Back in February, I wrote about ChatGPT ads just as they were beginning to surface. At the time, a lot of it was educated guesswork – reading between the lines, interpreting early signals, and trying to anticipate how this new channel might evolve.

Three months later, things look very different.

This isn’t theoretical anymore. Ads are live, money is flowing, and – importantly – we’re starting to see how this actually works in practice.

So here’s an updated view: what’s changed, what we got right, what we didn’t, and what marketers should be paying attention to now.

This is no longer a test – it’s a real, working ad channel

In February, ChatGPT ads felt like an experiment, but now they’re alive and kicking.

OpenAI has been running a live pilot (primarily in the US), and early reports suggest:

• Significant advertiser demand
• Rapid revenue generation
• Strong interest from both brands and smaller businesses

So – not very surprisingly – they are proving they can attract budget.

What this means:

If you’re in marketing, this has officially moved from “keep an eye on it” to “start thinking about how this fits into your channel mix.”

The format is more conservative than expected

One of the biggest surprises is what ChatGPT ads aren’t.

Back in February, there was a lot of speculation around:

• Ads embedded directly into answers
• Sponsored recommendations woven into responses
• Fully “native” conversational advertising.

So far, that hasn’t happened.

Instead, ads are:

• Clearly separated from the main response
• Placed below the answer
• Labelled as sponsored content

It’s a much more cautious rollout than many predicted.

Why that matters:

OpenAI is prioritising trust over short-term monetisation. That’s a big signal about how this platform will evolve – they’re going to do it slowly, and with guardrails.

They’re being careful about trust

If there’s one theme that’s become clearer over the last three months, it’s that trust is important to this whole thing.

OpenAI has been explicit about a few things:

• Ads do not influence responses (some disagree…)
• Conversations are not sold to advertisers
• Users have control over what they see

This isn’t just PR positioning – well it might be, but they’ve got to practice what they preach on this one, as it’s foundational to whether this model works at all.

Because the moment users feel like answers are being “bought,” the product (the “trust”) breaks.

Early performance signals are strong

We’re still early, but initial signals suggest:

• Users aren’t aggressively dismissing ads
• Engagement is holding up
• Relevance improves with context

That’s notable, because the expectation was that ads in an AI assistant might feel intrusive.

Instead, when done well, they appear to feel more like:

• Timely suggestions
• Useful next steps
• Contextual options

That’s what’ll help this platform sink or swim – at the moment it isn’t interruption-based advertising. It’s proximity-to-intent.

They’ve launched a self-serve platform

I had an email, I think it was last week, to say OpenAi have now launched a self-serve platform (in the US) for their ads platform. Which is massive because you used to need to know a guy, and get actual OpenAi to update your ads!

Assuming they let anyone use this platform now – and you can sign up here to get your name on the list of people who want to use it – it means suddenly there’s a much lower barrier to entry, you can do faster experimentation – and there’s a lot more competition.

Expansion beyond the US has begun

Initially, this was very much a US-only story but now we’re seeing early signs of international rollout.

For UK marketers, we’re not in this next wave, but hopegully the one after that so we’ve got a short window to prepare before this becomes widely accessible.

The big open question: influence vs integrity

There’s an emerging tension here around subtle bias in recommendations, and sponsored influence vs neutral answers. OpenAi are adamant it doesn’t happen, but people aren’t yet convinced.

With Google Adwords we’ve never been worried that someone having an ad at the top of search results influenced the rest of the serps below it – but that kind of effect does seem possible now in ChatGPT.

Time will tell, and people will be looking for it!