Managing Partner Jonathan Yantz reflects on how AI will impact search.
In 2009, Microsoft launched its long-awaited Bing search engine – a decade later than the industry-leading Google. Since that day, it has sought an opportunity to make up the ground.
Now that moment may have arrived with the sudden popularity of generative AI tools. Bing has partnered with the most hyped tool – ChatGPT – and thus achieved a first-mover advantage. Google quickly followed suit to unveil its Bard Generative AI, powered by its proprietary LaMDA technology, but Microsoft had scored the first point in a long-term battle over the next evolution of search.
In this imagined future, incredibly powerful and accurate chatbots deliver exact information without needing, in most scenarios, to visit a page beyond your preferred search engine. This scenario is problematic if you’re in the business of selling products through search. It’s leading some to question whether or not search engine optimization (SEO) will be phased out by a world built by generative AI.
Those concerns are premature. First of all, neither search engine has definitively said they anticipate this technology to eventually replace the traditional search experience they’ve offered for years. But more importantly, there are several significant issues that make generative AI-only search a long way off.
Here are six reasons why…
Potential antitrust issues: Microsoft settled a landmark case in 2001 and is currently being sued by the FTC over its proposed acquisition of Activision. Google will be going to trial in September 2023 over various antitrust complaints – Google Maps was recently added to a list of complaints. Imagine if Google and Microsoft one day remove all third-party links for just one answer provided by their chat. That could be considered an example of monopolistic behavior.
Potential legal issues from publishers and content creators. AI models do not just appear out of nowhere. They are trained over millions of millions of pieces of content. For text-based ones, such as ChatGPT and Bard, that means “reading” articles from publications, web pages, and others. Visual models like OpenAI’s DALL-E, Stable Diffusion, and Midjourney are trained on millions of photos, paintings, and other multimedia.
Simply put, these tools would not be worth much without the content on which they base answers. Publishers have begun raising concerns about their content being used to train these models. Getty Images has already filed to sue Stable Diffusion. The clamor and legal objections will likely increase if the search engines change to delivering definitive answers without the user leaving the page.
The results are still too often wrong or incomplete. Ask ChatGPT whether one pound of something weighs the same as one million pounds of the same thing. Bard got a fact wrong about the James Webb Space Telescope in a launch promotional video. There’s a reason why both have released AI search as either beta products or otherwise framed as research: they’re not quite ready for prime time yet.
People will still want choice.
When traditional search is working well, it provides several different options to a user’s question, and that user gets to decide which option they access. What AI-based chatbots will eventually do well is narrow down a response to a question to one best-possible answer. Much like people sometimes go to a store to browse and other times they know exactly what they want, we all crave the ability to make that decision. As furthered in the point below, many times people will choose choice over being spoonfed an answer.
Every Internet user is different. For every question, some users will want options; others just want one good answer. Imagine a person who needs to bring a vegan dish to a potluck. If you love cooking, you may want to browse many different options and recipes to find the perfect one to test your creativity and cooking prowess. If you hate cooking and are just looking for a dish – any dish – that you can make, you just want a well-working generative AI to give you one response. A completely generative AI-oriented search engine would rob people of that choice or the ability to browse for the things that are most important to them.
Won’t do a good job of telling you what to buy. Asking generative AI questions about what to buy will lead to frustrating responses around “it depends.” Some of this is due to how the user construes the query, but, nonetheless, they do not build AI models to have opinions about things; they prefer dealing in facts. Considering 43.2% of people say researching products and brands is a primary reason they use search engines, many people would be underserved if the search engines pivoted to just generative AI results.
AI will likely have a profound impact on the future, remaking industries, and giving certain companies or industries dominant positions. However, it is still incredibly early in the evolution of these technologies, and they have only shown a glimmer of their power. They still have a long way to go and what they can do will be somewhat limited to what government regulation and societal interest say they should do. For now, there are still few disciplines more important than SEO and SEM to cost-efficiently engage potential customers.
