Did Google Get Blindsided?

2023 / 03 / 01

Recently there's been a lot of talk about ChatGPT and its potential impact on the way we search the web. Shortly after its release, Microsoft announced that they would be adding a version of the popular "chat bot" into their Bing search engine. A common narrative at the time of writing is that Microsoft have finally found a tool that can challenge Google's dominance in the search space and that the tech industry is about to experience a major shift.

It's likely that a version of Microsoft's Bing Chat was already in development by the time ChatGPT started making headlines. It's also likely that Google already knew about Microsoft's intentions. So then why did Google's Paris presentation in February fail to impress its audience and subsequently their investors [1]? Did Google get blindsided by Microsoft's rapid productisation of ChatGPT, or is there more to this?

The Arrival of ChatGPT

OpenAI's ChatGPT is now the fastest growing web site by traffic according to Similar web's research [2]. The interest generated by the new chat bot was not lost to Microsoft, who invested billions into OpenAi to help them "responsibly advance cutting-edge AI research" [3]. OpenAI is far from Microsoft's only investment in the space though. In fact their own development teams have been successful in incorporating AI research projects into their products. A great example is Bing Maps, which uses deep learning to enhance the quality of satellite imagery and the same technology was also added to Microsoft's Edge browser to help users improve the quality of images found on the Internet [4]. There's also the introduction of a number of AI-powered tools in their Office 365 suite, aimed mainly at increasing accessibility [5].

I'm highlighting this because it seems to me that people are unaware of the massive investments that Microsoft have been doing in the AI space. They have a history of successfully translating research projects into products. With that in mind, it's not surprising that a version of the successful ChatGPT (and GPT 3.5) has also made its way to one of their existing products, Bing Search, in the form of Bing Chat [6].

Google's Product Development Approach

Google on the other hand never struck me as a company which approaches new product development in a reactive way which is influenced by external market forces. They always seemed to experiment with new ideas and while few of them didn't survive [7], most have gone on to become foundational pieces of technology that we all use everyday. Billions of people use Android OS on their mobile phones [8], millions of websites are built using Angular [9] and millions of applications are hosted on Google Cloud. There's no doubt, however, that when it comes successful products, few can rival Google Search. Google Search is ubiquitous and is part of our vocabulary.

Google themselves are also investing heavily into AI research, with their own take of the large language model in laMDA. In fact their own Google research paper was at the heart of the rapid improvement in the large language generative models, including the one powering ChatGPT [10].

With all of that said, it's difficult to ignore the fact that one of the tech giants delivered a working product, while the other said that they're working on a product. The Paris presentation showed that Google was not ready to respond to Microsoft's threat, with their response to Bing Chat, only getting 15 minutes of fame. Is it lack of a cohesive product vision from Google which led to this, or is there more to it?

The Problem of Scale

The cost of a Google search query is very likely lower than a ChatGPT query. Looking at OpenAI's language model pricing, I won't be surprised if on average a single chat session costs them a few dollars [11]. Given the volume of daily searches on Google, if the cost wasn't lower than ChatGPT, then they'd go bankrupt in days. Although cost is a key consideration when building new products, there are ways to limit them, by staging the rollouts, simplifying the language model, or even limiting the response length, just to name a few. In Google's case the reason we haven't seen a product similar to Bing Chat earlier is likely scale, but not in the context of "cost". Rather it's likely the impact that such a product may have on the way users interact with Google Search.

Changes in the Search algorithm can have a huge impact on how companies rank in organic searches [12]. There are people and businesses who specialise in SEO (Search Engine Optimisation) and their only goal is to push a page up in search rankings. There's a reason why people would rather write a blog on an already established platform (like Medium.com) than create their own blog site and that's because the latter likely won't ever find an audience due to poor search rankings.

A chat bot which provides users with answers to their questions by pulling data from different web sites is great as a time saver, but it doesn't really drive traffic to said sites. Less traffic, means lower ranking in search results and in turn lower revenue.

A product like Bing Chat in Google's ecosystem can have a massive impact on it. A chat bot available at Google's scale which, for example, can recommend your next favourite book may have a noticeable impact on traffic to sites that specialise in book recommendations, such as Goodreads.com.

Microsoft's Bing doesn't really suffer from the same problem, which according to independent estimates accounts for only about 9% of global searches, with Google handling more than 80% of the global search queries [13]. Roughly speaking, Bing's annual search volume is about the same as Google's in a month. In this highly skewed scenario, Microsoft has the opportunity to introduce a highly disruptive product to its ecosystem in an effort to increase its user base, even if it means that the cost for supporting that user base is higher.

Bard

This finally leads us to Bard, Google's response to Bing Chat, which was announced during the aforementioned Paris presentation [14]. While the impact of Bard is still unknown (at time of writing it is still in closed testing), it is clear that Google is recognising the threat in the search space. Without that threat, however, would Google still consider adding Bard to Search? Absolutely, yet probably not in this format.

These days we mostly search for answers, not for specific websites or services [15], so it's not a stretch to say that having the ability to ask any question and get straight to the answer is going to add a lot of value to the end user experience. Yet from a purely practical point of view, such a product can also be too disruptive and can quickly cannibalise ad revenue from Search, which remains Google's biggest revenue stream, without too much upside. When you command more than 80% of the market, can you really justify capturing an additional 10% percent of it, by increasing your Search costs by 100% percent, for example? Probably not, but looking at it from the opposite side, can you justify retaining 10% of the market, by increasing your costs by 100%? This is probably harder question to answer, but ultimately an existential one if it's about the survival of your flagship product. Seeing the massive interest generated by ChatGPT likely forced Google to recognise that the way we search online is changing and the best course of action would be to accept the risks, higher costs and lower margins in order to protect market share.

Conclusion

In this battle for the search space, Google has a lot to lose and Microsoft much to gain. Google may have preferred to have launched Bard on their terms, in a format which doesn't directly compete with their main revenue generator. The small slot allocated to Bard during the Paris presentation highlights that they had to include a response to Microsoft's Bing Chat, but were not confident enough in their product strategy for it. They had to acknowledge the threat that Microsoft's Bing Chat represents to their core product and in turn decided to respond in a swift and decisive way. Overall, I don't think that Google got blindsided in any way, but it's very likely that this time they had to approach the development of a new product in a more reactive way.