Chatbots make Search interesting again
They could disrupt all four of the pillars that Google's business rests on.
One of my claims to fame is that I was the product manager at Google who launched “Featured Snippets”. This is the feature in Google Search that directly answers your question using a quote from a web page. While it’s very different from a chatbot, it’s probably the most similar product that Google has launched, and so I’ve found myself getting into a lot of conversations about what chatbots1, and the large language models that power them, mean to Google.
It’s worth starting by taking a step back and looking at why Google makes as much money as it does.
Google makes money because billions of people search Google to find things online, and sometimes they click on ads. For this to work, Google needs to maintain four pillars:
Information Habits - Most of Google’s revenue comes from a tiny fraction of search queries. The reason Google puts so much effort into making other queries good is so users will form a habit of searching Google for anything they want to find online.
The Entry Point - Even if Google was the best search engine, users might use a rival if the rival’s search box was easier to access. This is the reason for Google’s substantial investments in Android, Chrome, and Google Home.
User Interaction Data - Google has a lot of very smart engineers, but the main reason Google has the best search results is because Google has the most data from observing how users interact with search results.
Dysfunctional Markets - The searches that make Google the most money are for things where the user is going to spend a lot of money but it’s hard to know which product is best - like car insurance or mesothelioma lawyers.
In the rest of this post, I’m going to address each of these in order.
Information Habits: Most Queries are Worthless
Google makes the majority of its money from a tiny fraction of queries. These are queries like “mesothelioma”, “car insurance” or “emergency plumber” where the user is telling Google that they want to spend money to solve a problem. If Google has an advertiser that sells a product that solves the problem the user has then they will pay Google lots of money to introduce them to those users.
The reason Google puts so much effort into giving good results for other queries (e.g. “why is the sky blue”) is that Google wants you to form a habit of using Google every time you want to find something online. If Google is the tool you always use to find something online then it’s the tool you’ll use when you want to find something to spend money on.
But “find things online” isn’t the only use case habit that these high value queries fit into. They are also examples of “buy something”, “help me do something”, “get advice”, “understand something” and likely other use cases. Google could potentially remain the dominant search engine while losing almost all of its revenue if another product dominated one of these other use cases. That is why Amazon is such an important competitor to Google. Amazon dominates the “buy something” use case, and “car insurance” and “mesothelioma lawyer” are things you buy.
The obvious way chatbots could threaten Google is if they allowed a competitor (e.g. Bing or Perplexity) to create a search engine that was so much better than Google that people used to ‘look things up’ instead of Google. However another risk is that chatbots allow users to form an entirely new habit that happens to include all of Google’s highest value use cases, leaving Google with just the queries that don’t make money.
Looking at my own personal use, I’ve found that I still use Google for quickly looking up facts and navigating to websites, but I’m increasingly using chatbots to discuss things I want to understand better, to get advice on something I’m doing, and to get product recommendations. If this becomes a broad trend, and those use cases include a large fraction of monetizable queries, then Google could be in trouble.
Of course, the ability of chatbots to create new things (including code, songs, slogans, summaries, ideas, and explanations) has the potential to disrupt far more than just Google - extending all professions where people are paid to understand things or write things.
The Entry Point: Chat is not Search
Google has gone to great lengths to make sure that it is easier to get to Google than to rival search engines. This is why Google built Android, Chrome, and Google Home. It’s also why they pay huge amounts of money to Apple to be the default search engine for Safari.
A user will typically use Google multiple times a day, for a few seconds each time. Because each use is so brief, a slight delay in getting to the search box can be a significant fraction of the total time spent using Google. If a competing product had a search box that was slightly easier to access then people might use it even if it was less good.
But people don’t use chatbots in the same way they use Google Search. The normal way to talk to a chatbot isn’t through a search box but using a chat app with a similar interface to WhatsApp. This reduces the value of all the work that Google has put into making their search box easily accessible.
Interactions with a chatbot are also usually longer than interactions with Google. I talk to chatbots the way I would talk to a friend or a co-worker, having back and forth conversations where the chatbot explains a concept to me, or where I ask for refinements to something the chatbot created. Since the interaction is longer, I’m more willing to tolerate that extra second needed to open the app, and having the lowest-friction entry point becomes less important. Moreover, since each chat response is shorter than a web page, I’m more keen to keep the same chat open and keep adding to it, rather than open a new tab every time I ask a question.
On the other hand, Google’s investments in voice entry points could pay off well. Prior to now, voice interfaces have been great for looking up quick facts and doing simple actions (e.g. playing music) but have been a much worse way to research a topic than reading a web page. Chatbots change that by making it possible to learn about something by having a conversation and asking questions.
User Interaction Data: Usage is Gold
The secret sauce that makes Google search so good isn’t its algorithms, but its user interaction data. Every time you do a search query, Google notices what you click on, what you scroll past, what sites you search for by name, how you refine your query, and a ton of other things that teach Google what sites you trust, what pages you like, and what your query means.
There is a positive feedback loop at work. Google has the most users, so it has the most user interaction data, so it has the highest quality search results, so it has the most users. This loop, more than anything else, has made it hard for anyone else to compete at search.
Chatbots create a whole new feedback loop. Every time I talk to a chatbot, I’m giving it more information that could help it get better in the future. Did I refine my question to phrase it a different way? Did I ask a follow up question that wouldn’t have been needed if the first response had been better? Did I ask for an improvement to something it created that wouldn’t have been needed if it had better guessed my intention? Did I call out the chatbot for getting a fact wrong? Did I give an explicit thumbs up/down that shows how well the chatbot did?
Whoever gets the most initial traction is going to have the most such data, and that could kick off another feedback loop that gives that product an insurmountable advantage. At the time of writing ChatGPT has millions of users and Google hasn’t yet launched a product.
Dysfunctional Markets: Do we need ads at all?
Google and Facebook both make money from advertising, but they do it in very different ways. Facebook makes money because you are bored, you want to read something interesting, and an advertiser has a product that they think you will appreciate based on what they know about you. Google makes money because you are looking for something specific, and an advertiser wants to sell you their version of that thing.
So why do people click the ads, rather than clicking on one of the organic search results, or doing deeper research to find out which product is good? Part of the reason is that the search results for commercial queries are usually pretty bad. Google doesn’t intentionally make their search results bad for commercial queries, but they don’t try very hard to make them good either. In the absence of good results, it’s hard for a user to know which products are good or to learn how to choose a good product, and so users often just click on one of the ads and make Google some money.
The queries that make Google the most money are ones where the user wants to spend a lot of money on a product, but it is hard for them to know which product is best. Sometimes this is because it’s in an industry (like insurance) where considerable effort has been put into making it hard to compare prices. Sometimes it’s because it’s something (like lawyers) which people find confusing. Sometimes it’s because it’s a situatation (like emergency plumbers or urgent care) where the user doesn’t have much time.
If it was easy for people to know which product was best then it wouldn’t make sense for companies to be spending so much money advertising on Google. Instead that money could be spent on reducing prices, or improving the product.
It’s possible that future chat-bots will behave similarly to Google for commercial uses - they will have poor quality answers coupled with ads for specific products. However it’s also possible that chatbots might make it significantly easier for people to know which products they should buy, both by directly recommending the best product, and also by helping users understand how to choose a good product. People sometimes joke that only dumb people click on ads. Maybe chatbots will make people less dumb.
Conclusions
Some of the predictions I made in this post will doubtless turn out to be false. However I’m pretty confident that chatbots will be a significant disruption to Google, and not just by allowing someone like Bing to make a better search engine.
Please comment with any thoughts you have. I read every comment, and will likely reply to interesting ones.
In this post I’m going to mostly use the term “chatbots” to refer to products like ChatGPT, Claude, and LaMDA that use large language models to take part in a back and forth conversation with a user. Pedants might point out that the “chatbots” have been around for ages and that what is interesting is the large language models that power this new generation. That is true, but “Large Language Models” is too much of a mouthful, the acronym LLM is too acronym-y, and I think the chat-bot interaction model is a core part of what makes these products interesting.
Man, I remember the first time I heard you explaining the concept of MashMaker 15 years ago at Intel, I was windblown, just like with this post. The clarity an succinctness in the way you present ideas and concepts is unparalleled. Besides your predictions, this is the best explanation on how Google works I've ever read.
This is a bit of an aside, but one likely outcome seems to be a nosedive in AdSense rev—both to Google and bloggers / content farms?
If you take a category like cooking or travel, being able to ping a bot with questions seems like a *much* better user experience. It can strip out just the recipe I need, and then I can ask it contextual questions about ingredients or prep methods in a messy / custom progression etc.
I wonder how this changes the economics on both sides of the table? Eg. travel bloggers will presumably see a lot fewer clicks on new content, and will thus take a dive on AdSense and referral links etc. Google would also lose rev, along with fresher content to incorporate.
The main model of content monetization has long been “win a click through search, then convert some % to subs, purchases, etc”. But if people never land on your page to begin with coz they’re just querying the content in aggregate form...