Search has become a conversation, not a query
The way that people find information online is changing.
Strides forward in LLM capabilities over the last few years have transformed the types of questions that people can ask online. Rather than having to learn to “speak search engine” and enter truncated queries into a search bar, AI can provide effective responses to far more multi-faceted queries – whether those queries are grounded in natural language, or visual & multi-modal search.
This shows up in a variety of data points. Firstly, on Google search, we’re seeing consistent growth in more open-ended queries – for instance, queries containing “what”, “how” & “best”. These are subjective questions that require complex comparison skills, which search engines of old would have struggled to address. Compare this to the more closed, “command”-style queries, such as “buy” or “cheap”, which are trending either flat or in marginal decline.
We also know that the queries people are asking are growing in length. Whilst Google are seeing growth across query strings across all lengths, nowhere is that growth greater than in the longer-tail. This is happening even more in AI environments, with users of Google’s AI Mode asking queries that are on average 2-3x longer than those on regular Google search.
All of this change has drummed up a lot of interest in the search industry, and is leading clients to ask one question time and time again: How can we ensure our brand is present within AI responses to queries that are relevant to our brand?
Until now, the answer has mainly been confined to the realm of SEO practitioners, because paid advertising opportunities have been incredibly limited. When we look across the AI answer engines from a paid advertising perspective, it’s only Microsoft’s Copilot that today has an operational advertising unit.
Perplexity has launched its “Follow Up Questions” format, but so far only in a very limited testing phase to some US brands.
Sam Altman, CEO of OpenAI, has historically taken a very anti-advertising stance to ChatGPT. Whilst his view may be softening, ads in ChatGPT are certainly still not on the immediate horizon.
And finally, in Google’s sphere, whilst we’ve known for some time now that Ads in AI Overviews have been in testing, the launch had always looked like being a fair way away. Google had been very tight-lipped around any potential timelines.
All that changed last week. At the beginning of August, Google revealed that Ads in AI Overviews could be landing in English-speaking markets in as soon as the next 5 months – ie, before the end of 2025.
It’s hard to overstate just how big a change this will represent for Google and for clients.
Why is this such a big deal?
The short answer is that this will open up the floodgates on a huge amount of previously un-monetised Google search inventory.
To understand why this is the case, and why this won’t just cannibalise existing advertising opportunities on Google, it’s helpful to build a segmentation of all of the different types of text-based queries that Google receives today.
A 2024 report by SparkToro revealed that only 15% of such queries demonstrate a clear commercial intent. A further 32% of queries are “navigational”; this encompasses brand searches, as well as more generally any queries where the user has a known destination in mind that they are looking to reach.
And then, finally, making up over half of all traffic on Google, are the “informational” queries. In these queries, users are asking a question that doesn’t – at least, on the face of it – demonstrate an obvious intent to purchase a product or service. Questions like the ones users are asking even more often within AI experiences, as discussed above.
With our segmentation in place, it’s then helpful to think about what the SERP looks like for each of the three query types. Whilst there are some exceptions, by and large it is the case that ads are bought against commercial & navigational queries, whereas AI Overviews are the most suitable response for Informational queries.
The crucial callout here is that there is very little overlap between these two result formats. When ads do arrive within AI Overviews, they will appear on queries which have largely been un-monetised – and so will genuinely present a net new, incremental opportunity for marketers to access their customers.
Of course, it would be a mistake & an exaggeration to propose that every single one of those informational queries is going to be monetised. However, in even our most conservative scenarios, we’d expect this to unlock a 20% expansion in ad inventory on Google search.
Why can Google put ads on these queries all of a sudden?
There is a crucial difference in how ads will work within AI Overviews compared to how they work today within “regular” search.
In a regular search context, by far and away the most important signal to determine whether an auction will be initiated is the search query itself. If one or more advertisers are deemed to have an ad that could be relevant for the query, then an auction will be triggered.
However, Ads in AIOs can also be triggered based on the content that is included within the gen AI response itself. Think of the AI Overview as a bridge: first of all, it will answer the query that the user asked of it, and then it will suggest “oh and by the way, if you did happen to be thinking of buying a product or service to help you out with your question, then here are some suppliers you could consider.”
It’s a simple enough idea, but it’s hard to overstate just how significant a conceptual leap this is.
What do marketers need to do about it?
Firstly, it’s worth being clear that there will be no “AI Overviews” campaign type appearing in your ad account any time soon. In keeping with their general direction of travel over the past few years, Google will be consolidating all advertising inventory in AIOs into existing campaign types.
However, that’s not to say that we can afford to stand still and have the +20% uplift land in our laps. These auctions are materially different from the ones we’ve played in before, and trying to access this new inventory with your existing search tactics will see you quickly run into some hurdles:
The good news though, is that everyone else is in the same boat. This opens up the potential for a huge first-mover advantage – but what do we need to do to capture it?
From the three challenges above, it’s not difficult to see the direction we need to move in for our Google Ads campaigns:
Targeting: Adopt intent-based targeting solutions that assess ad relevance based not just on the query, but on Google’s understanding of your business
Messaging: Lean into creative assets generated in real-time that allow you to show up in the most relevant possible way
Bidding: Provide bidding strategies with more flexibility to target never-before-seen queries
Specifically, the tactics that Brainlabs is recommending its clients to test & implement between now and the end of the year are:
Somewhat understandably, many advertisers have historically been wary about some of these features: in particular, when it comes to sharing control over assets & messaging with Google. And up until now, there hasn’t been a pressing performance need to overcome this reticence. But with Ads in AIOs just down the line, we’ll very soon reach a point where this hesitance will put the brakes on advertisers achieving the visibility and performance that they desire from AI environments.
As a practitioner, what’s most exciting about these tactics is that, with AI Max in particular, Google have learned from the feedback they received throughout the launch of PMax a few years prior. Almost right out of the gate, Google launched AI Max with an array of reporting and experimentation features that will provide clients with greater levels of transparency & influence over how the campaign is behaving and performing. A full breakdown of these features is for another post – but four particularly notable callouts would be:
1. Cookie-based A/B testing functionality for AI Max campaigns vs. regular Search campaigns;
2. An “AI Max expanded searches” row within Search Term Reports, to indicate what performance has been driven through AI Max’s targeting capabilities in addition to your existing keyword targeting;
3. Asset reports that allow for a comparison of performance between advertiser-uploaded and automatically-created assets;
4. Asset control functionalities, such as Asset Exclusions & Asset Removals. You can mentally liken these to negative keywords, but for assets rather than search terms.
Call to arms
Ads in AI Overviews will almost certainly represent the largest overnight expansion of search inventory that we have seen within the last decade.
The growth of mobile devices & Broad Match keywords are the two other events that had a comparable impact – but both instances required gradual adoption from users and advertisers, respectively. Contrast that to AI Overviews, which already reach 1.5 billion people globally. When Ads in AIOs & AI Mode do arrive, it’s hard to see Google opting for anything other than a wide-scale “turn on the tap” approach. The whitespace ahead for brands will be huge.
However, by committing to a sprint test & learn roadmap between now and the end of the year, marketers have a chance to tip the scales even further in their favour and to build distance between themselves and their competitors. The five-month countdown has begun.
English 
简体中文