Not long ago, searching for a recipe online meant typing “best chocolate chip cookies” into a search bar and getting a long list of blue links. You’d click a few, dodge some pop-up ads, scroll past a lengthy personal story, and eventually find the ingredients. Today, you’re just as likely to see a complete recipe—ingredients, steps, and cooking times—summarized neatly in a box at the very top of the page, long before you see the first blue link.
This is more than just a convenient new feature; it’s a fundamental change in how the internet works. That box is an AI-generated answer, and it represents a massive shift from an internet of links to an internet of answers. This article will explain what Large Language Models (LLMs) and Google’s AI Overviews are, how they find information, and why this transformation matters for everyone who uses the internet—from casual searchers to the websites and creators who have built the digital world we know.
1. What Are These “AI Brains”? A Simple Guide
At the heart of this change are powerful new artificial intelligence systems. To understand their impact, we first need to understand what they are.
1.1. Meet the Large Language Models (LLMs)
A Large Language Model, or LLM, is a type of AI trained on a colossal amount of text and data from the internet. You can think of them as incredibly advanced prediction machines. According to expert Barry Adams, LLMs are essentially “advanced word predictors.” Their fundamental job isn’t to understand information in the human sense, but to produce a “reasonable continuation” of any text they’re given. They are statistical models that have learned the patterns of human language on a massive scale.
Several major players have emerged in the LLM space, each with its own model:
- ChatGPT from OpenAI
- Gemini from Google
- Claude from Anthropic
- Perplexity AI from Perplexity
These are the engines powering the new wave of conversational AI chatbots and search features.
1.2. Google’s New Feature: AI Overviews
An AI Overview is Google’s version of an AI-generated summary that appears at the top of many search results pages. Its goal is to provide a direct, synthesized answer to your query by pulling information from multiple websites and combining it into a single, easy-to-read block.
This feature is part of a larger strategic shift at Google called “AI Mode,” which the company sees as the future of search. The aim is to make finding information more conversational and direct, moving away from the traditional list of links and toward providing immediate answers.
Now that we know what these tools are, let’s pull back the curtain on how they gather their information and generate answers.
2. How Do They “Learn” and “Answer”?
LLMs aren’t born with knowledge; they are trained on data collected from the internet. The methods they use to gather this data are both powerful and controversial.
2.1. The Internet’s Biggest Reader: Understanding Web Scraping
The process AI bots use to read and collect massive amounts of data from websites is called web scraping. These bots systematically “crawl” the web, downloading text, images, and other information to use as training material.
The scale of this scraping can be staggering and places a significant burden on website owners. The experience of the tech review site Trusted Reviews provides a stark example. The site discovered it was being hit by OpenAI’s web crawler with:
- 1.6 million scrapes in a single day
- An astonishing rate of 18.5 scrapes every second
This firehose of automated traffic put a strain on the site’s servers and degraded the experience for human visitors. And for all that data taken, the return was minuscule. The 1.6 million scrapes resulted in:
- Just 603 users sent to the site
- A click-through rate (CTR) of only 0.037%
Furthermore, the traffic that did arrive was of low quality, with users spending 58% less time on the site and viewing 10% fewer pages than the average user. This highlights a growing conflict over the “fair value exchange” online. AI companies benefit from the content, while the websites that create it are often left with the costs and very little traffic in return.
2.2. The Good, the Bad, and the Made-Up: Citations vs. Hallucinations
When an LLM provides an answer, the information it presents can fall into two very different categories: verifiable facts linked to sources, or confident-sounding fabrications.
| Citations (The Good) | Hallucinations (The Bad and the Made-Up) |
| These are links to the original websites where the AI found the information it used in its summary. For content creators, this is the best-case scenario, as it can still drive some traffic. | Because LLMs are “advanced word predictors,” they can, as expert Barry Adams notes, “make stuff up and very confidently present it to you.” This is a hallucination. It’s a fundamental weakness because the AI doesn’t understand facts; it only predicts the next most likely word in a sequence. |
| For users, citations provide a crucial way to verify the AI’s claims and dig deeper into a topic by visiting the primary source. | An AI might invent facts, misattribute quotes, or create fake sources because those words are statistically probable, not because they are true. This makes trusting an AI’s output without verification a risky proposition. |
With these powerful (and sometimes flawed) abilities, are people abandoning traditional search engines for AI chatbots?
3. Is AI Replacing Google? The Surprising Answer
A common question is whether AI chatbots like ChatGPT will make traditional search engines obsolete. The data so far points to a more complex reality.
3.1. More Tools, Not Fewer Searches
According to a major clickstream data study by Semrush, the evidence supports a theory known as the Expansion Hypothesis, which posits that ChatGPT adoption does not reduce Google Search usage. Instead of substituting one for the other, people are simply using both, expanding their overall information-seeking behavior. This suggests users are treating AI chatbots and traditional search as complementary tools, likely using conversational AI for complex idea generation and Google for quick verification or navigational searches.
3.2. What We’re Really Using AI For
Studies of user behavior on platforms like OpenAI’s ChatGPT and Anthropic’s Claude reveal clear patterns in how people are incorporating these tools into their daily lives. The top three use cases account for the vast majority of conversations.
- Practical Guidance: Asking for advice on how to do something, from planning a trip to fixing a leaky faucet.
- Seeking Information: Using the AI as an advisor to learn about a topic, similar to a traditional informational search.
- Writing: The most common work-related task, primarily involving requests to edit or modify existing text, like emails, rather than generating new content from scratch.
These uses can be grouped into two new user intent categories that define the AI era:
- Asking for information or advice (49% of use)
- Doing a specific task, like writing or coding (40% of use)
While people are using both AI and search, the way AI presents information is causing a major shake-up for online publishers and creators.
4. The Big Shift: What This Means for Websites and Creators
The move from a list of links to a single AI-generated answer is creating a new set of rules for anyone who publishes content online.
4.1. The Disappearing Clicks
The most immediate and significant impact of AI Overviews on websites is traffic loss. A groundbreaking user experience (UX) study conducted by Kevin Indig found that when an AI Overview appears:
- Outbound clicks on desktop can fall by as much as two-thirds.
Furthermore, the study revealed that most users never even see the sources cited at the bottom of the answer. The median scroll depth was only 30%, meaning the majority of users only read the top third of an AI Overview. This makes it incredibly difficult for websites cited lower down in the summary to ever be seen, let alone clicked. Notably, the study found that the small number of users who do scroll further are also the ones who report higher trust in the AI’s answer, suggesting that visible, authoritative citations play a key role in building confidence.
4.2. A New Goal: From #1 Rank to AI Citation
For decades, the holy grail of Search Engine Optimization (SEO) was achieving the #1 blue link ranking. That is no longer the primary goal. In the new landscape, the objective is to become such an authoritative source that your content is used and, crucially, cited within an AI Overview.
As news SEO expert John Shehata explains, the key metric is shifting from “traffic” to “visibility.” Being mentioned in an AI answer is the new form of top-ranking placement, even if it doesn’t result in a direct click.
4.3. Why Trust and Brand Matter More Than Ever
In a world where AI can confidently “hallucinate” incorrect information, users are becoming more skeptical. The UX study revealed that users now apply a two-step mental filter when looking at search results:
- “Do I trust this brand?” (Is the source cited a name I recognize and believe to be credible?)
- “Does this result answer my question?” (Only after trust is established do they evaluate the information itself.)
When faced with a block of AI-generated text, users actively look for familiar, authoritative names cited as sources—like government (.gov) sites, academic institutions, or well-known publications—to feel confident in the answer. This means that building a strong, trustworthy brand that people recognize and actively search for is becoming one of the most important survival strategies for creators and publishers.
5. Conclusion: Navigating the New Digital World
The rise of generative AI is not just another update; it is a fundamental restructuring of how we access information online. The landscape is shifting in three critical ways:
- The internet experience is moving from navigating a list of links to receiving direct answers.
- The goal for content creators is evolving from driving traffic to earning visibility and citations within AI summaries.
- The most important factor for users is no longer just the relevance of the information, but the brand trust of its source.
This change is an evolution, not an endpoint. The rules are still being written, and the technology is constantly improving. For students, creators, and consumers alike, understanding this new landscape is the essential first step to adapting and thriving in the next era of the internet.
