Is Google Getting Worse? (Update)

Hey there, it’s Stephen Dubner. About a year-and-half ago, we published an episode called “Is Google Getting Worse?” We talked about Google’s near-monopoly over internet search, and we looked at a startup called Neeva that was trying to compete against Google. Well, Neeva has already shut down. So maybe it will be impossible to dislodge Google. Or maybe not. Since we put out that episode, a new kind of competitor has emerged: search engines powered by artificial intelligence. You’ve probably heard about ChatGPT, but the one I was interested in hearing more about is called Perplexity. So today, we’re going to replay that earlier Google episode — but first, I’d like you to hear a new conversation we just recorded, with this would-be rival.

Aravind SRINIVAS: Google created the greatest business model in the internet. It just has to be rethought. 

That is Aravind Srinivas.

SRINIVAS: I’m the co-founder and C.E.O. of Perplexity.

Stephen DUBNER: Perplexity is what? Describe it to someone who — like, a grandmother, an uncle, someone who doesn’t know a whole lot about how this works. 

SRINIVAS: Perplexity is a conversational answer engine. So you ask it a question, and it just gives you the answers, and it also tells you where it pulled the relevant information for the answer from, the sources. So, it’s saving you all the time that you otherwise would have spent browsing the internet and opening these different tabs and reading all of them for yourself and being pretty confused about what you should take away.

DUBNER: Now, if I didn’t know that you were the C.E.O. of Perplexity, I might think you were the C.E.O. of ChatGPT, because what you’re describing sounds the way that ChatGPT works to a large degree. Are you simply a later, slightly more up-to-date version than ChatGPT? Or are you substantially different?

SRINIVAS: When we came out, the primary difference was the fact that ChatGPT hallucinated and gave incorrect information for queries that actually needed factual grounding. And we could do it because we always pull up relevant facts from the internet and only use those facts to answer your query. So, accuracy is about two things. One is like up-to-date information. The second thing is the speed. When ChatGPT browses, it almost takes, like, an eternity to get the answer back. Six or seven seconds, to be more precise. But in the paradigms of computer interaction and user experience, closer to 10 seconds for a response is pretty bad.

DUBNER: What do you think are the odds that Perplexity and ChatGPT or OpenAI, and maybe five other A.I.-driven search or answer engines, as you call them — what do you think are the odds that all of those collectively will essentially put Google out of business within 10 years?

SRINIVAS: I don’t think Google is going out of business. All the infrastructure they’ve built for the traditional ranking-based search interface is still going to be valuable, even for the next generation of interfaces that are going to be more A.I. chatbots-like. All I can say for sure is that this new sector of A.I. chatbots and answer bots and potentially agents too, will not be a monopoly.

DUBNER: And why do you say that? 

SRINIVAS: Because it already is not. First of all, leave Perplexity. Maybe you can be skeptical that Perplexity can make it that far, to be considered a Google competitor. But already OpenAI is not letting Google be a monopoly there. Meta won’t let Google be a monopoly. Microsoft was there. So, there are just going to be more players in this ecosystem of answer bots and chatbots and agents and more devices in the market that power these. And Apple is also not going to be quiet here. So, Google could have been the monopoly here. For what it’s worth, they could have built ChatGPT in 2020 or ‘21 and had the whole profits from this new sector largely driven for themselves. But they were just too lethargic, and other people had the first-mover advantage. 

DUBNER: But, with a few billion people used to having search for free, how do you approach the notion of training people to pay for something that they used to get for free?

SRINIVAS: We’re still always going to have a free version of the product. They would pay for a better service. And once people do realize that they have been cheated all along where all their time has been taken away and all their data is being marketed to the advertiser who’s making the money, they’re going to realize that this is worth at least a few lattes a month, you know.

DUBNER: Let’s pretend that three years from now, Perplexity is acquired by a much bigger firm. What’s most likely, that that firm is Google or the rest of the field — the rest of the field we’ll say: Microsoft, Apple, Meta, etc. 

SRINIVAS: I hope not Google because, you know, the whole point is the world needs alternatives, right? If you just want one company to forever be the de facto source of information and knowledge on the internet, then that doesn’t seem like a good future. 

DUBNER: Do you see a company out there that you think, “If we were to be acquired, that would be a fantastic fit?”

SRINIVAS: Firstly, I just want to caveat by saying I am interested in being that company myself — like, Perplexity to be one of the providers of information on its own and building a sustainable business. But since you asked a hypothetical question, I think it would be very interesting to work with Apple. Because Apple has such an amazing stronghold in terms of access to American users who are what advertisers are really trying to target. And, the ability to have native integration across the Apple ecosystem of devices. Siri will finally be able to ground itself in real information and answer questions. And that’s one company Google fears, right?

DUBNER: What about privacy — how Apple treats customer data versus Google and Meta and others?

SRINIVAS: Even there, yeah. We have been very transparent about, you know, if you want to turn off your data usage, just go to the settings and turn it off. We’re not going to use that information. Which is actually way better implemented than ChatGPT. In ChatGPT, if you turn it off, you lose your history too. Whereas in Perplexity you can still retain your history, and we would not use the data for training. In fact, there was a recent incident where Google by default indexed all people’s Bard chats without telling them. Whereas in Perplexity by default, every query is private unless you choose to share it only then it becomes publicly indexable.

That, again, was Aravind Srinivas, the C.E.O. of Perplexity. He tells us they now have about 10 million active monthly users, with 70- to 100,000 of them paying $20 a month for the premium service. ChatGPT is much, much bigger, and Google recently introduced its own A.I. model, called Gemini — so, this is only getting started. Okay, now it’s time to hear the original episode, called “Is Google Getting Worse?” We’ve updated facts and figures where necessary. And we’ve also cut out the section about Neeva since they’re already dead.

*      *      *

Do you remember when using Google to search for something online felt like magic?

Ryan McDEVITT: I’m old enough to remember before Google, and it was really hard to find anything on the Internet. 

That is Ryan McDevitt; he’s an economist at Duke.

McDEVITT: Google was a revelation because it made this information accessible, and it was so useful.

The power of that revelation faded, as revelations do, and we all began to take Google for granted. When you needed some information, you just typed a few words into the search box and, very quickly, you got the answer you were looking for, usually from an authoritative source. But today? To me, at least, it doesn’t feel the same. My search results just don’t seem as useful. I feel like I’m seeing more ads, more links that might as well be ads, more links to spammy web pages. Do you also feel like Google isn’t what it used to be? Today on Freakonomics Radio: how did Google come to dominate web search in the first place?

Marissa MAYER: People would be like, “Your search engine is so good and you’re not making any money, and we just wanted to pay you.”

What happened when Google essentially became a monopoly?

Jeremy STOPPELMAN: The problem with monopoly power is that you can degrade the experience because you’ve locked in that user. 

And we will hear Google’s defense — or is it more of a threat?

Liz REID: If I took Google away, you would go nuts.

Is Google getting worse? Or is it just that — as Milton Friedman might have said — there ain’t no such thing as a free search?

*      *      *

In the beginning, computer scientists created the world wide web, and the web was without form, and void; darkness was upon the face of the Internet. And someone said, “let there be search,” and behold, there was search. And, in time, the search engine.

MAYER: There was AltaVista, there was one called AlltheWeb or Fast that was working out of Norway. There was one called Dogpile that was actually a metasearch and hit many of the other searches.

That’s Marissa Mayer. She runs a tech startup called Sunshine. Before that, she was C.E.O. of Yahoo! And before that, she worked at Google, for 13 years. She was the 20^th employee at Google, and their first female engineer; she joined right after getting her master’s degree from Stanford in computer science.

MAYER: I was very attached to a search engine called Infoseek. I just felt like it really helped me on my papers. And so when I even started working at Google, for about the first month, I would do every search on both. It took me a while to convince myself I wasn’t going to miss out on anything by not doing the search on Infoseek. 

Infoseek and other search engines were okay at crawling the web and finding pages that contained the words you searched for. But they weren’t great at figuring out which of those pages you’d actually want to see. A search query produced a long list of sites, in seemingly random order, and you would have to page through hundreds or thousands of links to find what you needed. Sergey Brin and Larry Page thought they could do better. In 1996 — when they were grad students at Stanford — they created a search engine built upon a clever algorithm called PageRank. As in Larry Page but also as in ranking your search results by page. Theoretically, the most useful result would be at the top of the first page of results. Think about a newspaper, if you can remember back that far: the important stories of the day were on the front page, A1; the less-important stuff was tucked inside. That’s what PageRank was going for. And how did it do that? Brin and Page had applied an insight from the world of academic research, where the most important research papers are the ones that are most often cited in the bibliographies of other papers. The web didn’t have bibliographic citations, but it did have links. So when PageRank looked at a given web page, it would see how many other pages linked to it. The most important pages, they theorized, would have lots of links pointing to them from other pages — and then the algorithm looked at how important those pages were, based on incoming links. Their mission, they said, was “to organize the world’s information and make it universally accessible and useful.”

MAYER: Even when I was deciding to go to Google, the refrain I heard most often from people who knew I was thinking about working there was, “Why does the world need another search engine? There’s already a dozen or so that are good enough.” And I think it was Larry and Sergey rejecting that premise, that good enough isn’t good enough for search. You can be great. And the one that’s great will make a marked difference, is really what led to Google’s success.

Brin and Page were not the only researchers who thought about ranking search results this way, but they did get their version to market first. And people really liked it.

MAYER: Every now and then, we would get fan mail that would have like a check or cash in it. People would be like, “Your search engine is so good and you’re not making any money, and we just wanted to pay you.”

So how would they make money? In 1999, some big venture capital firms invested $25 million in Google, in a deal that valued the company at $75 million. They considered or attempted a variety of money-making strategies: licensing their search technology to other websites; selling a hardware product that would let companies search their own internal files. If you had read Brin and Page’s original research paper about their search engine, you might have thought there was one source of revenue they wouldn’t pursue: advertising. In a section of their paper called “Advertising and Mixed Motives,” they wrote “advertising income often provides an incentive to provide poor quality search results.” So which revenue model did they pursue? Yes, it was the ads.

Tim HWANG: One of the things that Google really innovated early on and was really successful in was getting ads to scale on search engines. 

That is Tim Hwang, who used to work at Google on technology policy. Now he is a media researcher and the author of a book about digital advertising called Subprime Attention Crisis.

HWANG: And what that basically meant is that Google as a search engine became a lot more profitable than a lot of its competitors.

What made them so profitable? As it turns out, the mechanics of the Google ad-tech system were just as innovative as their search engine.

HWANG: The way the Google ad-tech system works is that there is essentially a marketplace where people are bidding for the right to show me the search ad. 

McDEVITT: Google was one of the first to auction off the first few listings on the search results. 

Ryan McDevitt again:

McDEVITT: Their position auctions, they call them. That’s how Google made all their billions of dollars.

HWANG: Now, one of the innovations that Google has is that that auction is based not only on how much money you’re offering, but also on a number of characteristics around the quality of the ad. And in some ways, this is kind of Google’s genius, that they were able to set up a market where people bid for the right to show ads, but then there’s also an incentive for those ads to be ones that are not automatically annoying ads. 

So when you click the “search” button, Google finds relevant ads and chooses which ones to display above the search results, based on several factors: how much the advertiser bid, how relevant Google thinks the ad will be to the user’s search, as well as what Google calls “context” and “expected impact.” Google says that 80 percent of searches result in no ads being shown along with the results. But how important is that 20 percent to Google’s bottom line?

HWANG: It’s huge. About 81 percent of Google’s revenue comes from advertising on its properties and in Search, which is really remarkable if you think about it, right? It still remains their core driver of revenue, despite all the other things Google has done in that time. The kinds of things that Google is known for, like artificial intelligence and self-driving cars and all this kind of stuff — they all pale in comparison to the continued dominance of ads as the thing that funds the company. 

“The company” is actually called Alphabet these days; that’s the conglomerate that includes Google and a bunch of other properties, including YouTube. But most people still call it “Google.” And considering the financial strength of Google itself, this makes sense. But mixing search and advertising — as Google founders Brin and Page had themselves argued — could create “mixed motives.” Think about what happens when you do a Google search. You’re looking for the answer to a question. The advertiser, meanwhile, wants the user to see information that promotes the advertiser’s interests. Those two motives might overlap — but they very well might not. Marissa Mayer says that in its early days, Google did consider a subscription model versus an ad-sales model.

MAYER: As we started to do a sort of thought experiment around how to monetize, we were like, “Should we just have people pay us to run this search service?” You know, we looked at it, we were like, okay, $20 a year. That seems like a reasonable price point for that time. And then we were like, “Theoretically, what if we could make a penny a search?” And by the way, our search ads are sold by the thousands, they use mille in Roman numerals, C.P.M.s. 

C.P.M. meaning “cost per thousand.”

MAYER: So we were like a $10 C.P.M., a penny per page. And we were like, “Well, we think people are going to start doing more search, not less. And as people start doing searches, let’s assume they do 20 searches a day, five days a week, they take the weekends off. That’s 100 searches a week. And there’s 52 weeks in a year. So that’s, you know, 5,200 searches, which means if we could do a $10 C.P.M., we could make $52.” So then you’re like, well, wait. That means with ads, we could probably make $52, where consumers are really putting a value on this that’s closer to $20, right? So it’s two and a half times as lucrative to do it as ads, as monetizing the consumer directly. And by the way, in Q4, in the retail quarter, in the run-up to Christmas, the C.P.M. that Google sees in North America today, I would guess is well north of $100 C.P.M. So it’s not a penny per page. It’s a dime per page. So then you’re saying, well, now it’s the difference between, like, the user might pay us $20 or maybe because of inflation and time, maybe now they pay us $100 a year. But Google can make $500 a year on those same searchers with that same set of assumptions.

DUBNER: So in the end, it was not a hard decision.

MAYER: It wasn’t a hard decision.

But Mayer and others at Google were concerned that showing ads with search results would degrade the experience for users. So she helped design an experiment to see how true that might be.

MAYER: We said, “Okay, 99 percent of users will get ads and we’re going to hold 1 percent not getting ads,” just to make sure that if we start to see, in various search-health metrics, a degradation in how the searchers were feeling and using the search engine, we could pick up on it

This was back in 2000. The experiment was put in place — 99 percent of Google users getting ads, and a small group of users getting no ads at all. But then Google sort of forgot about the experiment.

MAYER: I think it was 2008, one of my colleagues came to me and said, “We have this issue, we have advertisers who call our customer-support line and they’re very upset because they pay for ads and then they get assigned into this experiment group called “No Ads at All,” and they can’t see their ads and verify that they’re running.” And I remember feeling — like I turned kind of white at my desk as I heard this, because I said, “Do you mean that experiment that I coded in 2000 is still running?” And they were like, “Apparently.” But I was like, “You know, before we turn that off, this is the longest-running split A/B experiment in the history of the internet. And it is on this fundamental question about our company and our business. So before we turn it off, let’s analyze it.

And what did the analysis show?

MAYER: What we saw was 3 percent more searches from people who had ads than didn’t. So basically, there was an appreciable difference over a long period of time that people actually liked Google search results more and did more searches when they had ads than when they didn’t, which I thought was really validating. And then we turned the experiment off. 

When you consider that users who got ads did more searching, you realize this was probably the most expensive experiment Google ever ran. Think of all that lost ad revenue! But there’s a bigger question: why would ads improve the search experience?

MAYER: Because you often get a higher quality result experience from someone who’s willing to pay for it. I will say now as an entrepreneur, there’s a point when you’re building an app that you’re like, “I’m willing to put an ad out for our app. I think it’s that good.” Up until then, when you’re in beta, you’re like, “You know, we’re still ironing out some kinks. I’m not sure it’s worth to pay to put our app in front of someone.” And that said, there are some queries that just have a remarkably commercial intent. At the time, it was like, Madonna tour tickets. If you did “Madonna tour tickets” and you looked at the organic results, they were terrible. There’s no tickets to buy. But if you put search ads there, the people who actually have tickets to Madonna’s tour are happy to pay a lot to get those expensive tickets in front of users. And they actually make the search quality better. 

That’s an interesting argument, and I don’t disbelieve the data Marissa Mayer is citing. But: you could imagine a different story about ads and search. If you’re looking for something specific and hard to find, like Madonna concert tickets, an ad might be the best search result available. But what if you just need someone to unclog your toilet? In the old days, you might open the big fat phone book known as the Yellow Pages.

McDEVITT: I got a couple of them right in front of me. 

Ryan McDevitt again:

McDEVITT: A-AAAA Sewer And Drains. AAAA Scots Plumbing. It’s just, like, ridiculous names.

Why did all these plumbers have such ridiculous names? Well, in a format like the Yellow Pages — whose listings were alphabetical — those plumbers’ listings would appear at the top of the plumbing category.

McDEVITT: They’re trying to attract customers who have an urgent need. They don’t have time to search. They probably don’t use this very often. And so that’s the one way to grab attention, is to be at the front of the Yellow Pages.

McDevitt began studying plumbers when he was a graduate student; he was looking for a dissertation topic. He and his wife were house-sitting for another couple when he suddenly got an idea.

McDEVITT: And so I jumped out of bed in the middle of the night and they actually had a Yellow Pages. And so I looked and sure enough, there in the plumbing category, a bunch of A’s. Locksmiths, a bunch of A’s. And then you start to think about the logic behind that. So I spent the better part of two years calling every plumber in Chicago.

And what was McDevitt hoping to learn? He wanted to know if there was any sort of relationship between the kind of plumber who would use a ridiculous name to get to the top of the Yellow Pages listing and the quality of the plumber. The paper he wound up writing is called “‘A’ Business by Any Other Name.”

McDEVITT: I thought it was a clever title because I put “A” in quotation marks. But I guess that’s kind of lame outside of economics.

It’s okay, Ryan, your humor is safe with us. Anyway: McDevitt set out to compare plumber quality, as measured by data he got from the Better Business Bureau, and plumber name, based on how they listed themselves in the Yellow Pages. What’d he find? The plumbers with a lot of A’s in their names tended to be more expensive and lower-quality than other plumbing firms. He also found that the low-quality firms listed themselves under multiple names.

McDEVITT: You’d call up AAA Plumbers and it’s like, “Hi, I’m Mike, what do you need?” And then you go down to A Best Plumbers. “Hey, it’s Mike, what do you need?” And then you go back to the Best Plumber, Chicago, and, “Hey, it’s Mike.” And those plumbers are the worst possible plumbers. They’re really trying to manipulate the Yellow Page listings to get a lot of results. 

The Yellow Pages still exist, online, but that’s not where most people go today when they need a plumber. They go to Google. And there, McDevitt argues, the same logic applies: firms that do everything they can do grab a customer’s attention are not competing on quality.

McDEVITT: Now you’re blanketing Google with a bunch of listings. You are blanketing Google Maps with a bunch of listings. You’re trying to win the first ad slot on Google just like you try to get the biggest ad in the Yellow Pages. All these behaviors are consistent with the story — you’re trying to attract uninformed consumers who aren’t going to search a lot. 

DUBNER: So if I’m reading this correctly, your research showed that plumbing firms that advertise on Google have 13 times as many complaints as firms that don’t. First of all, am I misreading that? Thirteen times is the actual magnitude?

McDEVITT: That was the right number back in 2007, ’08, when I was pulling the data. So it’s a big difference. And again, the logic is the ones that are trying to attract these one-off customers who are searching quickly, those are the ones that you can rip off. 

DUBNER: I’ve seen other research on position auctions for Google advertising — this is from Susan Athey and Glen Ellison, I believe — which says that position auctions actually make search more efficient, and they sort firms from high- to low-quality. Do you agree with that?

McDEVITT: You know, it’s a theory paper and as a theory paper, a beautiful paper. It depends a lot on the assumptions in the setting. And so they did not have plumbers in mind when they wrote that paper. What’s different in the setting of plumbers is the costs are very different, the profits are going to be very different if you’re not providing the service in a legitimate way, if you’re a fly-by-night plumbing firm, you know, those are the ones are going to be most profitable and get the most value from winning the auction. 

A few years ago, Google was in the news for a locksmith scam. A shady call center would create fake listings for locksmiths across the country advertising very low prices. If you clicked one of their links, the call center would send out a poorly trained subcontractor who then demanded three or four times the listed price. Google eventually cracked down on these call centers by requiring special verification to advertise locksmith services. This kind of problem goes way beyond shady locksmiths and plumbers — and beyond Google as well.

McDEVITT: I’ve noticed pretty plainly over the past couple of years that any time you try to search for anything on Amazon, it’s really hard to find what you’re looking for. The first dozen or so listings is mostly junk. 

McDevitt, we should say, actually worked with Amazon for six months a few years ago.

McDEVITT: I’m looking for a phone charger, AirPods, whatever it is. These sellers have found a way to manipulate their search algorithm and Amazon is making a lot of money from the ad placement now. They’ve taken a page out of Google’s playbook. It’s quite lucrative for them to auction off those slots, and it’s not always with the customer’s best interests in mind.

So maybe these are just intrinsic properties of the modern web, the spammy and junky landscape you have to slog through in order to glean the benefits. But Tim Hwang — the former Googler, now a media researcher — Hwang argues it did not have to be this way.

HWANG: It was a dream in the early 2000s when we saw Wikipedia coming up, like, “Wow, maybe this is an alternative way to generate all sorts of things on the internet.”

Wikipedia launched in 2001. Wikipedia is hardly perfect, or definitive; it is, however, an example of a vast website that provides valuable information to a huge audience — and it doesn’t accept advertising. It’s a non-profit, and it relies on volunteers to create and maintain content. But, let’s face it, Wikipedia is pretty much a one-off; it is not a model for how the web is won. The Google model, built on billions of dollars of ad revenue — that is the model that’s winning.

HWANG: I think what that has effectively ended up doing is sucking the air out of other experiments that could happen. There is basically a generation of investors in Silicon Valley where you pitch them on a tech idea, and they say, “Well, wouldn’t ads work better? We know ads already work. Why don’t you do ads?” So, if I right now were, on the air, to propose to you that we should start a subscription-only search engine, you’d say, “That’s ridiculous. People would never pay for that because they’re so used to having a search engine for free.” And I think that that is also a factor of historical accident, right? That basically, since advertising became so dominant so quickly so early, we have sort of set the norms of the market in ways that make it really difficult for alternative models to form. And so I think it’s almost a failure of imagination to believe that things could not have turned out on the internet any other way.

Would it be ridiculous to launch a subscription-only search engine? When we first published this episode in 2022, we talked about Neeva, an ad-free and subscription-based search engine, which has since shut down. Then again, Perplexity and ChatGPT are offering premium versions for $20 a month — and they can do more than just search. Coming up after the break: Could the Internet have turned out another way?

*      *      *

We first spoke with the economist Ryan McDevitt in 2021, for an episode about kidney dialysis. His research had found that the consolidation of the dialysis industry into just two big companies had led to worse outcomes for dialysis patients. Here’s the McDevitt quote that stuck with me from that episode: “Bad things … can happen when there’s no competition.” So I asked him if the lack of competition in online search can also lead to bad things.

McDEVITT: This is a very standard theoretical result, where if you have almost absolute market power, you’re going to increase prices and you’re going to reduce your quality to the extent that it maximizes your profits. In healthcare, that’s very bad for obvious reasons. In a tech setting like search, the implications are a little bit different.

As of today, Google handles about 90 percent of online global search activity. What’s wrong with that?

McDEVITT: Mostly you’re going to not have that same incentive to continue to innovate, to offer the highest quality products, because you don’t have to. You’re just a little bit less disciplined. They still are innovating, of course. They’re hiring many Ph.D.’s and they’re putting out new great products all the time. But that rate would be even faster if they were worried about losing all their customers to the competitor down the street. 

There are a few search competitors — Bing in the U.S., Yandex in Russia, Baidu in China. There’s also DuckDuckGo, a privacy-focused search engine — but again, Google’s global market share is 90 percent. Google’s Chrome browser is also easily the biggest browser in the world, with around two-thirds of global market share. In second place, with just under 20 percent of the market, is the Apple browser, Safari, which runs on iPhones and iPads and MacBooks. In 2012, Google co-founder and then-C.E.O. Larry Page argued that Google couldn’t really be a monopoly because — in his words — “It’s easy for users to go elsewhere because our competition is only a click away.” But Google pays Apple a reported $18 billion a year to be the default search engine on Safari. And that’s why, if you type into your iPhone “Chinese food near me” or even “Why is Google Search so bad?” you will get results from Google — not Bing, not any other competitor.

STOPPELMAN: And what’s that designed to do? Well, it’s designed to make sure that consumers don’t have to make an explicit choice or that Apple doesn’t have to think about providing other alternatives on a level playing field with Google. It’s protection money, essentially. 

That is Jeremy Stoppelman, C.E.O. and co-founder of Yelp. Yelp is a huge, crowd-sourced review site where tens of millions of users rate restaurants, stores, even plumbers and locksmiths. Stoppelman came up with the idea in 2004, when he had a hard time finding a good doctor.

STOPPELMAN: In 2005, we started to get traction, particularly in San Francisco, with random people finding our site and contributing local reviews that had never really been created. You could look up a dry cleaner and see what people in the neighborhood thought of that dry cleaner, and that was an incredibly novel concept. So it wasn’t long before Google took note. 

Early on, Google licensed Yelp’s content.

STOPPELMAN: Ultimately, that partnership was not long for this world because Google decided it really didn’t want to just organize the world’s information, at least when it came to local search. It felt like it needed to own that space.

Google then tried, and failed, to buy Yelp. And then Google began inviting users to post reviews of local businesses directly to Google. If you use Google Maps, you’ve seen these reviews. And how did Yelp feel about that? Here is Jeremy Stoppelman testifying before the Senate Judiciary Committee, accusing Google of unfairly prioritizing its own content and even stealing Yelp’s content.

STOPPELMAN/C-SPAN: Allowing a search engine with monopoly market share to exploit and extend its dominance hampers entrepreneurial activity.

That was in 2011. Stoppelman still feels strongly about Google.

STOPPELMAN: They felt like no rules applied and they could do whatever they want, which I think is a tell-tale sign of an abusive, dominant monopoly.

Stoppelman, like some of us, is old enough to remember a different Google.

STOPPELMAN: If you go back to the early days of Google, it was ten blue links, minimal ads. And those links were intended to be the best content on the web that they could find. And they were known for crawling every nook and cranny of the web. And users loved it. That’s what made them successful. The problem with monopoly power is that you can degrade the experience because you’ve locked in that user. So in the early days, you have to delight users, as Google did. You have to give them a product that is just way better. But once you’ve become wildly successful, you might not degrade the results everywhere, but there are certain areas you could cut some corners and dial some knobs and generate a heck of a lot more money. And at that point, you really become a tax collector. And that’s what Google really is, they’re a tax collector. 

So if you’re Jeremy Stoppelman, what would a truly competitive market for local search look like?

STOPPELMAN: Get Google and these other big tech companies to stop putting their thumb on the scale for themselves. Don’t self-preference. Pick objectively the best content or the best website, period. There are other websites out there. Yelp is not the only one. There’s TripAdvisor, there’s Angie’s List, HomeAdvisor, take your pick, Zocdoc. There’s lots of different players that have consumer reviews, but Google makes sure that the prime real estate is dedicated to two things: ads and themselves. 

Google is now facing several antitrust cases in the U.S. alone, including one brought by the Department of Justice alleging that Google is “crippling the competitive process, reducing consumer choice, and stifling innovation.” They’ve already been fined for antitrust violations in the E.U. and in India. And just recently, back in the U.S., Google agreed to pay nearly $400 million to settle claims from 40 states that Google kept collecting users’ location data — valuable stuff if you want to serve local ads — even after users had opted out of being tracked.

STOPPELMAN: I’ve been shocked at how fast the winds on this topic have shifted. When we began speaking out about Google, we got a lot of eye rolls, both from the industry as well as from regulators, frankly, and from the White House. The No. 1 most frequent corporate guest in the Obama White House was Google, to give you some idea of where it was not so long ago. 

In early 2022, a Brooklyn man named Vitaly Borker was arrested for running a fraudulent online store that sold eyeglasses. It was the third time he’d been arrested for pretty much the same thing. He would send customers faulty or counterfeit glasses, overcharge them — and then, if they complained, he would harass or threaten them. Believe it or not, this was part of Borker’s business model. As he told The New York Times, a mistreated customer might register an online complaint, which would show up in Google’s index with links to his site — which, in turn, would boost his ranking on Google and bring in even more customers. Google eventually announced it had changed its ranking algorithm to prevent this sort of behavior. But you can bet that plenty of other people are still finding ways to game the system. After a couple of decades with Google as the main gateway to online search, the web is constantly shape-shifting to fit inside Google’s algorithms.

MAYER: I do think the quality of the Internet has taken a hit.

That, again, is Marissa Mayer, one of Google’s first software engineers.

MAYER: When I started at Google, there were about 30 million web pages, so crawling them all and indexing them all was relatively straightforward. It sounds like a lot, but it’s small. Today, I think there was one point where Google had seen more than a trillion URLs. 

So is that URL inflation what’s responsible for worse search results?

MAYER: When you see the quality of your search results go down, it’s natural to blame Google and be like, “Why are they worse?” To me, the more interesting and sophisticated thought is if you say, “Wait, but Google’s just a window onto the web. The real question is, why is the web getting worse?”

DUBNER: Okay, Marissa. “So Google’s just a window onto the web. Why is the web getting worse?” 

MAYER: I think because there’s a lot of economic incentive for misinformation, for clicks, for purchases. There’s a lot more fraud on the web today than there was 20 years ago. And I think that the web has been able to grow and develop as quickly as it has because of less regulation and because it’s so international. But we also have to take the flipside of that. In a relatively unregulated space, there’s going to be economic misincentives that can sometimes degrade quality. And that does put a lot of onus on the brokers who are searching that information to try and overcome that. And it’s difficult. It kind of has to be more, in my view, of an ecosystem-style reaction, rather than just a simple correction from one actor. 

One way Google has tried to fight the overall decline in quality is by supplementing its index of a trillion web pages by showing you some selected content rather than just showing you links. If you ask a simple question about cooking or the age of some politician or actor — or even “what’s the best podcast?” — you may see what Mayer calls an “inline result,” or what Google calls a “featured snippet.” It’s a bit of text that answers your question right there on the search-results page, with no need to click on a link.

MAYER: I think that Google is more hesitant to send users out into the web. And to me, you know, that points to a natural tension where they’re saying, “Wait, we see that the web sometimes isn’t a great experience for our searchers to continue onto. We’re keeping them on our page.” People might perceive that and say, “Well, they’re keeping them on the page because that helps them make more money, gives them more control.” But my sense is that recent uptick in the number of inline results is because they are concerned about some of the low-quality experiences out on the web. I think that the problem is really hard. You might not like the way that Google’s solving it at the moment, but given how the web is changing and evolving, I’m not sure that the old approach, if reapplied, would do as well as you’d like it to.

Coming up after the break: it took a while, but we did finally get someone from Google to speak with us.  

REID: Yes, you may ask one question. 

She’s just kidding — we get to ask plenty of questions.

*      *      *

It’s easy to be nostalgic for the old days of Google Search: 10 blue links, simplicity. But the simplest interface might not be able to handle today’s Internet — and the declining quality of a lot of online content. And here’s one more problem to solve: what you want from Google and what I want from Google may be different. This is Liz Reid.

REID: You come in and you might type a question, but actually what you meant by that question is actually very different than what someone else might mean. 

Reid is vice president of Search at Google.

REID: So I joined Google in 2003 in the New York office actually, and started there as a software engineer out of college.

DUBNER: So we all know that the fortunes of Google and Alphabet were built on search, which creates this massive audience for advertising. Since you’re running Search now, do you fantasize about how Google Search could be different/better if it were not an ad-supported service? Do you stay up thinking, “Oh man, if only we had gotten around to some kind of a premium, paid version, just imagine what I could do with this sucker”?

REID: In all honesty, one of the things that I love about working on search is that Google’s mission around “universally accessible” is something we take very seriously. And I think that’s really one of the powers of the ads model. Like a premium model would say, great, it’s available for a small subset of the world. And I think that’s one of the best parts about the internet, is that the addition of advertising really, truly democratized access to information. If you think about things like a billion users use us a day, like that’s a lot of people.

DUBNER: I’m surprised it’s not more. What are the other six billion using? 

REID: It’s more than a billion. And also some of the 6 billion are three years old. But over a billion people are using it per day. That ability to provide information to the whole world, to help empower them, to help them go about their daily life or to help them find a better job or build new skills, I think that’s an amazing part of Google’s mission. And the ad system really fundamentally is what supports that.

DUBNER: Look, I am not in a position to complain because A) I’ve used Google every day from the minute it was available. So thank you, first of all. Second of all, when it comes to ad-supported — like this show survives, it exists because of advertising. So I’m familiar with the model. I think the model is a really good model and I think it makes all kinds of businesses viable. That said, when you think about the original paper from your founders, there is a famous line: “Advertising income often provides an incentive to provide poor quality search results.” So how do you reconcile that early reckoning of what’s optimal with what has become real?

REID: Behind the scenes in Search, we take great pride in not only trying to provide high-quality results, but ensuring that they are not influenced by ads. So we have a very strong culture that says whether or not you’re an advertiser, does not allow you to change the results we show. You cannot pay us to change how you show. Those results are very pure, and we go to great lengths to ensure that that is the case. And so the quality of the search we provide is run separately from the ads to do that. Relevant ads are actually quite helpful to people and people often click on them and if they don’t click on them, they disappear. We don’t get paid if people don’t click on the results, it’s not we just get paid for showing them. But fundamentally, what we can do with the search results isn’t affected by the ads. We build out our search results separately.

DUBNER: So here’s my recent-ish experience with Google, having used Google for a million years — but in the last year-and-a-half or two, I felt more and more that whereas in the old days, a Google search would typically produce results that were really useful to me if I wanted to learn something about the search topic, and these days, it feels to me that a Google search will more typically produce results that are useful if I want to buy something related to the search topic. And then I find myself going down further and then yeah, I find what I’m after. But it feels like a different environment. That is one person’s experience. And I’ve read a lot about other people who have related complaints about Google. So for those of us who are whining, like me, at the moment, is this our perception that’s changed or is the reality changed? Are there more ads, or more ad-related or more sponsored-related or even Google-generated snippets and so on, on the first page of a Google search result than there used to be? 

REID: So the number of ads we show on a page has been capped for several years, and that hasn’t changed. I do think people sometimes come to buy as well as to research a topic, and we have done more to provide sort of non-ads-related-information about products. I do think we’re trying to continually ask the question, how do we help those who have the intent to buy successfully buy, and those who don’t have the intent to buy, not buy, right?

DUBNER: Is there a secret word that I can type in that indicates that I’m not looking to buy, I’m looking to learn?

REID: It’s a great question. So there’s not a magic word that we’re like, “Aha, this is the magic word that filters it.” But we’ll think about this. But we are trying to think about how do we do that. And so sometimes people will use, like, “Research X,” but I wouldn’t promise that that’s the secret word. I think that is something that we are actually continually exploring on.

DUBNER: Now, what about the other non-ad-results that I’m seeing? I think you call these “snippets,” correct? 

REID: We typically use the word “snippet” internally to refer to some additional context or reference from the page itself that helps you decide if the results you’re about to click on is going to be more relevant. So, we can save you a lot of time by not having you spend your time going to a page you’re not really interested in.

DUBNER: So for instance, I looked up the other day — I looked up Premier League standings, the English Soccer League, and the result I got, it pulls up a table that looks like it might be a summary from some Premier League-affiliated site, but it turns out that it’s a Google table. And when I click into it, I get a much larger version of that. And again, as I keep clicking around it, I just stay within this little island of Google-generated results that are, no offense, not great. It’s as if I was looking for The New York Times or The Wall Street Journal, and instead I get like a pretty good high school newspaper. Like, it’s trying, but it’s not the same thing. It looks pretty real because it’s got the Premier League lion wearing a crown there — but then when I click the feedback link, Google is asking me, “What do you think? This is helpful. This isn’t useful.” But then, “The score is wrong or missing.” And I think, “Wait a minute, I don’t want to be the one to have to tell you that the score is wrong or missing. I want you to tell me what the proper score is.” So again, how is this a better search result for me, for the user, than something that’s editorially curated versus algorithmically created?

REID: Okay, so this unit is not algorithmically created. The data would be sourced from a high-quality provider. We go through a pretty extensive vetting process to ensure the overall accuracy of any place we’re sourcing, the data is high-quality, highly accurate, one of the things that is really important to us at Google is that we not only show relevant results, but that they’re very reliable, and very trustworthy. Now, our assumption is not that the data should be wrong, but users get very frustrated if we are wrong and they can’t tell us. We all make mistakes, right? And sometimes our providers make mistakes. So if there’s a mistake, great, then users can tell us and then they can help us improve our product.

DUBNER: So I get that and I appreciate that. I just don’t understand why if I’m searching for something that seems prima facie, a request to go to the source, essentially, why I would want to get a secondhand version of that. I don’t see why it’s in Google’s interest to do that.

REID: So in your statement, you said, “my request is to get to the source.” That is your intent, okay. With all due respect, it’s a very good and well-founded intent.

DUBNER: But you’re saying I’m a weirdo.

REID: No, no, no. I’m just saying, not everyone is identical in this world, right? Some folks may want to go and check on this, like, every day. “I just want to go check, what’s happening. What changed, what changed, what changed?” And they’ve got 5 seconds to go look at it, right? And so when we think about really trying to help people with their information needs, we need to recognize that both people care about depth, but they also often care about speed. If you go back in time at Google, this was one of the things that Google prided itself on, was that it was fast. But fast isn’t just about how fast the server renders the page, right? Fast is about how quickly you can get the information that you seek out. Why would this be useful for Google to do? Because some subset of the people who are coming really just want to find out the answer at a glance. Other people really want to do more in-depth research. It is still the case that we send more and more traffic to the web every year, but we also give people multiple options so that they can get their tasks as easily as they can and really meet their need.

DUBNER: So here I am complaining about Google Search, which is absurd, because it’s such a bounty for me and everybody. But not only that, I also use the other products that have been created because Google Search is so profitable. Google Maps, for instance, which I know that you’ve worked on. This is a live atlas in my pocket it’s unbelievable. So do you think the moral of the story is someone like me who’s complaining that Google Search doesn’t seem quite as good as it used to for research purposes, should I just shut up and be grateful for everything you’re giving me every day? 

REID: I think one of the great parts about working on a product like Search is that users keep raising their expectations about what’s possible. In the early days in Google, if you entered a very simple query and you didn’t get the good results, then you thought you did a bad job entering the query. Now you can enter a super-long, complex query and if Google doesn’t get it immediately, you’re like, “Google is an idiot.” But that’s great because that challenges us to do better and better, but the bar of what users are expecting raises. If you asked a query like “When does Starbucks open?,” we would show you nearby Starbucks and their hours of opening. You have to understand that it’s not like Starbucks opens at the same time in every single branch. But if we didn’t understand it, you’re like, “What are you doing, Google?” Users’ expectations rise and that keeps us on our toes and that’s what honestly drives us. If you didn’t complain, if none of you pushed us, we might not make better and better products. So I love the provocation, to challenge us to make search better and better every year. We’re in it to help you. 

DUBNER: Do you ever find yourself frustrated by search results in your civilian life and then use that experience to tweak the algorithm or tweak the results? 

REID: Yes, I find my examples. We don’t tweak it to fix my needs, to be very clear. I send complaints probably every two or three days. You might think that we all think it’s perfect and it’s not. One of the things I will continually push us to get better on is how do we help people with more complicated tasks and how do we help people with exploration, right? Google is both immensely useful if you want to do something like buy a house, and it’s still really hard, because it’s actually not just about buying a house. You actually have to figure out, can you afford a mortgage? Where would you actually live? And then you buy a house and you’re like, “Oh my gosh, nothing works.” And you have to find a contractor and these things go on, right? And so if you think about that, if I took Google away, you would go nuts. But you still have to do a lot of hard work. And so how do we continually push to go and say, okay, well, if that’s what you’re trying to do, how do we make that as easy as possible?

As we noted earlier, there’s no such thing as a free anything. And Google is no exception. It’s all about whether you find the tradeoff acceptable. I went back to Ryan McDevitt, our Duke economist friend, to ask what he makes of the Google tradeoff, overall.

McDEVITT: I’m very happy with that trade-off. I think we’re getting the good end of that bargain. You know, clicking on a few bad links doesn’t cost you a whole lot in the scheme of things, as long as you are then careful when you actually go to make the transaction. So a few junky links — I mean, it’s not so bad. 

What do you think of the Google ecosystem, this unbelievably vast network of products and services and software and ideas, most of it funded by the simple fact that billions of people use their search engine every day, whether they choose to or not? Is that trade-off acceptable to you? Do you even consider it a trade-off? I’d love to hear what you think — about this episode or any other. Send us an e-mail to radio@freakonomics.com. And if you want to learn about other challenges to Google’s search monopoly, you can check out my interview from last summer with Microsoft C.E.O. Satya Nadella; it’s called “Satya Nadella’s Intelligence Is Not Artificial.”

*      *      *

Freakonomics Radio is produced by Stitcher and Renbud Radio. This episode was produced by Alina Kulman and Zack Lapinski. Our staff also includes Eleanor Osborne, Elsa Hernandez, Gabriel Roth, Greg Rippin, Jasmin Klinger, Jeremy Johnston, Julie Kanfer, Lyric Bowditch, Morgan Levey, Neal Carruth, Rebecca Lee Douglas, Ryan Kelley, and Sarah Lilley. Our theme song is “Mr. Fortune,” by the Hitchhikers; the rest of the music this week was composed by Luis Guerra. You can follow Freakonomics Radio on Apple Podcasts, Spotify, Stitcher, or wherever you get your podcasts.