Will ChatGPT replace Google?

ChatGPT burst into the scene as the new kid on the block circa 2022 and, since then, has rapidly established itself as a query-answer solution, enough to chip at Google’s dominance.

Google processes 5 trillion searches annually — 14 billion daily searches — according to an article published earlier this year by the company describing the reach of its tools.

The battle for the web search space

In recent months, however, Google has seen a decrease in its searches over Apple’s native browser Safari, per The Wall Street Journal.

The two Silicon Valley rivals have a deal that balances competition and collaboration. As reported by Bloomberg, Google paid Apple $20 billion in 2022 to remain the default search engine in Safari, court documents showed.

Since 2020, the Department of Justice has been pursuing two separate antitrust lawsuits aimed to break up Google’s monopoly on web searches and digital advertising.

On the penalty trial phase of the most recent lawsuit, Apple executive Eddie Cue attributed the decrease in Google searches to the rise in popularity of AI services such as ChatGPT, according to The Wall Street Journal.

ChatGPT’s growth

With over a billion queries per day, ChatGPT processes only a fraction of Google’s daily inquiries — no small feat, given the recent boom of the Large Language Model in the global digital space.

Since its release, ChatGPT’s user base has been growing steadily in leaps and bounds with an estimate of 800 millions users, per Forbes, and $12 billion in annualized revenue, per Reuters.

In March of this year, OpenAI announced two-month free access for students in the U.S. and Canada, which only increased its traffic.

For some, ChatGPT has become the new Google, but looking at how these technologies differ is essential to understand the tug-of-war for a market share.

Search engine model vs. large language model

Even though it might seem like ChatGPT and Google Search serve the same purpose — to provide information — they are fundamentally different at their core. How they arrive to that information is the key differentiator.

Google/search engines

Search engines such as Google function as an entry-point to the internet.

According to Google’s guide on how search works, Google Search follows a three-step process to deliver a result for a user’s query — crawling, indexing and serving.

First, Google Search surveys the internet through “crawlers” that find and download web pages, which are then indexed and added to Google’s database, i.e., a directory.

For Google to present a result, the website page has to exist and be indexable. Google is constantly crawling the web to update its directory. Most of the time, these new websites are linked through websites that Google has already added.

Once indexed, Google proceeds to “serve” and rank the websites according to the relevance to the user’s query. In addition to relevance, there are many other metrics that determine where a webpage is ranked by Google, some of which include the quality of the content, location, device type, language, etc.

Most search engines work more or less in a similar fashion, each one using their own algorithms and criteria to determine how the results are shown.

It is a matching game. When Google receives a query, it will look for the pages that are the most trustworthy, highest quality and most recent.

Put simply — it regurgitates, it does not generate.

ChatGPT/LLMs

Large language models, or LLMs, such as ChatGPT are pieces of software that attempt to emulate human-like text. They are massive in scale, and very sophisticated.

LLMs are trained using vast amounts of information that allow them to make connections within the data and, thus, predict the next word in a sequence. However, not all language models are generative.

In the case of ChatGPT, it uses parameters to, first, understand and, then, reproduce identified patterns from its dataset to generate human-like text. This is called content awareness.

ChatGPT’s capability to analyze relationships within the data is similar to a teacher’s ability, after extensive training, to explain concepts and extrapolate ideas “without reproducing the original materials verbatim,” parent company OpenAI said in an article about how its LLM works.

According to the same article, ChatGPT does not keep the data it was trained on. Rather, it uses the data to adjust its built-in parameters and improve its predictions. That’s how the model is able to keep “learning” and teach itself. Feedback makes it better, sharper.

Put simply — it generates, it doesn’t regurgitate.

Where ChatGPT and Google Search meet

Despite these inherent differences, the overlap lies in the users’ needs — answers, responses, solutions, explanations, etc.

The main difference is that ChatGPT gives you a textbook-ready answer, whereas Google shows where to look for an answer.

According to OpenAI, ChatGPT can assist with a variety of tasks including simple queries, all stages of the writing process from brainstorming to rewriting, creative suggestions, logical reasoning problems and language translation.

Differences

Both tools will provide information, but the information will be different. The reason why generative LLMs are known for "hallucinations" is because the LLM will try to predict what the next word is based on the training data it already has. Another feature that makes LLMs like ChatGPT appealing is their conversational nature, which has prompted people to turn to it as a step-in therapist.

Search engines are not free from faults. Google’s ranking system has been impacted by the ads that crowd the online real estate. Nowadays, the first search option often will be connected to an advertiser that paid to be there in the first place.

It is important to note that Google searches are more electricity efficient. Each ChatGPT query uses about 10 times more electricity than a traditional Google search, according to Euronews.

Google’s response

Google has been on the offense. It rolled out its revamped AI model, Gemini, at the end of 2023 and it followed this development with the implementation of AI overviews on its web searches the next year.

The web search space has been dominated by Google, but as recent tech history shows, there’s no guarantee that today’s technology won’t become tomorrow’s memory.

Where the race is headed

Before ChatGPT, Google — so predominant that it became a verb — used to be the go-to for answers to all kinds of questions (it still is). For the last decade, Google’s market share of global searches has been sitting comfortably above an astonishing 90%.

But the current user demographics are shifting toward AI-powered searches, and it’s making a dent in Google’s dominance. At the end of 2024, its shares fell below 90%, over a year after hitting an all-time high of 93%, according to Statcounter data. One has to go back 10 years to find the last time Google’s shares were below 90%.

On top of that, Google stock went down more than 7% on May 7, The Wall Street Journal reported.

These fissures are showing an unforeseen vulnerability in the tech giant, further echoed during Cue’s testimony in the penalty trial phase of the DOJ antitrust trials. The Apple executive acknowledged Apple is exploring search alternatives among AI industry leaders, including OpenAI, per Bloomberg.

However, Sam Altman, OpenAI’s CEO, stated during a U.S. Senate subcommittee hearing on AI in May that, despite some advantages of ChatGPT, it will “probably not” replace Google.

As more players have entered the AI race in the last year, it remains to be seen how it will play out.