AI-driven search platforms are rapidly evolving and are poised to disrupt advertising strategies that fail to adapt to the changing internet landscape.

Marketers now have to consider the inevitable ad placement opportunities on these AI platforms, their tendency to favor different content patterns when surfacing web results and the need to monitor existing SEO traffic in light of their surging influence.

For the past year, conversational search has been a prominent use case of AI thanks to platforms such as Perplexity and Google’s growing implementation of its web summaries feature, AI Overviews. These offerings are meant to provide real-time information to users, such as news and stock prices, in a way that feels more human-like and personalized than the transactional process that searching the internet has always been. 

Over the past two weeks, the AI search space has turned a corner. OpenAI integrated its uber-popular ChatGPT within search engine capabilities, with plans to bring AI-powered search to the platform’s 250 million-plus active monthly users over the coming months. And Meta was reported to be developing its own AI search engine, potentially bringing millions of more users into the fold. These major entrants, with their massive audiences, have turned AI search into a phenomenon that advertisers can no longer afford to ignore. 

Of particular interest are the ways that search advertising opportunities may expand in tandem with the growth of the space, said Jason Hartley, head of media innovation and trust at agency PMG.

Each AI search platform will incorporate offerings for advertisers in some capacity, Hartley said. These will inevitably have their differences based on the platforms’ unique search engine results page (SERP), as well as the niche ways that consumers use the platforms. For example, Perplexity and ChatGPT have mostly been geared toward information discovery, rather than behavior that constitutes a readiness to purchase. This means that advertisers showing up in these platforms may need to deploy different kinds of messaging, Hartley said, such as prioritizing brand awareness or affinity over concrete promotions and items for sale.

Google remains a destination for purchases, and the ad opportunities it has already rolled out inside AI Overviews reflect this. Brands can deploy shopping ads, labeled as “sponsored,” within relevant summaries—one example, previously demonstrated by Google, shows how a user inquiring about how to get a grass stain out of jeans might see a series of links to purchase stain removers via websites like Walmart and Albertsons. 

Despite AI platforms showing impressive usage numbers, Google is still ways ahead of the competition, accounting for nearly 90% of the U.S. search engine market share as of last month, according to StatCounter. Rather than viewing the influx of new players as a harbinger for the demise of their Google-dominated search ads strategy, advertisers should simply expect the overall pie to enlarge, Hartley said. This is because AI search could give users more and differentiated reasons for using search engines. 

“Search is now an integrated marketing strategy, whereas it used to be just a channel,” Hartley said.

Also read: Google and post-cookie ad tech trends for marketers

Perplexity, for instance, is best suited for conducting thorough research for a particular topic, including surfacing material that would likely be buried as irrelevant by Google, said Tobias Zwingmann, managing partner for AI consulting firm Rapyd.AI. As such, Perplexity plans to help advertisers with metrics of success less traditional for search, such as driving awareness over conversions. This will include the ability to sponsor a “related questions” section that appears in a user’s inquiry.

“We’ll launch with a few select brands across different categories to experiment with formats and gain initial user feedback,” a Perplexity spokesperson told Ad Age.

OpenAI has yet to announce advertising offerings for ChatGPT’s search capabilities and did not respond to a request for comment.

Exploiting AI’s data preferences

Not everyone is sold on the promise of AI-powered search. Rapyd.AI’s Zwingmann, who has tested ChatGPT’s new search capabilities, thinks that the feature is too niche to succeed with mass audiences, especially when search giant Google is already as reliable as it is. That AI platforms prioritize criteria other than reliability when surfacing search results means that consumers are likely to be burned, which could ultimately erode trust in the technology, Zwingmann said. 

But this very issue is one that savvy advertisers can use to their advantage should AI search continue to gain popularity. According to Zwingmann, when AI models are scraping the internet to return results for a search, what they’re looking for is more about content density than content authority. That is, content that appears with high frequency and semantic overlap (or consistency of meaning) are more likely to be surfaced, even if their sources are less reliable. ChatGPT’s search capabilities, for example, are vastly underperforming Google’s with respect to short internet queries, in which meaning and intent is vague, according to TechCrunch.

This means that one way brands can increase their chances of being surfaced by these platforms is by simply increasing their presence on the internet—or, “playing the numbers game,” Zwingmann said. Providing more information on one’s website, for example, as well as posting more on social media could have an impact on how the AI sees that brand, and its likelihood to mention the brand in search results. 

The SEO-related incentives for this tinkering is quite clear. In a basic test conducted by Ad Age and Greg Asman, founder and managing principal of marketing consultancy firm The Asman Group, a search query for “Dove” on ChatGPT first surfaced results for the bird, while on Google the same query first surfaced results for the Unilever-owned brand. And even when ChatGPT did surface results for the brand, only a link to its Wikipedia page was included and not a link to its actual website. Dove declined to comment.

Providing more information about one’s brand may not only improve visibility when AI performs real-time data searches, but may also do so with respect to its underlying training data foundation, Zwingmann said. AI systems are always being trained on more data in order to generate better responses, but even the size of the internet is no match for their appetites. According to a study published this summer by the Data Provenance Institute, the amount of potential training data available to these systems is drying up, in part because publishers and other online entities are restricting their data from being used. Some platforms, such as Perplexity, have been accused of continuing to hoover data without consent.

In this data-scarce environment, brands could find advantage by showing up on the web in a larger way. Another path to do this is by increasing one’s advertising. While traditional formats, such as display ads, are ignored as training data, brands can apply workarounds by publishing more sponsored content, which AI systems will gladly train on, Zwingmann said. 

This approach may feel dubious for web advertising, but it is not without precedent. Zwingmann compared it to the common tactic of placing sponsored ads alongside similarly looking non-sponsored material, such as in Google search results, Instagram feeds and Amazon listings. 

With AI search, “the lines [between advertising and non-advertising] are getting blurred more and more,” Zwingmann said.

In the meantime

While the impact of AI search is already being felt, advertisers still have non-AI search strategies that need attention. Monitoring these strategies in light of the mounting influence of AI search platforms is something each brand should be doing right now, said Brian Yamada, chief innovation officer at VML.

One area of particular importance is bot traffic. In the past, brands have focused on minimizing the negative impact of bots, such as skewing audience data across various channels. But now, bots in the form of AI search platforms are a positive presence, seeking to flag information for consumers that may very well draw them to the brand. This means that brands need to keep an eye on the ways that bots are showing up on their website and other properties, in order to discern helpful ones from malicious ones, Yamada said.

The technology required for such identification purposes is still incipient, but it will improve along tasks such as determining source and intent, Yamada said. Team One, a Publicis-owned ad agency, recently launched a version of this kind of tool called Faikcheck that identifies AI-generated images. 

Understanding that users will be both human and non-human, and the non-human cohort both beneficial and unbeneficial, brands may want to reconsider the barriers they have in place for entering their online properties. For example, if a brand does not want bots under any circumstances to be included in their audience traffic metrics, they could require logins for entry or some other form of registration that would stop AI in its tracks.

“Every company will need to make decisions about what they’re allowing from a search standpoint,” Yamada said.

While a new variety of SERPs are already emerging, brands should keep in mind the kinds of disruptive search experiences that could ultimately come down the pike. One such technology is AI agents, which would effectively do a user’s bidding on the web for them, Yamada said. The idea is that a user would have a set of personalized agents, each programmed to autonomously carry out a specific task, such as book flights. This model of using the internet would look far different than the way it is currently used.

ChatGPT and Perplexity do not currently offer agent-like mechanisms, but their AI-powered search capabilities can be seen as beginning to shift the way that people use the web toward a less demanding, more conversational dynamic.

“At a minimum, we’re starting to see how the search experience will improve,” Yamada said.