Browsing obsolete: Examining the AI search era

Every week brings a new wave of AI innovation, and search is no exception. So it wasn’t a surprise when Google recently announced AI Mode, a search mode that answers queries with AI responses and curated links.

“We’ve been hearing from power users that they want AI responses for even more of their searches,” Robby Stein, Google Search’s VP of Product, wrote on LinkedIn. AI Mode runs on a custom version of Gemini 2.0. According to Google, this will allow users to ask more nuanced questions, get more links and have their context taken into account.

“I don’t even use Google anymore—I just use ChatGPT,” says Ash Minhas, a Technical Content Manager at IBM. The value, he says, is the ability of ChatGPT to scan and synthesize a vast amount of sources in a short amount of time.

This, of course, marks another shift in the ever-evolving search landscape. Google remains dominant—according to Similarweb, it registered more than 81 billion monthly visits between December 2024 and February 2025, while OpenAI’s ChatGPT saw fewer than 4 billion. The AI search engine Perplexity, over the same period, recorded 101 million visits. Even still, experts see great promise for these LLMs in the years ahead.

“Adoption took off, and it’s only going up. The number of people who know about these tools is growing,” says Ahmed Malik, CEO and Co-Founder of ScalePost, in an interview with IBM Think. “Few things in the world captured corporate and consumer imagination like this,” he adds. Founded in 2024, ScalePost serves as a marketplace between AI companies and content sources. The startup has already begun working with top publishers, helping them facilitate and scale partnerships with AI firms like Perplexity.

Meanwhile, Perplexity remains bullish on building its own search experience. That much is clear with the recent announcement of its upcoming browser, Comet. The goal, according to Perplexity’s CEO, is to "build the future of internet browsing, with AIs doing deep research and tasks for us."

“Who owns the browser? Chrome, Bing, Safari. Perplexity wants to own a channel so they can grow things. That’s where the journey starts,” explains Chris Andrew, whose company Scrunch helps large organizations adapt to AI-powered search, in an interview. “Browsing is inefficient by its definition. It’s only natural you would outsource this to an agent.”

As AI-driven search tools multiply, they introduce a new challenge for publishers and content creators: monetization. Traditional search engines drive traffic to websites, where content creators can make money through ads, subscriptions or lead generation. But with AI agents summarizing information directly, fewer users are clicking through to source pages.

“AI search has broken the Internet. More specifically, it has broken the contract between search engines and content creators [like with] scraping content for free [and] sending referral traffic,” observes Jared Rand, Principal Data Scientist at Everstream Analytics and Founder of Skillenai.

Now that AI search engines provide answers (with citations) instead of a list of links, referral traffic has predictably plummeted.

“Agentic browsing and search exacerbate the problem. In both cases, human eyes are no longer on the publisher’s website,” Rand says. “The trend continues that more people can get more answers to more questions without ever visiting the source.”

AI-powered and agentic search doesn’t send much referral traffic. A first-of-its-kind report from TollBit calculates that, on average, AI chatbots drive 96% less referral traffic than traditional Google search.

“When you do a Google search, you might read one or two websites,” explains Josh Stone, VP of Partnerships at TollBit, a startup that works as an intermediary between content owners and AI companies. “But when you use an AI-powered search tool, it might pull information from 10 or 12 different sources and return a synthesized answer.”

For content creators, this raises concerns about revenue. AI agents don’t engage with ads or sign up for subscriptions. But it’s also a challenge for agent builders—large-scale web scraping isn’t free.

“Web scraping isn’t free, it’s actually relatively comparable to paying for content,” says Olivia Joslin, who co-founded TollBit with Toshit Panigrahi. “A lot of the companies building AI agents are trying to automate workflows. They don’t want risk, and they’d rather pay for the content,” she adds. “They’re not playing litigation roulette.”

Another issue is the accessibility of content through AI tools. Minhas notes that even if he uses ChatGPT for quick or deep search, he also sees the lack of ability to use specific sources (such as paywalled news publishers, for example) as a limit. “How do I know it’s true or not? If it can’t scan sources, is it scanning junk?” asks Minhas. “What’s going to end up happening is more deals made between AI frontier models and data sources.”

Similarly, having a human interact on a website could become a premium in the future, says Chris Hay, a Distinguished Engineer at IBM. “But at the same time, businesses are going to start thinking about how they make sure their content is consumed by the AI.”

Looking ahead, Panigrahi believes the need to provide agents with reliable and deep data will be greater than ever. “Tesla, Amazon, Meta: all the big companies are working on humanoid robots. All of them will have almost a GPT-power consciousness that connects to the internet. They will also need data access. This is the future we're trying to prepare for.”