AI Ads are selling. Are you buying?

It has become a cliché to say that AI is changing everything. Most of the time, it isn’t. But ad placements inside AI responses may be one area where something genuinely shifts.

At first glance, AI ad placements look like search all over again: a user asks a question, and brands appear alongside the answer. The similarity is real—but incomplete.

AI placements sit earlier in the decision process, shaping consideration before a user has committed to a search or a click. That creates opportunity—but only if the approach changes. Simply extending a paid search strategy into AI environments is unlikely to deliver the same results.

Perplexity was the first major AI search engine to test paid placements in 2024. Despite early interest, with CPMs exceeding $50, revenue growth stalled. They have since taken a cautious approach, citing user trust as a key concern.

But where Perplexity led others soon followed.  

Microsoft Copilot’s advertising has launched in several languages, integrated into Bing inventory via existing Microsoft Ads campaigns.

Google AI Overviews began testing in 2024 and now appear alongside around a quarter of all AI Overview responses, up from around 5% in early 2025. Shopping ads with direct offers are also live inside AI Mode. 

In Feb 2026, OpenAI launched its US ad pilot on ChatGPT, targeting free-tier users.

Each of the major AI players then, with the notable exception of Anthropic has developed an ad offering. As might be expected, each is slightly different. 

Microsoft Copilot, perhaps the most mature option, benefits from familiar infrastructure, CPC bidding, and measurable performance through Microsoft Ads reporting. However its reliance on existing Microsoft delivery infrastructure means it is not yet widely embedded in high-frequency customer journeys. This combined with its skew towards desktop means it has struggled to embed itself among habitual users. 

Google AI Overviews is not so constrained. Being embedded in the world’s most popular search engine makes these placements unavoidable. Management and measurement sit inside Google’s own widely used ads infrastructure. Thus far Google has viewed its AI overviews as merely an extension of its search and PMax campaigns, rather than a distinct placement, sacrificing clear attribution and measurement for set up convenience.

OpenAI with ChatGPT, lacking access to an established delivery network, has been slower off the mark but has big plans and the reach to meet them. Until now: OpenAI’s March 2026 update which dramatically increased the available inventory showed no impact on consumer trust metrics and low ad dismissal rates. A self-serve ads manager just launched, removing the $200,000 minimum entry bar.

Is this for you – and if so, how do you show up?

Not every brand should be testing AI ad placements right now. The channel is real but it’s still forming, and the cost of getting it wrong isn’t low as the attribution and the learnings aren’t there just yet.

Before you commit, ask yourself:

Is your category search-dependent? AI ad placements are closest in behaviour to search. If paid search is already a meaningful part of your mix, AI placements are a natural extension. If you’re primarily social or display-led, the case is harder to make right now.

Do you have enough experimental budget? Copilot aside, most AI ad placements are not yet held to the same performance standards as mature channels. If every line of your budget needs to justify itself against a CPA target, this isn’t the right moment. If you have 5-10% allocated to emerging channel testing, this is exactly what that budget exists for.

If you have two yeses and an audience that’s already asking AI questions – That’s your green light.

Let’s make sure you’re set up to measure success.

Measuring AI ad placements with the same framework you’d apply to search or social will give you the wrong answer. Last-click attribution was already not the most accurate form of measurement, add conversational AI to it and it’ll become completely useless.

LLMs excel as research tools. It is at this stage of the journey that the user is most likely to utilise an AI Chatbot, rather than at the bottom of the funnel where paid search is likely to remain supreme. Although ChatGPT and Copilot’s interface may feel reminiscent of a search engine, its role in marketing is closer to that of a comparison site. The role of ads is therefore to enhance visibility, rather than directly convert.

If last click can mislead, what can you measure instead? 

Assisted influence – look for lifts in branded search, direct traffic and organic conversions in the period following AI placement activity. Not perfect, but helpful.

Share of AI citations – are you appearing in organic AI responses for relevant queries? Paid placements and organic citations are increasingly connected. Brands that show up organically are better positioned when paid inventory opens up.

Incrementality – where possible – test. Split your audience and measure the delta. It’s the only clean way to isolate what AI placements are actually contributing.

For mobile-first advertisers, MMPs like AppsFlyer are increasingly relevant here. AppsFlyer’s probabilistic matching and view-through attribution can capture conversions that last-click misses. If you’re running app campaigns alongside AI placements, your MMP setup should be configured to account for assisted and view-through touchpoints, not just last-click.

One development worth flagging: with OpenAI’s new self-serve ads manager, measurement setup becomes a must rather than optional. Get your attribution framework in place before the inventory opens up.

The main principle is – set your measurement framework before you spend. This matters for any channel, but it’s non-negotiable here.

AI ad placements are not a channel that will stabilise and sit still.

The next 12-24 months will reshape what’s available, what’s measurable, and what share of budget it deserves.

A few directions worth watching:

The inventory will open up. ChatGPT’s self-serve manager lowers the entry bar. Other platforms will likely follow. What’s currently a managed, medium to high value buy becomes a self-serve channel accessible to mid-market and growth-stage advertisers. The question shifts from “can we access this?” to “how much should we allocate?”

Organic and paid will converge. AI answer layers are already blurring the line between SEO and paid search. As paid placements scale, brands that have invested in appearing organically in AI responses, through content and AEO, will have a head start. The two strategies will need to be planned together, not in separate workstreams.

Attribution will get better – but not quickly. The measurement gap won’t close overnight. MMPs and incrementality testing will improve, but we’re still miles away from any industry standard for AI placement attribution.

The creative frontier Brands testing ChatGPT and Copilot placements right now are finding that the formats that work on search and social don’t translate. The ad that performs in a conversational interface looks different, reads different, and works differently. 

Watch this space – we’re still designing it.

If your media plan looks the same as it did 18 months ago – let’s talk about where AI placements fit, what to test first, and how to measure it properly.