What this covers

Sridhar Ramaswamy, former head of ads at Google, argues in conversation with Russ Roberts that ad-supported search has fundamentally inverted its purpose. Rather than serving the person who searches, the system now serves advertisers—a shift enabled by the CPC model, which lets advertisers bid on queries and measure whether clicks convert to sales, creating a self-optimizing machine for extracting value. Ramaswamy traces how Google's acquisition of DoubleClick positioned it as a middleman across the internet's advertising ecosystem, yet search itself remained the crown jewel, generating over $120 billion annually. This comes at a cost: roughly $160 per person per year in foregone value, extracted as reduced search quality and mortgaged attention. The conversation moves from how the ads economy works in mechanical terms to what it does to search quality and human agency.

The core tension Ramaswamy identifies is structural. On commercial query searches—those with economic intent—the dominant engine cannot serve both advertiser and searcher; it chooses the former. Google's market dominance cannot be dislodged by switching costs alone: browser defaults and ingrained habit create what he calls a myth of "competition is just a click away." Traditional antitrust remedies founder because current law uses a consumer welfare standard that does not apply well to free, ad-supported services. Ramaswamy's answer is Neeva, a subscription search engine with no ad business and no separate data products—a structural guarantee of alignment with the user. The conversation also touches on how the Optimization Mentality behind engagement metrics has turned every site into an Attention Merchant, surfacing provocative content and conspiracy theories, and how remarketing and algorithmic feed curation exploit human manipulability in ways surveys cannot capture.

Sharpest takeaway

Ramaswamy argues that ad-supported search has hijacked search from serving the searcher to serving advertisers, imposing a hidden economic and attentional tax, and that a subscription model like Neeva is needed because conventional market competition and antitrust cannot dislodge entrenched search monopolies.

  • The CPC ad model aligns incentives toward maximizing attention and conversions, not serving user curiosity
  • Free search is not free: it imposes a per-person economic tax and mortgages user attention to advertisers
  • Competition is not a click away because user habits and locked browser defaults prevent challengers from being tried

The argument · threads8 threads · 36 claims
0.85

Ad-supported search inevitably prioritizes advertiser and platform interests over user needs.

6 pointscentrality 5/5
  • The taking-the-ad-model-to-its-logical-extreme means that for any commercial query the user will eventually see a page full of ads, and Google Search's ad model fundamentally does not work for serving the user.

    “There is a demand for more and more money, and so you end up taking more and more space. It's very relentless. And, the logical conclusion of search is that for a commercial query you are going to see a page full of ads.”

  • Because ad revenue scales with time spent on a site, every site became an Attention Merchant, making headlines more provocative and surfacing conspiracy theories and anger because such content drives engagement, clicks, and responses, leaving users with less agency over their attention.

    “every site has figured out that the more time that they can get you to spend on them, the more ads money they're going to make... There is a reason why you see conspiracy theories and anger on Facebook. You know what? It drives attention.”

  • Search is no longer about the searcher; the moment a user enters a commercial query, the dominant search engine serves itself and advertisers rather than answering human curiosity, meaning the core search product effectively no longer exists -- which was the foundational observation behind Neeva.

    “The first issue is that search is not really about you the searching anymore. It's about serving Google. It's about advertising... The dominant search engine in this role has become much more about serving itself, serving advertisers. So, that core product simply does not exist.”

  • Search costs roughly a billion dollars to run and could be a profitable transactional product on $5-6 billion in revenue, but the ads economy mortgages users' entire attention so Google makes over $120 billion (over $50 billion in the US), amounting to roughly a $160 annual tax per living person for a service that costs pennies per person.

    “Google makes over $120-billion, more than $50- just in the United States. If you work out the math, that's a tax of $160 dollars per living person in the country for a service that costs, like, pennies per person per year.”

  • Even services the user pays for, like the New York Times and Wall Street Journal at $15-20 per month, fill their mobile apps with ads making them unpleasant, because there is no limit to ad encroachment when the incentive exists -- a model Neeva refuses to adopt.

    “even the services that I pay money to... fill themselves up with ads, which is something that we, for sure, never want to do, because there's no limit to that. I find sites like the New York Times, but even the Wall Street Journal, to be unpleasant experiences on my phone, even though I shell out $15, $20 bucks every month”

  • Much of the benefit of scale in technology that could have flowed to broader society has instead gone to individual companies because of the ads model, since the ad-funded firms reap the profits rather than passing efficiencies to the public.

    “much of the benefits of scale in technology that could have come to larger society have gone to individual companies because of the ads model. Because they're the ones that reap the profit of these models.”

0.71

Incumbent dominance persists because browser defaults and user habits lock out competitors despite claims competition is one click away.

6 pointscentrality 4/5
  • The incumbent claim that 'competition is just a click away' is a myth, because users are ingrained with the habit of searching by opening a new tab and typing a query, and locked browser defaults (Safari/iOS, Chrome) prevent challengers from being tried at all -- so the free market is not working in search.

    “competition is not a click away. That's, like, a myth that is perpetuated by incumbents... People are ingrained with their habits. People search by opening up a new tab and putting in a search query. And, if you cannot become a search provider there, you might as well not exist.”

  • Compared to enterprise software, cloud software, or crypto, the amount of actual competition in adtech and search is minimal, and 23-year-old entrepreneurs avoid building search engines because they know they won't get funding for such a daunting idea.

    “compared to say other... whether it is enterprise software, whether it is cloud software, or whether it's crypto software companies, the amount of actual competition in adtech, in search, is minimal.”

  • When a company comes of age determines its culture and business principles, which is why Yahoo built a search directory rather than a search engine, since it emerged when there were few enough sites that a curated list was feasible.

    “when a company comes to be, determines its culture, determines its business principles--this is the reason why Yahoo set up a directory for search and didn't think about a search engine, because it came of age at a time... when there are not that many sites”

  • Search is the ultimate expression of human curiosity because there is essentially nothing that passes through people's minds that they will not type into a search engine.

    “another way to think about search is that it is the ultimate expression of human curiosity. There is literally... nothing that'll go through many people's heads that they will not type into a search engine.”

  • Google's PageRank insight can be distilled as 'you're popular if every popular person says you're popular' -- ranking pages by the endorsements of other already-popular pages.

    “their idea of Page Rank, which I can roughly distill as, 'You're popular if every popular person says you're popular.'”

  • Deep-learning models can assimilate billions of documents and generate useful summaries, so search may become a conversation in which a model that has digested much of the internet synthesizes new information for the user rather than just returning existing pages.

    “The deep-learning models are another very fascinating angle. They have the power to assimilate billions upon billions of documents and generate useful summaries. So, who knows? We might actually be producing content that doesn't even exist, where search is more of a conversation between you on this model that has digested much of the Internet”

0.66

Click-based performance advertising creates measurable accountability absent from impression-based models.

5 pointscentrality 4/5
  • The cost-per-click model aligns interests because Google has no incentive to show impressions that don't get clicked, advertisers have no incentive to buy clicks that don't convert, and advertisers can buy clicks, measure resulting sales, and tune the system accordingly.

    “the CPC model actually aligns interests. It's very powerful. Google has no interest to keep showing ads, because you're not paying them on impressions. You have no interest to get back-clicks, because you know that it's not going to lead to great conversions.”

  • Advertisers should never pay for impressions on social media but always pay for clicks, because clicks are trackable and verifiable for effectiveness while impressions are not.

    “You should never be paying for impressions on social media. You should always be paying for clicks on social media, and sure as heck, you can track whether they're any good or not.”

  • Online advertising's deeper power is that advertisers can back-calculate cost per acquisition (CPA) by tracking whether clicks led to purchases, then deploy software to automatically tune bids and discover new keywords across millions of terms, creating a self-optimizing performance advertising industry.

    “the next logical step is, they can back-calculate what is called CPA--the cost per acquisition... you set in place software that's going to automatically tune these things, find new keywords, adjust bids up, adjust bids down.”

  • The click is the core concept of online advertising because it creates three-way observability: the user takes the action, and both the platform and the advertiser observe it, which is one of the few events in life with such mutual visibility.

    “the core concept in it is a click. You as a user express interest with an action, the click... It's one of the few events in life there is a three-way observability. It's like the three parties interacting.”

  • AdWords inverted the traditional advertising model: instead of the publisher guaranteeing audience reach to advertisers, Google offered an unending stream of queries and required advertisers to do the work of figuring out which queries were relevant, showing ads only if their quality was deemed good enough.

    “AdWords turned this model on its head. It told advertisers, 'We have inventory. We actually don't know exactly what. We have an unending stream of queries. Why don't you, the advertiser, do the work to figure out which queries are relevant to the product that you are selling?'”

0.56

Manipulation through tracking and remarketing exploits human psychology despite users' belief they cannot be persuaded.

3 pointscentrality 4/5
  • People are eminently manipulable: in surveys, every individual claims they can never be persuaded to buy anything, yet 50% admit anonymously to buying junk they didn't need because of remarketing ads, showing 'free' attention-driven products carry real costs through manipulation.

    “Every single person, if you talk to them, individually is 100% convinced that they can never be, like, persuaded to buy anything. But, if you ask in anonymous surveys whether they bought stuff that they didn't really need because of remarketing ads, 50% of people will say, 'Absolutely.'”

  • Engineers optimizing conversion rates will choose to show the same ad many more times (e.g. 30 vs 10) because the principle is purely conversion optimization; even when reducing remarketing impressions 12-15% would cost only 0.25% of revenue, the team's logic is that since no harm is measurable and a competitor would show it anyway, there is no reason to leave money on the table.

    “We can reduce the number of impressions, ads, that we show on these remarketing ads by some ridiculous number... 12 or 15%. 'And, if we do that, we will lose about a quarter percent of revenue. But... we can't really see that these ads are causing any harm. So, why would we want to leave .25% on the table? And, by the way, if you don't show it, someone else is probably going to.'”

  • Remarketing often targets people with no genuine purchase intent (mere idle curiosity), yet being pounded with the same ad hundreds of times will eventually persuade them, exposing the common fallacy that we are invincible to manipulation.

    “if you're pounded with something 200 times, you will be persuaded. This is a common fallacy of our invincibility. We are not. We are eminently manipulatable.”

0.43

Google's scale and structural advantages in search infrastructure create barriers only a genuine alternative product can overcome.

3 pointscentrality 3/5
  • Around 2005-2006 Google concluded the advertising market outside its own properties was far larger than Google Search revenue and bought DoubleClick to become the middleman for ads across the internet, but this assumption was wrong because Google Search remained its biggest and most profitable business.

    “Google concluded that the market for advertising--outside of its properties, outside of Google Search--was far, far larger than the money that it could make on Google Search. Turns out, as I said, this is a terrible--this is a bad assumption. Google Search is still way bigger.”

  • Outside of Google and Bing, Neeva is the only company that is a genuine product company focused on search, having moved from using Bing's APIs to building its own large-scale infrastructure, which lets it innovate on programming and product queries and even index a user's personal data rather than just slapping a privacy layer on existing products.

    “outside of Google and Bing, we are the only company that is actually a product company focused on search. We started using Bing's APIs, but we have since made very large investments in all of the infrastructure needed to create a better product.”

  • Working on Google's search ads system meant every line of code could bring down the largest business Google had, with conference calls losing thousands of dollars every second while engineers scrambled to restore systems.

    “I've also been in lots of conference calls where we were literally losing lots of money, thousands of dollars every second, and you have to figure out how to bring systems back up.”

0.34

Neeva's structural commitment to no data products and encrypted search offers earned trust through business model constraints.

4 pointscentrality 2/5
  • Early research suggests search could be run in a true end-to-end encrypted fashion, allowing a search engine to search across a user's data on the user's behalf without the provider ever being able to see the content of that data.

    “There is some early research here that will let us run search in a true end-to-end encrypted fashion, meaning that we will be able to search on behalf of you all of your data in such a way that we cannot see the data.”

  • Neeva's biggest competitive hurdle is getting people to try the product (the top of the funnel), not converting free users to paid subscribers, because once users experience search without tracking and creepy ads, conversion from free to paid tiers is very healthy.

    “The biggest hurdle that we face, by the way, is not the point of converting Neeva users to paid subscribers... It is getting people to even try Neeva. That's the biggest competitive hurdle that we face.”

  • Ramaswamy personally invested as much as Greylock and Sequoia did in Neeva's first round, giving him substantial personal skin in the game beyond his time.

    “I put up my money for the first round of Neeva. I invested as much as Greylock and Sequoia, so there's more than--I'm just not talking. I care about the outcome and I'm personally invested in a very real way”

  • Trust cannot be asked for but must be earned as an ongoing relationship, and the real guarantee Neeva offers is structural: it has no data products and no other lines of business involving data, so being open about what it does not do is what makes it trustworthy.

    “trust is earned. It's cannot be asked for and it has to be earned. And, I see trust as an ongoing relationship... I think it's the 'what we don't do'--have any other lines of business involving any kind of data--that I think is the guarantee.”

0.33

Optimization systems without considering externalities cause harm independent of malicious intent or explicit evil.

1 pointcentrality 3/5
  • Picking a set of simple quantitative principles to operate by is dangerous because things outside your belief system happen and outcomes can go violently astray; harm arises not from evil people but from optimization mentalities that ignore externalities.

    “if you pick a certain set of principles by which to live, especially simple quantitative principles, you have to be very careful about things going violently astray, because lots of things happen that are outside of your belief system.”

0.29

Government antitrust enforcement is too weak and underfunded to regulate tech giants effectively.

3 pointscentrality 2/5
  • Government is woefully under-prepared and under-invested to take on tech behemoths, illustrated by Google's General Counsel having more lawyers on his team than there are total employees at the FTC, leaving regulators unable to check companies that stifle competition.

    “Kent is GC, head of legal, at Google. He has more lawyers than there are employees in the FTC. That's the world that we're living today. And so, our government... is woefully under-prepared and under-invested to take on these behemoths”

  • The consumer welfare standard in antitrust is not law but an ad hoc interpretation that became broadly adopted, and it was originally conceived in economic terms of producers and consumers rather than ordinary individuals, so current antitrust law largely does not apply to ad-supported companies.

    “The consumer welfare standard isn't law, it's an ad hoc interpretation that's broadly gotten adopted.”

  • It is incoherent to expect competition problems to be solved by one two-trillion-dollar company chasing another two-trillion-dollar company to rescue consumers, while assuming it is not the government's job to protect them.

    “you're, like, 'Wait. Don't worry about competition. There is a two trillion dollar company that is out chasing other two trillion dollar companies to rescue us.' Listen to yourself. And, somehow we don't think that that's the government's job”