior | Self-serve, any budget |
| ChatGPT Ads | Mid-funnel research and consideration | Conversation context | $200K minimum, managed |
The gap between Google/Meta and ChatGPT on access alone is striking. If you're an indie maker or small business, this channel doesn't exist for you yet. OpenAI says they want to support "businesses of all sizes" eventually. Today, it's big brands only.
One more difference worth noting: retargeting. Google and Meta use pixels to know you visited a website. OpenAI confirmed that past chats can influence future ad selection if you opt in. That's conversational retargeting — the system knows you discussed a topic yesterday and shows a related ad today. But the detailed retargeting playbooks circulating online (same-day hot leads, two-week nurture windows) are frameworks people built on top of that single confirmed data point.
If you've been doing GEO, you're already halfway there
GEO and ChatGPT Ads operate on the same surface — AI conversations. If you've been doing GEO work, you've built the foundation for ChatGPT Ads without planning to. This is the part that clicked for me.
GEO gets your content cited as an organic source. ChatGPT Ads place a paid unit in that same conversation. Different mechanism, same environment.
But the connection goes deeper than that.
Same input: customer language
GEO pushes you to write around real questions people ask. Answer-first structure, FAQ schema, direct responses. ChatGPT Ads need the same input — if targeting is contextual, the businesses that understand customer language write better ad copy and get better matches. The content you've structured for AI citation is the content that informs good ad creative.
Same structure: answer-first and modular
Clear headings, atomic paragraphs, modular answers — the GEO playbook. If OpenAI builds interactive ad formats where users ask follow-up questions, the businesses with modular, answer-ready content adapt first.
Same strategic bet: AI conversations as the surface
Whether organic or paid, both are a bet that AI conversations are becoming a primary surface for research and decisions. Investing in GEO content and understanding ChatGPT Ads is the same strategic position viewed from two angles.
What ChatGPT Ads are bad at
Every new platform gets overhyped. Here's where the limits are right now.
Enterprise pricing with no SMB path. $200K minimum at $60 CPM. The ROI math is completely different from Google's $10/day starting point.
Weak measurement. Aggregate impressions and clicks only. No conversion tracking, no purchase attribution. Early guidance says to track branded search lift and direct traffic as proxies. That's a step backward from what Google offers.
Free-tier audience only. Plus, Pro, Business, Enterprise, Education users never see ads. For B2B products targeting professionals who pay for tools, the audience match could be off.
Visual and impulse products. Fashion, home decor, food — anything that sells on visual inspiration — Meta's visual discovery is still better. ChatGPT is text-based problem-solving.
Category creation. If nobody's talking about your product category yet, contextual targeting can't find those conversations. This works for known problems, not new ones.
No self-serve tools. Managed placement only. No testing, no iteration at your own pace. OpenAI has a sign-up page but no campaign manager.
How to prepare for ChatGPT Ads now
You don't need $200K to prepare. And the preparation pays off even if you never buy a ChatGPT ad.
Build a customer language library. Pull 100 real questions from support tickets, sales calls, and feedback. Organize by buying stage — curiosity, comparison, ready to decide. This is your foundation for GEO content and future ad targeting.
Map your sales process backward. Where do deals close? Where do they stall? Where do they die? Those moments tell you which conversations matter — and where you want your content, organic or paid.
Write answer-first content. Structured, modular responses to real customer questions. Works for GEO (gets cited by AI). Works for ChatGPT Ads (matches conversational context). Same content, two channels.
Keep doing GEO. ChatGPT Ads don't invalidate GEO. They raise the stakes. The more AI conversations become a surface for both organic and paid, the more valuable it is to have content that performs well there.
Watch the product, ignore the hype. The speculation ecosystem is moving faster than OpenAI's actual releases. Follow OpenAI's official announcements for what's real. Everything else is educated guessing — sometimes useful, not something to build a budget around.
GEO and ChatGPT Ads: the bigger picture
I wrote about how GEO is what good writing has always been — clear structure, real data, direct answers. Now there's a paid side to that same coin.
The structural advantage is real: no other ad platform targets a multi-turn problem-solving dialogue. Whether OpenAI builds the tooling to fully exploit that — granular phase targeting, interactive formats, self-serve campaigns — is an open question.
I'm still forming my take. The platform is days old. Pricing will change. Formats will evolve. Some speculated features will ship; some won't. But the underlying shift — ads inside AI conversations instead of next to search results — is worth understanding now, even while the specifics move.
The best move stays the same: understand what your customers ask, answer those questions directly, structure the content so both humans and AI can extract the useful parts. If ChatGPT Ads mature into what the speculation suggests, you're ready. If they stay a sponsored link at the bottom, your content is still better for the work.
Either way, the content compounds.