I've experimented with most of the "search chatbots". ChatGTP is by far the most satisfactory.
I wouldn't trust any chatbot for anything important, but most of our searches need to be quick, and "fairly accurate" is good enough.
I've asked questions to which I know the answers. ChatGTP has been close, consistently.
S E E D E D C O N T E N T
Jakub Porzycki/NurPhoto via Getty Images
OpenAI CEO Sam Altman arrived in Washington this week with a carefully crafted message for policymakers: Artificial intelligence is already boosting productivity for millions of Americans, and his company intends to keep it “democratic” by putting it in everyone’s hands. As the capital buzzes with debates over AI regulation, Altman is positioning OpenAI not as a disruptor to be feared, but as an engine for universal progress.
“It’s not about stopping disruption, but putting it into people’s hands so they have the opportunity to benefit,” a source familiar with Altman’s thinking told Axios .
The timing of this pitch could not be more strategic. ChatGPT is now handling a staggering 2.5 billion prompts every day, with 330 million daily queries coming from the U.S. alone, OpenAI told Axios. Just eight months ago, that figure was a billion daily prompts. For perspective, Google processes an estimated 14 to 16 billion daily searches. This means that in less than two years, OpenAI’s conversational AI has grown to handle a volume equivalent to one sixth of the world’s largest search engine.
The Stakes: A Changing Search Habit
For decades, search meant one thing: “Google it.” A new analysis by marketing researcher Rand Fishkin of Datos shows just how ingrained this behavior has been. In 2024, the average active American desktop user performed 126 unique Google searches a month. That includes everything from navigational queries like “Facebook login” to shopping, news, and local lookups.
But AI tools like ChatGPT are starting to chip away at that habit, and not just for power users. A small but growing cohort is using AI as a direct replacement for search engines, asking it to find, summarize, or create answers instead of scanning a list of blue links. Fishkin notes that while most users have not ditched Google yet, the threat is real enough that Google has defensively rolled out its own AI powered “Search Generative Experience” and even a “Web” tab for users who still prefer traditional links.
Why It Matters for Google
Google’s core business is search advertising, which generated $175 billion in revenue last year, accounting for more than half of its total $307 billion in revenue. If even a fraction of high value searches migrate to ChatGPT, Google’s economic engine faces a significant long term risk. The company is spending billions to integrate its own Gemini AI into search, but that strategy carries two major dilemmas:
The Stakes: A Changing Search Habit
For decades, search meant one thing: “Google it.” A new analysis by marketing researcher Rand Fishkin of Datos shows just how ingrained this behavior has been. In 2024, the average active American desktop user performed 126 unique Google searches a month. That includes everything from navigational queries like “Facebook login” to shopping, news, and local lookups.
But AI tools like ChatGPT are starting to chip away at that habit, and not just for power users. A small but growing cohort is using AI as a direct replacement for search engines, asking it to find, summarize, or create answers instead of scanning a list of blue links. Fishkin notes that while most users have not ditched Google yet, the threat is real enough that Google has defensively rolled out its own AI powered “Search Generative Experience” and even a “Web” tab for users who still prefer traditional links.
Why It Matters for Google
Google’s core business is search advertising, which generated $175 billion in revenue last year, accounting for more than half of its total $307 billion in revenue. If even a fraction of high value searches migrate to ChatGPT, Google’s economic engine faces a significant long term risk. The company is spending billions to integrate its own Gemini AI into search, but that strategy carries two major dilemmas:
For consumers: You may get faster, more conversational answers, but at the cost of seeing fewer diverse links and perspectives. AI could centralize information power even more than search engines did.
For creators and businesses: Google’s dominance once meant optimizing for a single algorithm. AI driven search means content could be summarized and stripped of attribution unless strong guardrails are built in. That is a looming threat to publishers already fighting for traffic.
For society: Altman argues democratization is key, asking, it’s who gets how much of a slice of the economic pie? But AI also raises the risk of misinformation, bias, and greater economic concentration in fewer hands.
The Future of Search
We may be watching the first major shift in online behavior since the smartphone. Fishkin remains skeptical that AI will replace Google for most people anytime soon, but he admits the early adopters are showing what is possible. If ChatGPT can handle one sixth of Google’s volume today, what happens when AI native search is built into our phones, cars, and voice assistants?
Google is not going away, but its once unassailable dominance is under pressure for the first time since the days of Yahoo and AltaVista. The fight for the future of search is about whether information online remains open and distributed, or collapses into a handful of powerful AI driven platforms.
OpenAI is probably just as bloodsucker as Google... but if it gives me better, quicker results... it's better.
I've been using Microsoft Bing for my searches, and I think it's pretty good.
I got rid of everything Microsoft some time ago. My OS is Linux Mint, Firefox and Brave are my browsers (Brave can block ads on Youtube
), LibreOffice replaces Microsoft Office (although I also use Google docs a lot, because it's s-o-o-o easy), and so on...
I experimented with Apple a decade ago. It didn't go well, with either the hardware or the software, so that won't happen again.
Google is making two significant moves: combining its to OSes (Android and ChromeOS) into a single one, and creating a "desktop mode" for phones and tablets - that means you just plug your phone into a screen/mouse/keyboard and shazaam, you have a computer.
The PC market today has three parts: enterprise, personal "ordinary", and personal "gamer". Only that last one requires a powerful processor, and probably a dedicated video card. The first two - the vast majority of units - need only a cheap everyday processor, less powerful that what's in your phone. My PC is a little box, about 12 cm on a side and 5 cm tall. It has a middlin' AMD processor. It can do everything I need. My Xiaomi13T phone has about the same power. (Both devices were made in China... OF COURSE.)
So I'm hoping that when I need to trade up for a new phone, it will also be my next PC. How cool is that??!!
Pretty cool.
I'm a failure when it comes to computer technology and even its verbiage.
It's probably because you started in the world of PCs and got used to that, (if your first computer was a MAC you probably would've disliked PCs (Windows).
I believe most people have computers and phones that have more power-- and more features-- than they really need.
I 've heard that now some people are turning away from even owning both a computer and a phone-- they use the phone for everthing they need!
The main thing is just to be able to know a few basic functions you need. For a lot of people that's many reading/writing things online (like NT and other news sources), doing searches, and watching videos.
I think many people don't really need the more powerful-- or more sophisticated programs.
I had a first-generation "Air" - the one Jobs pulled from a simple manila envelope, to the awe of the fanboys. It couldn't drive a decent-sized external screen. And then it just died. Not a great experience.
The OS was "Snow Leopard". Clunky. And of course, the app garden had very high walls.
Windows was more subtle in its self-aggrandizement, back then. A couple years ago, when adver-fucking-tisements appeared in Windows's task bar, I decided to leave.
Linux Mint isn't perfect, but it's at lest as good... and it's MINE!
Yes. For me computers are also toys, like miniature airplanes for some people. I invest far more time and energy than needed.
Have you tried Baidu's Ernie Bot (文心一言, Wenxin Yiyan)? Apparently Alibaba has LLMs that are accessible to Chinese internet users.
China is investing in developing LLM artificial intelligence to compete. Part of the development involves utilizing a broad base of users to refine and improve the capabilities of the LLM. So, you should have access to Chinese LLMs that are comparable to those available in the US.
I have used Baidu search. As to t what I said about computer verbiafe, to me LLM stands for Master of Laws, a step up from the LLB that I have, Bachelor of Laws.
Yep, I run into a lot repurposed acronyms, too. In case of artificial intelligence it's Large Language Model. Tech technology has matured to using specialized esoteric language in the same manner as law. I suppose if the language progresses to esoterically eccentric, the field will become a profession.
Guess I'm lucky. I'm not eccentric, I'm just nuts.
Aren't we all?
I most often use Google Gemini than the other popular LLM models. I tend to use the LLM models for testing speculative hypotheses, challenging conventional wisdom, and developing ideas. I want more than just a more verbose internet search. I tend to use an iterative interactive style to refine prompts and more fully explore information. That's more of a conversation (or debate) than an interrogation.
Gemini, ChatGPT, and Copilot exhibit biases resulting from their training data used. I have asked the LLMs how they were trained and what training data was used. All the LLMs have indicated they were trained on a curated body of information, prioritizing academic and public/private institutional sources, emphasizing consensus agreement on data, facts, and opinions. The training data also included public information from the internet and media sources; however, the greater weighting is given to academic, governmental, and institutional sources of material.
The three LLMs used for passive interrogation (like a search) will respond with what would be considered the 'conventional wisdom' response. It the user accepts the 'conventional wisdom' then they will get responses that affirm that 'conventional wisdom'.