AI & the future of search
In my previous article, I talked about content discovery & AI in the video streaming space and I will continue that discussion for other forms of audio and video entertainment in my upcoming articles. Meanwhile, I am back thinking about search and the potential of AI to disrupt search across consumer and enterprise products the way we know today. In this article, I will make a few predictions on how this market will evolve. I will cover popular consumer search products like Google, Amazon and LinkedIn, convergence of search and AI chatbots in new services like Perplexity, and the emergence of new enterprise search platforms like Glean.
Consumer habits will take time to change and the leading search engines are here to stay for a while
Google continues to be a leading global search engine today, (except in China). From a query like “capital of italy” to “ny weather” to “sofa coupons”, people use Google for anything and everything including their information, shopping, travel, and research needs. Similarly, search engines like Amazon search and LinkedIn search are used for searching for products, prices, people, companies and jobs. These search engines are strong brands, have a ton of user data and are integrated well with the internet ecosystem. Google, specially, due to its ownership of Chrome, Android and productivity tools, has leverage in making its search engine a default part of all its products. Consumers are habituated to using these engines and they are not going to change habits and move to new search platforms easily. This means, even though some user groups and certain types of user tasks, might move to newer AI platforms, the existing leading search engines will keep a lot of their traffic for a while.
Research, analysis, and idea and content generation tasks will move more and more to AI chatbots
Even though simple searches for people, places, products, jobs, etc. will likely stay on existing search platforms, tasks like research, analysis, comparison and deep expertise are already moving fast to AI chatbots like ChatGPT and Perplexity. Whether you are a developer learning to code, or a entrepreneur planning a new strategy, or a user trying to make sense of your bills, legal agreements or health reports, or a shopper comparing multiple products, or a college graduate preparing for job interviews, people are using AI chatbots for all their research, anlaysis and brainstorming needs. AI chatbots act as perfect companions for such tasks and allow for deeper, more nuanced conversations. In the past, a lot of this research would happen on Google through querying, re-querying and web navigation, but AI chatbots are much better equipped at this than a traditional “search, click and re-research” interface that Google, Amazon, LinkedIn and the likes provide today.
Note, however, that Google and Amazon already offer chatbots like Gemini and Rufus, and Google has also introduced AI answers on the top of its search page for relevant queries (e.g “compare top us cities in terms of quality of life”). AI chatbots can also analyze an image or video and provide summaries, detect objects or answer questions, and find answers from a uploaded document or report, which was not possible with traditional search engines. I expect more specialized and domain-specific and enterprise chatbots to emerge in the next few years that cater to doctors, nurses, lawyers, accountants and more, and that specialize in specific types of work or tasks like education, sports, medicine, etc.
AI chatbots will improve on quality and personalization, and strengthen trust with users
Even though more complex tasks like research, summarization, coding, analysis, brainstorming and idea or content generation have moved to chatbots, services like ChatGPT, Gemini and Perplexity hallucinate sometimes and do not provide accurate information or answers. For example, these models are known to be bad at math and cannot do simple addition exercises. The models are also known to generate a different answer every time they are asked the same question, making them more suitable for creative idea and content generation tasks than factual questions like “how far is the sun from earth”. AI chatbots have also not been around for as long as the popular search engines have been and will need more time to build trust. People still trust Google for being factual and accurate almost every time, and chatbots will need to both improve their quality (and fix issues like math and counting) to strengthen trust with users.
A great search engine offers relevant and personalized content while accounting for quality, popularity (vs deep cuts), diversity, seasonality, trends, and cultural/geo context. For example, searching for gifts during Xmas vs summer should lead to different search results (or products), and searching for snacks in west Africa should show different results than in northern Europe (while respecting personal preferences) and Google and Amazon do a good job at this (though they can be better). AI chatbots, however, are known to bias towards popular content and not account for diversity and deep cuts enough, though prompting techniques could improve that. Personalization based on geo, language and individual preferences is another important area of work, which leading search engines have learnt to do well. In my experience of using Large Language Models (LLMs) to generate personalized playlists based on text-based user prompts, it has not been easy to personalize the response playlists using prompting techniques. This article does a good job at defining opportunities at the intersection of personalization and LLMs, and this paper talks about personalizing LLM responses for a specific user.
Agents and new interfaces and AI devices will emerge
As people interact more with conversational assistants, new devices and form factors will emerge that will help them do this anytime, anywhere. For example, a voice-enabled assistant hanging from your neck or AI airpods in your ears auto-connected to ChatGPT can help you navigate the world around you, ask questions, engage in conversations and even do work on the go. With the first agents potentially coming out in 2025, these new interfaces powered by agents could create a huge productivity boost as everyone will have an assistant available to them all the time, who can complete tasks for them on the go. Specially, for people who are disabled or physically challenged, this would be a leveling platform enabling them to accomplish tasks faster than they can today.
AGI is coming soon
As AI models improve, researchers are OpenAI and Anthropic are predicting that models that as smart as PhD. students are only a few years away. The o3 model is the closest to that that OpenAI offers. The recent software and model innovation by DeepSeek is showing that we might much less compute power and energy to deliver AGI style models. This would mean that AI agents and models might be able to autonomously do most tasks that a white collar professional does today including interacting with each other and working in an ambiguous complex environment. We do not fully understand the consequences of this on the labor market, economies and communities, but it will not be small. Companies, countries and governments will have to re-think and re-organize themselves to operate effectively in this new world and keep humans happy and productive.
As I think about the future and the human need for information, productivity, and taking care of their self and their communities, I am excited about what AI has to offer. Search engines, just like other systems, will go through their own transformation in this world.
