The Rise of AI Search: Implications for Information Markets and Human Judgement at Scale
Based on a massive global study of 2.8 million search results, this paper reveals that the rapid expansion of AI search is driven by opaque corporate policies that significantly reduce source diversity and credibility while increasing political bias, thereby necessitating urgent global debate on the resulting economic and societal impacts.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the internet as a massive, bustling library. For decades, if you wanted to find a book, you asked the librarian (Google), and they handed you a list of cards (search results). You had to walk over, pick a few cards, read the titles, and decide which book to open. You were the judge, the jury, and the explorer.
Now, a new kind of librarian has arrived: The AI Search Engine. Instead of handing you a list of cards, this new librarian walks up to you, grabs a book, reads a chapter, summarizes it, and hands you a single, polished note saying, "Here is the answer."
This paper by researchers at MIT is a report card on this new librarian. They tested it 24,000 times across 243 countries to see how it's changing the way we find information. Here is what they found, explained simply:
1. The Librarian is Taking Over (Fast)
In 2024, this "AI Librarian" was only working in a few branches of the library (mostly the US and UK). By 2025, it had expanded to almost every branch in the world (229 countries).
- The Surprise: Some countries, like France, Turkey, and China, were told, "You can't have this new librarian yet." But for everyone else, the change was sudden.
- The Political Switch: The researchers found a weird switch. In 2024, if you asked about "Covid," the AI almost never answered (only 1% of the time). But in 2025, after a change in US government policy, the AI started answering 66% of Covid questions. It's like the librarian suddenly decided to start talking about a topic they were previously forbidden to discuss.
2. The "Zero-Click" Problem: We Are Stopping Reading
When the old librarian gave you a list of cards, you clicked on them to read the real stories.
- The New Habit: Now, because the AI gives you a summary right away, people are clicking on the original sources half as often.
- The Analogy: Imagine you go to a restaurant. The old way was: You look at the menu, pick a dish, and go to the kitchen to see how it's made. The new way is: The chef hands you a plate of food and says, "Here is your dinner." You eat it, but you never go to the kitchen.
- The Danger: The "kitchen" (journalists, researchers, and experts) gets paid based on people visiting. If no one visits the kitchen anymore, they can't afford to cook. This could kill the production of new, original information.
3. The "Echo Chamber" Effect
The old librarian gave you a list of 10 different books from 10 different authors. You might see a conservative view, a liberal view, and a scientific view.
- The New Voice: The AI librarian speaks with one voice. It synthesizes everything into a single answer.
- The Result: The researchers found that AI answers are less diverse. They rely heavily on a few big websites (like Wikipedia or Reddit) and ignore the "long tail" of smaller, niche experts.
- The Bias: The AI seems to lean more toward "Right" and "Center" political views and uses sources that are less credible than the ones traditional search uses. It's like the librarian only letting you read books from one specific section of the library, even if other sections have better answers.
4. The "Confidently Wrong" Trap
This is the scariest part. The AI librarian is very confident, even when it's lying.
- The Hallucination: Sometimes the AI makes things up (like telling people to eat rocks for nutrients).
- The Trust Trap: The study found that when the AI adds a little footnote or a link to a source, people trust it even more, even if that link is fake or wrong. It's like a magician who pulls a rabbit out of a hat and points to a picture of a rabbit on the wall to prove it's real. You believe the trick because of the "proof," even though the proof is fake.
- Who Gets Hurt: People with less education or less tech experience are the most likely to believe these confident lies.
5. What Should We Do?
The authors say we can't stop the AI librarian; it's already here. But we need to set some rules so the library doesn't burn down.
- For the Tech Companies (The Librarians): They need to show their work. Every claim should have a clickable link to the exact sentence it came from. If two experts disagree, the AI should say, "Some say X, others say Y," instead of just picking one side. They also need to pay the original writers for the information they use.
- For the Government (The Library Board): They need to force the companies to be transparent. They should demand reports on how often the AI is wrong and which countries are getting which answers.
- For Us (The Readers): We need to change our habits. The authors suggest a "Two-Click Rule." If you are making a big decision (like a health choice or a financial investment), don't just trust the AI summary. You must click through to at least two original sources to verify the story.
The Bottom Line
The internet used to be a place where we learned to navigate and find our own answers. Now, it's becoming a place where we are fed answers. This is faster and easier, but it's also making us lazy, less diverse in our thinking, and more vulnerable to being tricked. The researchers are asking us to wake up before the "AI Librarian" becomes the only voice we hear.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.