The internet is moving from being search-driven to answer-driven.
In the current era of AI-powered search, publishers are watching while their content is scraped and paraphrased by large language models, users hold concerns about not being able to easily opt out of AI features on large platforms, and a couple of Big Tech companies stand to benefit the most from this new paradigm.
A common search on Google means coming face-to-face with its AI Overviews, which populate the top of the search page with a summary of key information about the topic searched and links to where the answers came from.
A year after the initial launch of AI Overviews in May 2024, Google said the feature drove a 10%-plus increase in usage of the search engine in its biggest markets, like the U.S. and India, for queries that show AI Overviews.
This is good news for the company. Google claims that thanks to its generative AI features, people are more satisfied with search and using it more often. Along with fellow tech giants such as Microsoft and Meta Platforms, the company has spent the past few years since the public launch of OpenAI’s ChatGPT aggressively embedding artificial intelligence features across most of its services.
However, these same platforms have faced criticism from users and industry peers alike for what many say has felt like a heavy-handed approach.
“Forcing a publisher to consent or forcing a user to consent was not how we wanted to implement AI,” says Kamyl Bazbaz, chief communications and policy officer for DuckDuckGo, one of Google’s biggest search competitors, which stands apart from many of today’s tech platforms by offering an AI-free version.
In May, a week after Google announced an upgrade to its search with Gemini 3.5 Flash, DuckDuckGo received an increase in U.S. installs by 30% week over week.
Bazbaz says DuckDuckGo’s install levels are still about 30% above where they were before Google’s announcement.
“Some folks sort of thought, This is enough for me. I don’t like the direction this is going. It’s making me uncomfortable. I feel like I have no control,” he says.
Bazbaz believes what prompted people to switch to DuckDuckGo is its headline approach of being private, useful, and AI optional.
Fast Company reached out to Google, which shared information about its generative AI search features and its updates. Google said the decisions that websites make about participating in its AI features will be respected, but it also noted that people are increasingly gravitating to generative AI tools to aid them in finding and understanding information.
The company did not provide specifics about what opt-out features might be available for users of its AI tools in the future.
According to Bazbaz, user surveys show that a large percentage of people who switched over to DuckDuckGo in the last month or so are mostly dissatisfied with Google’s search results. Another large percentage are dissatisfied with the lack of control in AI.
These dissatisfied users are joined by another critic of Google’s AI search features—regulators in the U.K.
On May 30, in what it calls a “world first,” the country’s Competition and Markets Authority (CMA) required Google to allow U.K. publishers to opt out of their content being used to power its search engine’s AI features, and to make sure all content is properly attributed.
“With features like AI Overviews rapidly reshaping online search, it is crucial that content publishers, including news organisations, have appropriate bargaining power over how their content is used,” Sarah Cardell, chief executive of the CMA, said in a statement.
It remains to be seen how effective the CMA’s regulations will end up being. So far, Google has complied, saying it will engage with the CMA “to ensure website owners have the right tools as user preferences evolve.”
Further action was taken by the CMA on June 17, when it introduced two new conduct requirements for Google Search: Google must improve fairness in how search results are ranked and allow users to port their search data to authorized third parties.
What do people actually want in regulations?
While moves for regulation are being made in the U.K., on the other side of the pond, the U.S. has yet to see federal AI regulations. But as AI usage grows daily, more and more Americans say they want it.
A new survey by researchers at Johns Hopkins University found that most U.S. adults strongly support regulations on the technology. In April and May 2026, more than 2,000 U.S. adults were polled—with those who trust Al “a great deal,” “not at all,” or “somewhere in between” all in support of Al regulation at approximately the same level.
U.S. adults are also concerned with how true the information is that AI provides them, and many hold skepticism toward what AI serves up.
When asked how much they would trust AI to look up factual information, 41% of U.S. adults said “somewhat,” and 18% said “not at all.”
These feelings are not new. Concerns over AI in daily life have increased among U.S. adults, based on a Pew Research survey conducted in June 2025, in which 50% of those polled said they were “more concerned than excited.” That number grew from a 2021 poll, when only 37% were “more concerned than excited” about AI use in daily life.
It’s fair to say that U.S. adults would like to see AI regulations applied to search engines. But what would the path to federal regulations on AI search look like?
Pete Pachal, founder and editor-in-chief of The Media Copilot, a publication that reports on AI, is very cautious about attempts to regulate an economy of AI consumption.
Pachal (whose work is also featured on Fast Company) believes overregulation can sometimes do more harm than good, particularly in terms of competition.
Burdensome rules would pose a particular threat to smaller AI startups that might be innovating in the space, he notes, while Big Tech companies like Google and Meta would have the compliance resources to pull out even further ahead.
“It’s true of a lot of regulations,” Pachal says. “They will favor incumbents and larger players. So I think the right balance here is to try to get what is the simplest thing you could do that would help everybody that doesn’t put a bunch of clamps on a system that makes it immobile.”
Pachal likes the idea of requiring AI bots to identify themselves and be as transparent as possible. Identifying bots, he says, might keep the government away from making value judgments on content more broadly.
According to Pachal, online traffic will be primarily AI agent-driven in 5 to 10 years, with people most often talking to machines for answers—their phones, computers, or whatever else is cooking up in labs in Silicon Valley.
“We’re just talking to them, and they’re just giving us answers, and it’s intuiting things from whatever data we give it, and that just seems kind of obvious to me,” Pachal says.
As a self-described “power user of AI,” Pachal is already personally moving this way, talking more to his computer these days than typing.
Tech giants have more data than regulators
Back at DuckDuckGo, Bazbaz calls the CMA’s ruling the tip of the spear for Google. Defining the company as a “monopoly,” he sees the ruling leveling the playing field. He says it’s always a good thing to see regulators take action with specificity and conviction.
“A federal judge in the United States and many regulators all around the world, including those in Europe, have found Google to have engaged in anti-competitive behavior, and they are using a lot of the same tactics and tricks that they did in this last era to maintain this dominance in the AI era,” Bazbaz says.
Still, he understands that regulations can be tricky, particularly as regulators often don’t have the full picture of a given situation.
“Google will always have more information than the regulator. They’ll always know more because it’s their stuff,” Bazbaz says. “That power imbalance even exists in the regulatory process.”
His solution begins with a ban on surveillance or behavioral advertising in the context of AI. Referencing philosopher Shoshana Zuboff’s work on surveillance capitalism, Bazbaz touches on surveillance advertising being the original sin of this tech era. He believes part of the problem is the extent to which people naturally share more with AI than they do with a regular search.
“If you think about the surveillance economy and the extraction of your data that mines what you might do in the future and is able to sell you something in the future, [that’s] a cornerstone of all the problems,” Bazbaz says. “It’s sort of the original catalyst.”
A feasible solution
Senior attorney Jason Henderson says regulation is going to move at the pace that it tends to move, which is slow.
Henderson, a lawyer with an expertise in sports, entertainment, and streaming media, says requirements like the CMA’s are going to happen everywhere at some point, but it may not matter because publishers are going to do deals.
He believes the open web is dying before our eyes, because when people use AI search—across Google and Microsoft’s Bing—and receive an answer, they don’t click through to the original article.
Henderson notes that publishers are losing money as 68% of searches now end without a single click to the original source.
“They take the benefit of that content, but they don’t actually look at the original, so they don’t see any of those ads,” he says, referring to the common AI search user.
When publishers look to the government for what they should do to address their content being scraped, Henderson says there are a couple of approaches to take.
Publishers can ask for a chance to opt out of being scraped, allowing users to visit their sites directly. Or they can pursue licensing deals with AI companies, allowing them to scrape their content in exchange for paying them a license.
“The last [option] is you can sue,” Henderson says. “And all of those are happening, and they’re all happening simultaneously.”
He cites The New York Times establishing a licensing deal with Amazon versus filing lawsuits against OpenAI and Perplexity. At the end of the day, he says it’s just about the market. While publishers still own their content, they need to figure out by any means necessary how to get paid for it.
“Ultimately, what this means is the gold rush period where people were just scraping data willy-nilly and not having to pay for it is coming to an end,” Henderson says, “because a small provider might not have the heft to be able to demand to get paid, but a large provider absolutely would.”
The next big trend for publishing, Henderson says, will include licenses for data training. He cites the American Society of Composers, Authors and Publishers’ music licensing as an example. A business pays an annual license fee to ASCAP to receive access to its expansive music catalog. This license agreement gives a business permission to play music from any ASCAP member. Part of the license fee goes back to the ASCAP member as royalties, thereby paying them for their content to be used.
“I think that there will be something like that, but it will not be for reads or listens,” Henderson says. “It’ll be for scrapes.”