Aussie AI

AI Market Research: Search

  • Last Updated: 6 November, 2024
  • by David Spuler, Ph.D.

Research Papers on Generative AI and Search

Old Version about Search

[This was written in August, 2023] One of the biggest stories since the explosion of ChatGPT in Feburary 2023 has been the threat of upending Google's near monopology on one vertical: Search. What does generative AI ("genAI") mean for this vertical? This articles examines the threats and opportunities in this space for Google, Microsoft, and other less visible players in the search market.

Overall, the threat in the short-term to Google from Microsoft's use of ChatGPT seems overstated because it has numerous defensive barriers in the search market, and has itself the opportunity to go on "offense" with its own generative AI called Bard. However, the long game will take years to play out.

Google's Defensive Barriers

Google is a behemoth and has erected numerous competitive barries over the years.

  • Brand mindshare. The Google brand has such a deep mindshare in consumers that the phrase "to google it" is said without a thought to trademarks. Indeed, when Microsoft's Bing launched, one of the stated reasons for the name was that they hoped a short name would get verbized as "to bing it". But that hasn't really happened, and it seems likely that many people use Bing without even knowing its brand name (e.g. Microsoft Edge browser users who use the default search).
  • Chrome browser. The Chrome browser has been a masterful strategy move by Google, and gradually grew market share at the expense of Microsoft's Internet Explorer (IE) browser, mainly through better technical execution in terms of compatibility, performance, and stability. Indeed, Microsoft's newer Edge browser is now based on the same underlying engine as Chrome.
  • Apple iPhone search. Google reportedly pays a licensing fee to Apple in order to be the default search engine on iPhones. It's reportedly in the billions, although it is obviously confidential. One of the interesting things in the trade press when ChatGPT emerged, was that Apple was reportedly re-considering this deal in favor of Microsoft, Bing, and ChatGPT. It's unclear how true this was, since negotiations would clearly be closely held. Although this would seem to be a financial deal that Microsoft could outbid, it's important to note that consumers are comfortable with having Google Search on their iPhones, and Bing still may need to close some functionality and technology gaps before it can keep all of Apple's customers fully happy.
  • Android Phone O/S (and Samsung). Theoretically, non-Google search engines and browsers could penetrate this space, but it seems that Google has the advantage. The comments about an Apple iPhone deal probably also apply to a Samsung phone deal for Google Search.
  • Google has AI too. With all the hype about ChatGPT, it's easy to forget that Google has some of the most advanced AI tech on the planet, too. They have Bard and Gemini, and it was actually Google who invented the Transformer and open-sourced it in 2017 (don't you wonder if they maybe regret that?). Although Google has been playing "defense" against OpenAI's ChatGPT and Microsoft's Bing AI, Google will almost certainly quickly transition to go on "offense" with AI features incorporated in numerous areas of search and its other products.
  • Cash. In the unlikely event that Google Bard struggled with traction, Alphabet has the funds to go unicorn shopping. A more likely scenario might be that both Alphabet and Microsoft (and other behemoths) bid against each other for several of the major AI startups, as they begin to offer complementary functionality that the big foundation models can use to expand their AI offerings in new directions. Similarly, there will be competitive M&A interest in startups in the infrastructure segments (below AI models) and the application-level segements (above AI models).

Some Thoughts on Search Impacts of GenAI

AI Content Farms: There has already been an explosion of startups with new websites based on SEO content generated by GenAI. No doubt, there will be a boom in mediocre long-form content written by ChatGPT and other AI engines on the web. But there have been "content farms" for years, starting with Demand Media's growth in the early 2000's, based on thousands of "non-artificial intelligence engines" (i.e. human writers). Hence, Google has been fighting this battle against content spam for over a decade, so it seems likely to have this under control.

Pageviews-per-User: One of the less obvious risks to Google and a variety of other Internet players is that AI-based search could reduce the time-to-answer for consumers. Fewer searches per person means Google can show fewer ads, and also means fewer clicks through to websites (and their ads). People are well-known to visit dozens of websites via Google during extensive research tasks, such as those related to travel or health. The overall number of pages viewed by consumers in researching such a task could markedly decline if the answers were easier to find from a single query, nicely summarized by ChatGPT or Bard. On the other hand, if the search engine itself is better thanks to AI, more of those pageviews might stay on the Google and Microsoft sites, rather than distribute out to other sites on the web, so the two main players might benefit rather than lose out. Certainly, Google and Bing have both been doing various things over the years to increase the stickiness of answers to increase engagement directly with the search engines, and their secondary derivative site pages (e.g. shopping, maps), mostly at the expense of other third-party websites (e.g. travel websites, medical websites). The AI engines might exacerbate that trend for smaller websites. There might be less SEO traffic to go around.

AI Engagement: That users will do fewer searches remains an assumption, and it might be false. Since users love their AI answers, they might ultimately do more searches. It could be that complex searches lead to reduced queries (e.g. in travel or health), but that users do more simple queries. Or they might increase the range of activities for which they consult with Google or Bing (or ChatGPT directly). It also seems that Google and Bing have the opportunity to personalize searches in much more intelligent ways, toward the goal of a "personal AI assistant", thereby increasing user engagement overall. This all still remains to be seen.

Voice Interfaces: One of the keys to the future use of AI queries will be via voice interfaces to "virtual assistants". This is continually growing, and the advent of much smarter AI answers might herald a boom in this type of usage. The usefulness of that system is going to get better and better.

Biased AI training content: There's been a lot of talk about the "bias" of AI results, which is an issue in itself, but let's talk about the extreme: people intentionally tricking AI to produce biased results. Recently, ChatGPT started showing up in the survey results on how people find products (also, TikTok). Imagine, if you work as a marketer at Acme Computers. Firstly, you want ChatGPT to include your brand name when people ask: "What is the best PC?" There's already been some articles on how to get your product noticed by ChatGPT, which involves inserting your content into the training path. There's a whole industry around SEO trying to game the search engines, and the same will happen to AI. There'll be a whole industry around whatever it'll be called, is it "AIEO"? Whatever the name, AI model owners will have to deal with marketers intentionally trying to get their model trained with advertiser's very biased content.

AI Monetization early days: The user interfaces by which people see the results of AI on search engines are still rudimentary. And with changes in the displays, will come changes in the way that ads are shown for monetization. This remains one of the important ways that both Google and Bing may mitigate against the risk of users doing fewer searches per user.

Bias and Monetization: There are risks of intentional bias with AI monetization, too. Maintaining the integrity of the AI bot's answers will be important. Google has always maintained that it does so for its search results; i.e. you can't pay for SEO traffic, and that advertiser payments don't affect the display of non-ad results. The same concerns arise with AI-generated results, and the ads that will be near them, or inserted into their text.

Conclusion

Prediction: Alphabet, the company formerly known as Google, is well-positioned to defend its search market share in the short-term (years) and also seems sure to close the gap in AI technology in the long term (decades). Microsoft's Bing might make some short-term gains, but they'll be low single-digit percentages even if they occur, and Google will maintain its massive dominance of the Search market. And both Alphabet and Microsoft have even bigger opportunities to compete in the AI B2B market, which is another whole story.

Prediction: In terms of M&A activity to defend or attack in the search market, it's hard to see how the unaffiliated foundation model unicorns make good targets for Microsoft or Alphabet, since they directly compete with OpenAI or Bard. However, these unicorns may gain traction in certain use cases or segments, making them attractive, and they will also be attractive to several other big tech companies that lack their own major GenAI foundation model products (e.g. Databricks already acquired MosaicML for $1.3B). More activity in M&A may arise in the variety of AI-related startups as they gain business customers or a large consumer audience. This hasn't really started fully yet, but expect M&A deals at substantial valuations in the infrastructure segments (e.g. AI-specific GPU hosting startups, business AI training platforms, AI consultants and integrators, etc.) and in the application levels at the top of the AI tech stack, such as various B2C consumer use cases offering higher-level features by leveraging AI models. Even at unicorn valuations, these purchases will be considered as only "tuck-in" acquisitions for the big players, to be used against each other in competition for the motherlode.