Google is the front page of the Internet. The Internet is a central hub for knowledge, information, and economic activity. And Google is its biggest gatekeeper. It is where a lot of my online life starts.
I am a chronic googler. I search anything and everything that pops into my hyperactive mind. My search history is a deeply personal record of my weirdest ideas, questions, fears, and interests. May it forever remain private and never be exposed.
Google search has deteriorated. Results are riddled with SEO arbitrageurs, ads, and affiliates. It pushes users to the highest bidders. It prioritizes its own results over organic links and prioritizes SEO optimized sites over quality content. The quality and relevancy is declining. I am not the only one complaining.
And the results are often boring. Uninspiring.
Roughly half of my search activity is “subjective.” Meaning the queries are not for facts or specific results. “Subjective searches” don’t have cut and dry answers. They are Ideas. Opinions. Recommendations. Advice. Conversations. “Subjective search” encompasses everything from restaurant or recipe recs, to commentary on politics or new technology, to fitness tips and product reviews.
The best subjective search content often lives on forums, long-tail blogs, social media, and review sites. It doesn’t live in SEO optimized websites or big media publications. It is not readily surfaced on Google.
My “subjective search” activity has been steadily shifting to Reddit, Twitter, HackerNews, TikTok, and Pinterest. And most recently…. AI chatbots like Character.ai! GenZ is shifting off Google too. Some estimate ~40% of their search activity is now on platforms like TikTok and Instagram.
Back to my newfound penchant for AI chatbots.
No, I haven’t lost my mind. I know that the AI chatbots are generating content that may be unreliable. But it is interesting, interactive, and thought provoking. It is fun. I can essentially generate an engaging Reddit debate with my AI. Or have my AI chatbots recommend books, coffee shops nearby, etc. It surfaces recommendations in an interactive and personalized way. If I don’t like the first recommendation, I tell it why, it refines, and pops out a new one.
When Apple first released Siri, I wrote about it as a new era for how we access information. One that is more personal and interactive. Alexa and Siri took AI chatbots mainstream for basic q&a, digital assistance, and interaction. A decade later, thanks to natural language understanding, LLMs, vector indexes, and zero/few shot learning, the potential for truly personal, interactive, information retrieval systems is closer than ever.
The next Google won’t look like Google
This is a trillion dollar opportunity. Google will face innovator’s dilemma as its business model hinges on displaying a page of links littered with keyword auctions and PPC ads.
The next Google will look nothing like Google. While there is a big opportunity for companies like DuckDuckGo or Neeva that primarily compete on a single vector (e.g. privacy) but otherwise replicate the Google experience, I am excited about those creating a new, more engaging, personal, and contextualized search experience. This may look like a silly AI friend. Or a generative recommendation engine for a niche vertical. Or a conversational advice-giver. It probably starts in “subjective search” domains where Google is weakest. And in categories where accuracy is less critical so the stakes are lower, like fashion, food, and entertainment.
Here is a list of ideas and characteristics for a better and different information retrieval system:
- Ambient and in workflow I can conjure up my personal search bot via browser extension to generate information without leaving my workflow or opening a new browser, or like Alexa, has ambient presence and lives intelligently as a layer on top of daily activities.
- Generates summaries & dialog/answers, not a list of links with the ability to cite the original source of the information and supporting evidence. No need to pause to allow cookies or be blocked by paywalls. Intelligently deep-link and route users into specific apps when appropriate (e.g. deep link to Waze to show me the traffic getting to work).
- Come with a question, stay for entertainment users come in search of something specific, but stay for fun and interactive conversations, entertainment, games.
- Customizable user control empowers users to control their search experience by personalizing the algorithm, results, and UI (e.g. never show me recipes from Martha Stewart, surface more Reddit threads). Downvote and upvote results.
- Social see what my friends or people I follow are searching and interested in. Let me curate cool information, sites, or content I find and share it publicly, creating an ongoing taste board. I love the idea of a search product that is inherently viral, thrives on sharing and messaging.
- Context aware & personalized like TikTok or Pinterest, results are highly personalized and the system has a history of context about the user to better understand intent, Occasionally curates and surfaces interesting info that I didn’t even know I wanted to discover. Suggests and completes tasks like a personal AI assistant (e.g. send a calendar invite, order shampoo on Amazon).
- Conversational and visual interfaces that can move across modalities (e.g. speech to text, vice versa) and works with natural language, not keywords.
- Vertical focus creates a 10x better experience for a specific use case, like Yelp or Foursquare did for local, or Indeed for Jobs.
Consider this a Request For Startup.
If this resonates and you are working on something in this space or have feedback, please reach out!
Talia I’m absolutely floored at how consistent our vision at Lexii.ai is with what you laid out here. We’ve been at it for eight months already, and I’m just now reading this. I reached out to you privately on LI and Twitter — would love to hear from you.
–Andy, founder at Lexii.ai