WARC caught up with Michael Levinson, Vice President and General Manager at Google Ads, during Cannes Lions to discuss how generative AI may change the search advertising experience.

What are the focus areas for Google in AI as an enabler of better advertising? Are those focus areas guided by innovation in technology, or more what brands are asking for?

AI has been at the heart of Google Ads for years: we called it machine learning and it was about finding the right message for a user at the right time at scale. What is new technologically are the generative AI models. So much of it is about helping companies get started. How do they set up campaigns to reach the right people? Instead of manually creating dozens of images, can they create hundreds of thousands of different versions that are going to work with different audiences? This just wasn't possible before from a scale perspective.

But, to your question about the technology and clients, it's always going to be a little bit of both – it can be a trap just to let marketers have a hammer, as they’ll go looking for nails. But the fact is that the technology has changed so much around gen AI. Everyone is looking for ways to use it and we’re doing the same.

On the consumer side, where is Google up to in terms of figuring out where gen AI from an advertising standpoint fits in that search experience?

We announced at Google Marketing Live that we were going to start experimenting with ads in the AI Overviews experience. We haven't done that yet. It’s something that we are really focused on getting right, and we don't think we have the answer yet for exactly what that's going to look like. We do think that there's going to be a lot of opportunities to do new and creative things.

AI Overviews is a new way for people to engage with search. People are entering longer queries, more descriptive queries, and are having a more conversational experience with Google. They're getting a broader answer, not just some links. And, in that broader answer, it’s very easy to imagine the commercial message appearing alongside it. But we're going to go slow, and we're going to focus on responsibility. And I think we'll get it right over time.

What does ‘getting it right’ mean? Or, to flip that, what does ‘getting it wrong’ look like?

Well, we always have to start with the user. We want Google Search to be useful for people. If it's not useful for people, they won’t use it. Because the mode of engagement is so new, we're learning how people are going to use it, and people are learning how to use it. So be useful for users first, and if we do something on the outside that gets in the way of it being useful, that's not going to help anybody.

I’m sure brand safety is a big consideration when you're dealing with generative AI. What are Google’s guiding principles here?

With AI, it feels more like Google's giving you an answer, so we have to be really careful that we're giving people the right answer. When an ad comes in there, it's going to feel a little bit more like Google speaking on someone's behalf, so we need to be really careful with that.

When advertisers talk about brand safety, people's minds go to image generation. That's something where responsibility and safety is built in from the very beginning. We don't want people to use our tools in inappropriate ways and we also don't want to surprise advertisers. We're asking them to trust us, that we're going to use a bunch of these tools to make your assets perform better. Part of that is we're not going to surprise you with an image – it’s in your control to use the generated images or not, to pick which ones you'd like. Direct us in the interface. We don't want to accidentally have the system do something that an advertiser is not going to be comfortable with.

How can brands manage that at scale, when the AI can create hundreds of thousands of assets really quickly?

Those things are going to evolve. The first-gen AI products have been smaller scale: give us a prompt, and we'll create a dozen assets you can look at. If you don't like it, you can tweak the prompts. The next iteration is more logistical; it’s less sexy. Marketers can say, ‘Here's a couple of assets we really like, generate a whole bunch of different aspect ratios for me.’ That is not super risky. Being able to give you every possible aspect ratio, every possible permutation – that's pretty much all upside for an advertiser.

When we start getting into whether this person is more of a dog person versus a cat person, and let's create a version for that, that's a different thing. We're not there yet. But that's where, to your point, we're going to have to iterate our models for how people can review and set guardrails.

In the short term, at least, are generative AI tools going to be more useful for brands in the lower-funnel, and from that conversion perspective?

Creative is really important for performance. Especially these days, with the way people use the internet, scrolling really fast. When you show them an ad, it either arrests their attention or doesn't. Obviously, at Google, we're very performance-focused as an ads and technology company. Absolutely, this is going to be critical for performance.

How is the nature of commerce and the purchase journey in Google products changing, and how is the ad experience changing in relation to that?

Historically, it was really expensive for companies to create beautiful product imagery. It was a secret sauce for retailers and companies with more resources. With AI, even little mom-and-pop companies can create gorgeous imagery of their products and compete with any other company for people's attention. The quality of presentation in our shopping services and through Product Focus ads just gets better and better.

Do you get the sense that discovery is evolving, and that the distinction between upper funnel on YouTube and conversion through Google Search is changing?

On YouTube, you can show a person a product that they've never heard of before and they can just buy it right there. The whole funnel is collapsed. The more engaging your content is, the more likely you are to make that happen. It's not, ‘Let's teach you about our company and our brand, and then our category and then a specific SKU.’ It's, ‘Hey, how about this?’ You've never even thought about that category before, but you click it and you buy it. Is that lower funnel, or mid-funnel? It’s selling products to people.

What are marketers saying to you about the current situation with third-party cookie deprecation?

You know what, we've been talking about this for a long time. The Chrome team determined that the time wasn't right. It is still absolutely our plan to follow through with deprecating third-party cookies, but there are a lot of considerations to get it right. We will get there but it's complicated.

Do you get a sense that it's something that brands are now fully prepared for?

I've had conversations this week on both sides of that. We've had some companies say, ‘Come on, just do it already. We're ready. We spent all this money. We're ready.’ And we have other companies saying, ‘Oh, man, it's really going to change a lot for us.’ You can't please everyone, but it's clearly the direction of travel. We’ve got to get it right.

What are your predictions for how we’ll be discussing AI in advertising in 2025?

I think by Cannes next year we will have made a lot of progress towards the things that we're just starting to talk about now. What is the ads experience is going to be in the AI Overviews part of search? I also think we'll have made a lot of progress on YouTube as a performance tool. And I think we will have a lot more use cases for gen AI where it's not just, ‘We have this technology, let's just see what sticks.’ There's a lot of experimentation that we will do between now and then. I suspect a year from now it'll be much clearer.

Click here to read Creative Impact Unpacked: 11 effectiveness trends from Cannes Lions 2024