OpenAI just published its approach to advertising in ChatGPT, and it's one of the more important ad ecosystem updates in a while.
Not because "new inventory," but because it's a new interface for intent. And OpenAI is setting expectations early around trust, privacy, and where ads sit in the experience.
If you run a performance agency, this is the post to read before you pitch "ChatGPT ads" to a client or let your team chase rumors.
What OpenAI announced
OpenAI says it plans to start testing ads in the U.S. "in the coming weeks" for logged-in adults on the Free and ChatGPT Go tiers.
Plus, Pro, Business, and Enterprise remain ad‑free.
The initial format: ads at the bottom of answers when there's a relevant sponsored product/service based on the current conversation.
Users can see why they're seeing an ad, dismiss it, and give feedback.
Ads won't appear for users who say they're under 18 or are predicted to be under 18, and ads won't be eligible near sensitive or regulated topics like health, mental health, or politics.
Then OpenAI lays out five principles that matter for anyone who cares about how "ads in AI" will work:
That's the ground truth. Now, here's what it means for your agency.
The strategic shift: "intent" becomes conversational, not query-based
Search ads are built around a single moment: the query. Social ads are built around interruption and targeting.
ChatGPT is different. The user is already in a problem-solving flow. OpenAI's initial design puts ads after the answer, not inside it, and ties relevance to the current conversation.
That's not a small UX detail. It creates a different bargain:
If that holds, "winning" will look less like persuasion and more like being the most defensible next step when someone is already convinced they need something.
For agencies, this is closer to high-intent search than to classic feed advertising, but with a twist: the question can evolve across turns, so the context can get richer than a single keyword.
Why OpenAI's principles matter to advertisers (and why most people will miss this)
1) You can't buy the answer
OpenAI is saying the organic answer is optimized for what's helpful, and ads won't influence it.
For agencies, this matters in two ways:
2) The targeting model will not look like Meta
OpenAI is emphasizing privacy and saying it won't sell conversation data to advertisers. Also, users can turn off personalization.
So even if there's personalization, you should assume a large percentage of impressions will be driven by contextual relevance (the conversation), not hyper-granular identity targeting.
That changes how you win:
3) Expect "limited, premium" inventory at first
OpenAI says it's not optimizing for time spent in ChatGPT, prioritizing trust and UX over revenue.
If you take that seriously, early ad load likely stays constrained. That's an inference, not a guarantee, but it's consistent with the stated incentives.
Translation for agencies: don't pitch this as "we'll scale spend." Pitch it as "we'll win high-intent pockets and learn fast."
4) Regulated and sensitive adjacency will be stricter than most ad platforms
OpenAI says ads won't be eligible near health, mental health, or politics.
That's a big deal:
The agency playbook: how to approach ChatGPT ads without looking naive
Here's what I'd implement as an agency owner the moment this opens up to advertisers.
Step 1: Pick the right pilot clients
Start with clients who have:
Avoid for early pilots:
Step 2: Rewrite your "creative" as decision support
In ChatGPT, the user is already reading an answer. Your ad has to feel like an extension of decision-making.
So the unit you build is:
If the ad can't survive a user asking "is this actually true?" inside the same interface, it will underperform. OpenAI even points at a future where users can ask questions directly after seeing an ad.
Step 3: Instrument for intent, not just clicks
OpenAI hasn't published the reporting spec in this post, so don't assume it will look like Google Ads.
But you can still set yourself up:
If your client only measures last-click ROAS, you'll under-report impact on a conversational channel.
Step 4: Prepare a "brand truth doc"
Because the interface is conversational, your brand is going to be questioned.
Build a one-pager per client:
This makes you faster and keeps you out of compliance trouble when things get interactive.
Step 5: Set the right expectation with clients
Your pitch should be:
That framing protects retention when the channel is still early.
What to watch for as OpenAI expands the program
OpenAI has been explicit about principles, but not yet about the mechanics. Before you commit real budget, get clarity on:
If you run a serious agency, the first advantage won't be clever creative. It'll be operating discipline: fast learning loops, clean instrumentation, and clear client comms.
The bigger takeaway for agencies
Most ad platforms optimize for attention. OpenAI is stating it will optimize for trust and usefulness, and that ads won't influence answers.
If they execute on that, it creates a new kind of paid channel:
That's good news for agencies that run tight systems. And it's bad news for agencies that rely on volume, vague claims, and "we'll test creatives until something works."