A new advertising surface does not appear often.
Search changed how marketers captured demand. Social changed how they created demand. Retail media changed how brands bought closer to the shelf.
Advertising inside ChatGPT is different again. It sits inside a conversation where someone is exploring, comparing, narrowing options and making decisions in their own words.
After a week inside OpenAI Ads Manager Beta, here is what we think it actually means for media buyers.
The unit of targeting is no longer a keyword. It is a conversation.
Search targets a query. Social targets a profile. ChatGPT Ads targets a conversation in motion.
That matters because people rarely ask ChatGPT the way they search Google.
They do not always type:
“best CRM software”
They ask:
“I run a small agency and need a simple CRM that does not take weeks to set up. I mostly care about client follow-ups, reminders and keeping track of deals. What should I compare?”
That is a different kind of intent. It is longer, messier and often much closer to a real buying decision.
In Ads Manager Beta, you do not build targeting around keyword lists or predefined interest groups. You describe the conversations where your ad should be relevant. OpenAI also positions ChatGPT Ads around reaching people as they explore options, compare alternatives and make decisions, not just search for isolated terms.
For media buyers, this changes the core skill. The advantage moves from audience hacking to intent articulation.
A good context hint will not sound like:
“Target users interested in marketing software.”
It will sound more like:
“Show this ad when a small business owner, marketing manager or agency operator is comparing tools to manage paid campaigns, consolidate reporting, control advertising budgets or simplify work across several ad platforms.”
That is closer to a brief than to a keyword list.
Lifehack: write context hints as if you were briefing a smart junior strategist. Include the user, the situation, the problem, the decision moment and the category alternatives.
The best early buyers will build a library of “conversation moments”
In ChatGPT, the useful question is not only “who is this person?” It is “what are they trying to figure out right now?”
For example, a SaaS brand could test separate context hints around:
“comparing tools before buying”
“looking for alternatives to an existing provider”
“trying to solve a workflow problem”
“asking how to reduce costs”
“building a shortlist for a manager or client”
Each one can produce a different ad angle.
A user asking “what is the cheapest tool?” should probably not meet the same creative as a user asking “what should I present to my CFO?” One is price-sensitive. The other needs confidence, proof and language they can reuse internally.
Real-life example:
For an adtech platform, one context could be about budget control: “I run ads across Meta, TikTok and Telegram and need one place to track spend.” Another could be about market expansion: “I want to launch campaigns in a new geography but do not know the platform and policy requirements.” Same product, different intent, different ad.
Lifehack: before writing ads, create a simple table with three columns: “conversation”, “user anxiety”, “ad promise”. If the anxiety is different, the creative should be different too.
It is small today — and that is the opportunity
OpenAI says ads are currently shown to Free and Go users in the United States, Canada, Australia and New Zealand, and not to Plus, Pro or Business users, or users under 18.
That is a limited map. But “limited” and “early” are exactly the conditions under which serious learning happens. We've written a detailed article on how to create an ads manager account in OpenAI.
There are fewer benchmarks, fewer playbooks and fewer competitors. That makes the first phase less about scale and more about pattern recognition.
A media buyer should not come into this channel asking only:
“Can I scale this next week?”
The better question is:
“What types of conversations are commercially useful for us, and what language makes people click from that context?”
That learning will compound when the platform expands.
Lifehack: do not test ChatGPT Ads with one generic campaign. Build a small learning plan by intent cluster. Even with modest spend, you want to know which conversation type creates signal.
For example:
Campaign: Product discovery
Ad group 1: comparing solutions
Ad group 2: replacing an existing tool
Ad group 3: asking for budget control
Ad group 4: preparing a recommendation for a team
That structure gives you learning you can actually use later.
The interface is simple, so the craft moves elsewhere
Ads Manager Beta is intentionally uncomplicated. OpenAI describes it as a beta platform for creating, launching and managing campaigns, with workflows for campaign creation, bulk upload, performance monitoring, account settings and CSV exports.
That simplicity is good news and bad news.
Good news: there is less operational friction.
Bad news: the interface itself will not be your moat.
Soon, everyone will understand campaign, ad group and ad structure. Everyone will know where to set budget, upload assets and read basic performance.
The real craft will sit in two places:
- how precisely you describe the context
- how useful the ad feels inside the conversation
This is not a place for generic “Grow your business today” copy.
The ad has to feel like a useful next step from the conversation the user is already having.
Weak:
“Try our all-in-one marketing platform.”
Better:
“Launching ads in a new market? Get local platform setup, payment support and campaign guidance in one place.”
Even better:
“Planning TikTok campaigns in a new GEO? Evido helps with account setup, local requirements and campaign launch support.”
The more specific the conversation, the more specific the ad can be.
Lifehack: write ads as answers to unfinished thoughts. The user is already mid-decision. Do not restart the conversation from zero.
Measurement is honest about being early
The current reporting layer is still early. OpenAI says Ads Manager Beta lets advertisers monitor impressions, clicks and spend across campaigns, ad groups and ads, with table views, charts and CSV exports.
OpenAI has also introduced CPC bidding alongside earlier CPM buying, and says it is expanding measurement tools as the platform develops.
That means buyers should avoid judging the channel by mature-platform expectations too early.
This is not Meta after years of pixel history. It is not Google Search with decades of bidding behavior. It is a new surface with a new intent signal.
So the first tests should answer practical questions:
Which conversation contexts generate clicks?
Which ad angles feel native to the environment?
Which landing pages continue the conversation instead of breaking it?
Which offers make sense when the user is still comparing?
Do not send everyone to a generic homepage. If the ad appears in a conversation about “how to launch TikTok ads in a new market”, the landing page should continue that thought.
Lifehack: build landing pages around the same intent as the context hint. Context hint, ad and landing page should use the same mental frame.
Example:
Context hint: user is comparing ways to launch ads in a new country.
Ad: “Launch campaigns in new markets with local setup and platform support.”
Landing page: “What you need before launching TikTok Ads in a new market.”
CTA: “Check launch requirements.”
That is a much cleaner journey than sending the user to a generic services page.
The creative challenge is trust
ChatGPT is not a feed. People are not scrolling passively. They are often asking for help with something specific.
That makes the tolerance for vague advertising lower.
A strong ChatGPT ad should feel like:
“this might help me solve what I am asking about”
not:
“someone bought my attention”
The best early creative will probably be clear, practical and low-friction.
For B2B, that may mean:
“Get a feasibility check”
“Compare launch options”
“See platform requirements”
“Estimate media setup”
“Request market availability”
Not every CTA has to push a demo immediately. In many ChatGPT conversations, the user is still clarifying the problem. A softer CTA can match the stage better.
Real-life example:
If someone is asking how to start advertising on TikTok in Kenya, “Book a demo” may feel too heavy. “Check what is required to launch TikTok Ads in Kenya” is more aligned with the moment. The commercial intent is still there, but the action feels useful.
Lifehack: test one direct-response CTA and one “decision-support” CTA. In this environment, the second may perform better than expected.
What we would do this quarter
If you buy media, we would not wait for perfect benchmarks.
We would do three things now.
First, learn the room while it is still quiet. Run small, structured tests not to chase scale, but to understand which conversation moments have commercial value.
Second, treat context-hint writing as a new media buying skill. Build an internal library of strong prompts by product, vertical, pain point and buying stage. The teams that learn how to describe intent clearly will move faster as the platform opens.
Third, prepare the post-click journey. ChatGPT Ads will expose weak landing pages quickly. If the user arrives from a specific conversation, a generic “solutions” page will feel disconnected. Build pages that continue the logic of the question.
The early opportunity is not only cheaper media. It is cheaper learning.
Every new surface has a short period when the rules are still being formed. ChatGPT Ads is in that window now.
Evido is already inside the ads manager, and we will keep publishing what we learn. If you want a partner who is in the room early — or simply a heads-up the moment your market opens — that is exactly what we are here for.
.jpg)
.jpg)
.png)
.png)
.png)
