I decided to weigh in with my own take, based on nearly 20 years of experience related to on-page SEO and content creation as well as almost 30 years of simply being online (not all of it all of it terminally, as it was rather hard to be online all the time in the dial-up era). Depending on how you prefer to consume information, you can watch me make my case in the video below or scroll past it and get straight to the reading.
I have some bad new for Google bears: the search and ad giant is not going anywhere. As we've already seen so far with tests/experiments carried out by Google as well as Microsoft, the search engines are looking to use AI more as a sort of enhanced structured snippet generator rather than the end-point of a search query.
This makes sense since most searches will not have one single answer and many users will want the optionality of multiple results, even if the top ranking page still gets the most traffic.
One innovation I see being possible, if not likely, is the implementation of an AI-powered summarization feature for lower-ranking search results. Imagine if instead of clicking into a link, you could click a "summarize with AI" button next to it and get a quick synopsis of the contents that helped you decide if it was worth clicking or not. Think of this as an "enhanced meta description."
Another possibility is that you get additional information about the website or company behind the link, helping you decide whether or not that search result is credible.
The execution could be a bit tricky, but this kind of functionality would certainly add significant value.
As slightly touched on above, the impact of AI on search behavior will vary by query type. In the case of "one true answer" searches (e.g. "who developed the theory of relativity"), the AI box will likely reduce search clicks substantially. This will hurt traffic to Wikipedia and similar reference sites.
At the same time, product or service searches can't be replaced with a simple summary. Searches like "best basketball shoes" or "best B2B CRM" will always need to provide multiple potential results since the definition of "best" will vary by each person's evaluation criteria and the person making the search is looking for options, not a single answer.
Of course, AI could provide a list of recommendations based on a top-ranking result, but that's effectively the same as a search result page – there's no difference between an AI-generated top 10 list and the top 10 results except packaging since search results are also AI generated, in a way.
The death of SEO has been prophesied for as long as I can remember, but as long as there's a need for content to be indexed by computers, there will be a need for best practices that help structure content in a way that makes it easier for computers to index. AI doesn't change any of that since its still a computer system that indexes information – its just a smarter one that those we're used to. And the only way to provide indexable information in a way that's consumable by humans is to create content.
What will likely change, however, is how people approach content distribution. With AI taking over for long-tail queries, particularly question ones, that SEOs have long been fans of, other distribution channels will be necessary to get the necessary visibility. This sort of platform risk has existed for a long time, but many marketers have consistently ignored it to their own peril - and there are plenty examples of platforms rugging publishers that got overly dependent on them.