Bing vs. Bard — How Long Will the Adoption Curve Be for AI Search?
Depending on which you think matters more — accuracy or intimacy?
Last week, Microsoft kicked off an arms race to integrate generative AI technology into search. At a media event last Tuesday, Microsoft proudly unveiled a new Bing search interface, powered by upgraded AI capabilities enabled by ChatGPT. At this stage of initial rollout, the upgraded Bing search is available only to a select number of testers and accessible only via Microsoft’s Edge browser, although Microsoft promised that expanded access is in the works. Users with access will be able to chat with Bing as they do ChatGPT, asking questions and receiving answers in natural language. Unlike the answers that ChatGPT gives, Bing will cite its sources and link to them to ensure the integrity of its answers.
This announcement, of course, sent Google into action to defend its dominance in search. The day before Microsoft’s event took place, Google announced its own ChatGPT rival codenamed Bard, which the company called “an experimental conversational Al service powered by LaMDA (Google’s own large language model),” along with its plan to integrate it into Google search over the coming months. The rules of open market competition -unavoidably push Google to respond once Microsoft makes the first move, and as a result, the dawn of AI search is now upon us.
(It is worth noting that this AI search arms race is not limited to the U.S. — Baidu, China’s №1 search engine, has also announced its own AI initiative codenamed “Ernie Bot” last week that will enable users to get search results in a conversational Q&A manner. For the purpose of this article, however, we will focus on user adoption outside China, although the same arguments, with the exception of the competitive landscape, would apply to Baidu as well.)
Considering that search ads took up 13.7% of global digital advertising spend in 2022, per IAB data (second only to digital video at 19.3%), it is important for future-forward brands and marketers to wonder what this all means for the future of SEO and paid search ads. Considering that voice search via digital assistants has been around for years, yet no one has quite figured out how to put ads in the voice search results, it seems safe to assume that advertising would not be easy to pull off in this new AI-assisted, conversational search era either.
Naturally, for brand marketers, the next question is how long the adoption curve will be — do we have enough time left to figure it out before AI search hits mainstream and upends how brands use paid search channels to drive lead generation and conversion? Well, let’s take a look at some of the key factors determining the length of the adoption curve.
User-friendliness UI & Learning to Write Prompts
If the experience of online search were to transition from clicking through a list of blue links and browsing different sites for answers, to simply asking an all-powerful intelligent assistant to crawl through the vast information online and answer your questions in a concise, accurate manner, it’d seem logical that most users would likely choose the latter experience of less effort and friction. Chatbots have been around for nearly a decade at this point, and every mobile user has interacted with one at some point by now. Therefore, texting a virtual assistant for answers would almost seem an upgrade in terms of user-friendliness.
That said, in order to get the results you’re looking for, especially when it comes to more complex questions with nuances, there would likely be an additional learning curve for users to understand how to properly prompt the AI to look for the correct answer. Simply typing in a few loosely associated keywords off the top of your mind and hoping Google’s SEO engine will take care of the rest is no longer going to cut it. In other words, in order to get the right answers, users will first need to learn to ask the right questions.
In order to get the right answers, users will first need to learn to ask the right questions.
While that may sound easy to do, not everyone is great at that. And asking the right question of AI may mean something different than asking the right question of another human.. With in-person conversations, there is a lot of interpersonal and/or contextual information that is already known for both parties. (For example, if you’re my friend, when I mention the movies of Steve McQueen, you’d know I meant the British director who is one of my favorite working filmmakers, and not the late American actor that shares the same name.) In contrast, the search-assisting AI typically does not have the similar set of baked-in information to help it narrow down the scope of a search, which would require further clarifications that would add friction to the process — unless, of course, data about your locations, habits, and preferences is already fed into the search AI as well to provide more personalized answers.
Accessibilty & Shifting Competitive Landscape
Another key factor in determining the adoption curve would be how widely available and accessible the AI search tools will be in the near term. For now, Google promises to roll out Bard “within the coming weeks” to a wider audience (there’s currently a waiting list that you can join), and Microsoft certainly seems to be striking while the metaphorical iron is hot.
As a long-time also-run in the search engine space, Microsoft’s Bing (with a meager 9% global market share as of December) is undoubtedly the upstart challenger here. OpenAI, a startup in which Microsoft is a major investor, claimed that ChatGPT had reached 100 million monthly active users in January, just two months after its public launch, which sets a new record for the fastest-growing consumer application in history. With little to lose and a lot to gain, it makes perfect sense for the Seattle-based company to capitalize on the strong buzz that ChatGPT has been garnering in recent months, and come into this space with guns ablaze, ready to shake things up in search.
With little to lose and a lot to gain, it makes perfect sense for Microsoft to come in with guns ablaze, ready to shake things up in search.
Compared to Bing, Google’s approach to integrating its AI chatbot into search has been much more conservative and cautious. As a market leader in both search and advanced AI research, Google faces a lot of (deserved) regulatory scrutiny and market expectations. It is understandable that, even though the speculations about Microsoft potentially integrating ChatGPT into Bing have been swirling around for weeks, Google has been playing it relatively safe when it comes to launching a rival product.
But that alone won’t be all that much — a lot of online searches today now start on social media platforms or ecommerce sites like Amazon. For iOS users, the default search engine in the Safari browser is also responsible for the majority of the mobile searches. Therefore, for AI search to take off, companies like Meta, Amazon, and Apple (aka, the other three Big Five tech giants) will also have a role to play in this competitive landscape.
Things get even more complicated when you consider the rivalry between Microsoft, Google, and Amazon in their cloud services, which running a mainstream AI search engine would require quite a lot of. OpenAI’s popular generative AI tools are already being integrated into Microsoft’s Azure cloud service, which presents a pressing threat for Google’s cloud business in the long run. Meanwhile, Amazon could theoretically team up with Microsoft to integrate ChatGPT into its main search bar on Amazon.com, and compete against Google in an “enemy of my enemy is my friend” kind of mentality, But the direct competition between Amazon’s AWS and Microsoft’s Azure Cloud Service would certainly mean that would not be viable.
OpenAI’s popular generative AI tools are already being integrated into Microsoft’s Azure cloud service, which presents a pressing threat for Google’s cloud business.
Apple will also have a big decision to make in terms of which AI search engine it would end up integrating into its Safari browser as the default choice for hundreds of millions of iOS users. Google currently pays top dollar to the iPhone maker every year for this honor, but there’s a non-zero chance that Apple would end up going with the revamped Bing, if only to differentiate itself from the Android devices, into which Google is sure to integrate Bard.
Besides Apple, other browser integrations will help move the needle in adoption. Bard will surely spawn some Chrome extensions while Microsoft and OpenAI will try to integrate Bing/ChatGPT into more third-party browsers too. For example, Opera already announced it is integrating ChatGPT into its sidebar for a “‘shorten’ feature that provides bulleted summaries of the article or webpage you’re reading.
AI Search Accuracy & Trust Issues
Another big hurdle in adopting AI search tools is skepticism over the accuracy of the search results generated by AI. Cases of AI hallucinations have been well-documented, which raises the important question of how much users can trust the kind of AI-packaged answers to their search inquiries in general. Unfortunately, neither Microsoft nor Google is really helping in their respective rollout of the AI search tools so far.
Google’s team included an erroneous answer in its announcement video for Bard, which immediately drew criticism and cast doubt over the entire initiative, which, in turn, shaved $100 billion off Google parent company Alphabet’s share price last week. Meanwhile, early testers of the revamped Bing also found a series of factual errors in its answers — it even made financial errors during Microsoft’s first demos — not to mention some straight-up creepy or disturbing answers that some testers have coaxed it into replying. (More on this later.)
Obviously, this is just the beginning of a new era for search, and there will be a long road ahead for user education and fine-tuning the guidelines safeguarding the answers. For what it’s worth, Microsoft has published a blog post on Thursday in response to some of the concerns around the new Bing search, in which it claims that 71% of users gave AI answers a thumbs up, while acknowledging that Bing answers are not “necessarily helpful” or in the right tone after 15 or so consecutive questions. Reading this post, it is clear that Microsoft is positioning the new Bing as a continual learning experience that it hopes to mature into a product that could challenge Google and upend its de-facto monopoly in search.
Microsoft is positioning the new Bing as a continual learning experience.
Simulated Connections Over Accuracy? Sometimes, but Not Always
On another hand, perhaps getting too hung up on the accuracy of the AI search results is not the right way to gauge the adoption curve either. When we ask other humans questions, the answers we get wouldn’t be 100% correct, even when we are asking professional analysts or subject experts. And it is not like we are getting the correct answer all the time when we currently search for information online either. Granted, one’d hope the search results we get from an advanced chatbot would be factually accurate — our world already has so many problems caused or exacerbated by misinformation and fake news that we certainly don’t need more coming from Bing or Bard.
Yet, as analyst Ben Thompson pointed out in his viral blog post on his conversations with Bing’s conversational mode (aka Sydney), applying AI in search may be a distraction compared to what generative AI tools actually can do — convey (simulated) emotions and leverage that to forge a sense of intimate connection with users. After all, to err is human, and being capable of making mistakes would help humanize the AI. Theoretically, this capacity to make an emotional connection with users might just help some users to look past some occasionally inaccurate answers and be happy with a friendly research assistant.
Applying AI in search may be a distraction compared to what generative AI tools actually can do — convey (simulated) emotions and leverage that to forge a sense of intimate connection with users.
As alarmingly fascinating as it is to see some of Sydney’s answers give the appearance of being a sentient AI, researchers say we are still nowhere near crossing that Rubicon technically. Yet, in this digital age, perception is everything. (“If you can’t tell the difference, does it matter?”) And we humans are very good at anthropomorphizing inanimate — or in this case, insentient — objects. If Microsoft were to push forward with this supercharged Bing, then perhaps the audience attention it would end up garnering would go far beyond the volume of attention that a search engine would typically get.
So when it comes down to it, it seems clear that the adoption curve of AI search would work differently for different types of search — the ones that seek facts, news, and scientific information would likely hesitate to adopt AI search as their primary method of search (at least not without meticulous citations and double-checking), whereas people that search for opinions, recommendations, and creative inspirations may benefit a lot from engaging with an AI-powered Q&A-style search engine instead.
These two types of search have long been grouped into one behavior in the Google era, even though they are fundamentally serving two different “jobs to be done.” With the arrival of AI search, it may finally be time for our search engines to discern and unbundle these two types of search, and answer users in the format best suited for the goal.
In other words, the “ten blue links” are dead, long live the “ten blue links”!
