Big TakeSearch is the most lucrative part of Google’s business. So when ChatGPT launched in 2022 and offered to answer users’ every question, it posed an existential threat to the company, forcing Google to respond. Google had to figure out how to compete. But integrating generative AI into Google search would fundamentally change the company’s core product, and it introduced big risks.
Bloomberg reporters Davey Alba and Julia Love spoke to Google executives and former employees to get a look inside Google's plans to transform search with AI.
Read more: Google Is Searching for an Answer to ChatGPT
Bloomberg Audio Studios, podcasts, radio news.
It's twenty twenty one and some employees on Google's web search team go to their leadership with an.
Idea having a chatbot that could answer some of user's questions directly alongside the links.
Julia Love is a tech reporter at Bloomberg who covers Google, and so was Davey Alba.
This was several months before the launch of the massively popular chat GPT in late twenty twenty two. So Google Search was raking in billions of dollars for Alphabet the parent company, and engineers and leadership just didn't want to mess with that money making machine.
So when this group of employees present this proposal to use artificial intelligence to answer some of users search queries directly, leadership shoots them down.
It was so tech at that time to think about disrupting the way doodle search worked.
Generative AI just hadn't had its moment yet.
The next year, open ai rolled out chat GPT, and all of a sudden, Google faced a reckoning. It can no longer just embrace the status quo. It has to adapt well. Three years later, Julia and Davey set out to assess how Google's belated attempt to infuse their core product with generative AI is going.
We spoke with a lot of former employees who had worked on the product and moved on. Doodle also granted us considerable access, and then we also fanned out and talked to the publishers and creators who really live and die by Doodle's business practices.
We had had this creeping sense over the past few months and maybe even a year plus at this point, that Google Search had been changing. There's been a lot of inks build over this idea that it has been changing for the worse, that Google Search is getting quite bad in quality.
The story that emerged is a window into how incorporating generative AI into Google Search has become an existential issue, both for Alphabet employees and for the websites the company's business is based on. I'm David Gera, and this is the big take from Bloomberg News today on the show Inside the Changing World of Google Search, with reporters who have gotten a first hand look at the way that plain white search page has been reworked, and from a web creator whose livelihood is threatened by it. Just think about what it would have been like to be one of those team leaders at Google back in twenty twenty one. Maybe you'll sympathize with how little interest there'd be in mucking up Google's clean, reliable search interface with new AI elements. After all, search made Google nearly two two hundred billion dollars in twenty twenty four. That's around sixty percent of Alphabet's total revenue, and it keeps a lot of people employed.
There are thousands of people who keep Google Search humming smoothly. The codebase is massive, and it's kind of a creaky machine in some ways. It's been around since the late nineties, and there are some relatives of that time, code written by Marissa Meyer, who went on to.
Become the CEO of Yahoo.
And I think that there's so much attention devoted to keeping those results delivered.
Quickly, that instant response Google spits out when you enter a search query. It's central to the product's appeal, which made it a challenge to add any kind of feature to search.
They know they rest losing users with any small increase in the time it takes the page to load.
User had started to complain that Google Search was getting worse. It was being cluttered with bot created junk sites built to maximize clicks that degraded the quality of Google Search results, and it alienated users.
The example I would always go to is when you land on a recipe site and there's this long prelude ahead of the actual recipe and the food bloggers life story. So that's one quite illustrative example of what web creators were starting to do, which was create websites that were written for Google Search and sort of catered to Google's algorithms, and that started to change the way the Internet look end made it also in some ways a difficult problem for Google to rank high quality websites high up in its results.
Google employees were aware of this problem, and then open Ai launched chat GPT. How did that verberate around Google alphabet more and more broadly, how did they react to the debut of that platform.
They reacted with a lot of panic. They declared a code read, which means all hands on deck, and within the Search Org, they transferred something like a thousand engineers, which is a sizeable proportion of the engineers that work within that division, and set them loose on this project of coming up with an answer to chat GBT. So how do you make search more integrated with generative AI technology.
It was just this very, you know, all hands on deck situation.
It's not as if Google hadn't worked with AI before. They'd been using the technology under the hood to make results better for years. But incorporating generative AI right into its flagship products, including search, overnight Google went from feeling cautious about it to doubleing down.
I think that the response that we saw was in some ways akin to Doodle's response to social media. Everyone at Doodle had to have a social component to their product. That was a mandate. Bonuses were tied to that, and we heard that after chat GPT every product had to have a generative AI feature. But I think the states were higher. It just felt much more existential than even social media did.
Because if AI.
Can answer users questions directly, then who needs the list of links?
Who needs Doodle?
This sudden pivot wasn't easy. Google had built its product as an information aggregator, directing users to other sites who themselves provided information. Providing information directly came with new risks.
One former employee told us that Doodle is kind of weighed down by this notion of Doodle speaks the truth.
That's how the public often sees it.
And so when chat GPT launched, it was a fun novelty. It was something new, but if it occasionally hallucinated, that was something people were kind of willing to accept as part of the compact with using cutting edge technology. And for doodle Search to be offering up that type of experience, I think the public expectation is just very different, and executives.
Were very aware of that.
Fully aware of the stakes, Google charged forward and a few months later it started beta testing a new feature, AI Overviews.
It was quite a move. It was really Google signaling that they were no longer so paranoid and afraid about making mistakes as generative AI is wont to do, and that they were willing to sort of play with search. And I think the huge advantage that Google has had is that they already have these products and services that have been in use among the general population, and you know, whether or not the public was necessarily asking for generative AI to be embedded in those tools, Google sort of made that decision for people and rolled these features out anyway.
In other words, Google's response to chat GPT has been to roll the dice on AI next up, how that gamble has worked out for Google and for the web creators who rely on those all important search results. That's after the break. Liz Reid overseas Google's push into the wild West of AI.
Liz Reid is a really interesting figure within the company. She's been with the company since two thousand and three, when she was fresh out of college. She spent most of her time in maps. She told us a story about Sarah de Brin and how he was just really urging the team like to launch as soon as possible, and Liz said that that was the right call, that it made sense to get user feedback as early as possible, even if all the kings hadn't been ironed out. And I think that that's a philosophy that we really saw her apply.
Former employees told Julia and Davy that the search team under Reid seems to have found a new lease on life.
It does seem like there is renewed vigor and energy in the organization. It had been through kind of a long period of slumber, but now people I think are really motivated. They have a target to catch up to, and they have been rolling out products much faster. Ster soundar Pichai has boasted in the earnings calls about how they have really succeeded in driving down the cost of serving up these AI answers, which is always the strength of Google's scale and efficiency.
But that growth in the AI accompaniment to search has come with some notable hiccups. It's generated a lot of errors and continues to do so to this day. That's a big issue for users, but perhaps the biggest issue has to do with important collaborators the websites it directs users to through search websites like Housefresh, which Giselle Navarro runs.
What we have been told by Google is that the search results have changed, and that whatever we were receiving in traffic five thousand and seven thousand people come to a site every day from Google that's not going to come back because the search results have changed.
House Fresh publishes detailed reviews of air purified. The site depends on the revenue it gets when users buy the products it's reviewed, and it relies on Google to get users on its page in the first place. Giselle and her colleagues complained on social media and wrote articles about the way the AI search component was affecting their business. It caught Google's attention, and she was invited to attend a web creator conversation event at Google's Mountain View headquarters last fall.
At the beginning, we were all very angry.
At the morning sessions were very upsetting because I think we were just hearing again and again, it's not us me kind of thing. It's not you like, your websites are great, keep doing what you're doing, and we're like, you know that what we do cost money. We can't just keep doing it in the hopes that you're going to get it right.
Navarro says it felt like Google's engineers had been told to respond to any ideas the web creators offered by shooting them down, as did they progress.
I think lots of the creators in the room checked out us, this is a this is just a part exercise.
Julia and Davey spoke with many web creators in their reporting, and they found that Navarro's experience isn't unique.
And most of these websites simply haven't recovered. Some of the site owners that we spoke to have gone out of business and stopped publishing, and you know, sort of the only thing that has changed on Google's side of the equation is their increased focus on AI and generative AI. They really feel like the unspoken compact between them and Google has failed in a lot of ways, and that they can't rely on Google to be a reliable source of traffic anymore.
Apps like Chat GPT haven't had a meaningful impact on Google's market share yet.
But analysts tell us they are beginning to see some signs that chat, GPT and other products are starting to have an effect.
Analysts are forecasting Google's search REVS new growth will slow down. After all, for a company that was built on web search and for whom search remains the most lucrative part of its business, the changes wrought by AI are raising all sorts of existential questions.
We ask the question of executives, what does it mean to Google anymore when you say, oh, just google that or let me google that for you. Everyone knows what that means. But in this sort of age of generative AI, I think Google is thinking about that verb in a different way.
Even In other words, Google may have signed up to explore the AI universe, but it hasn't yet figured out what its role is in the expedition.
One thing that is undeniable that's come out of our reporting is that Google, for better worse, shapes the open Web as we know it, and it also reflects.
The open Web.
And so we are kind of in this cycle where creators of the Web are disincentivized to create, you know, sort of human written content to put out on the Internet that Google would then point traffic to. They're really contending with like a flood of AI slop that is invading the Openweb. And you know, Google's AI also trains on material that is out there, including this AI generated content. And so I think we're kind of in this cycle right now where the Web is slowly degrading.
Google was one of the pioneers in making the Web an integrated part of modern human existence. It's challenge now appears to be how to embrace this new phase of intelligence without deleting the humanity that made it possible in the first place. For more on Google's foray into AI, check out BusinessWeek's cover story searching for a New Google on the Bloomberg terminal Bloomberg dot com, or in the April twenty twenty five issue of Bloomberg BusinessWeek. This is the big take from Bloomberg News. I'm David Gura. This episode is produced by Julia Press. It was edited by Patty Hirsch and Joshua Brustein. It was fact checked by Agrana Tapia and mixed and sound designed by Alex Kura. Our senior producer is Naomi Shaven. Our senior editor is Elizabeth Ponso. Our executive producer is Nicole beemster borg Sage Bauman is Bloomberg's head of podcasts. If you like this episode, make sure to subscribe and review The Big Take wherever you listen to podcasts. It helps people find the show. Thanks for listening. We'll be back tomorrow
