I Used AI Chatbots for a Week to Answer Common Questions—Here’s What Happened

When AI chatbots like ChatGPT first launched, many of us were skeptical of their accuracy. Asking AI questions was more of a party trick than a legitimate use case. Since it’s been a couple of years of steady improvements, the question arises: Can we finally replace manual search with AI?

AI Is Faster and More Precise Than Google

When I say “precise,” I mean that chatbots rarely lose focus, regardless of how weird or specific your question is. Traditional web search engines like Google and DuckDuckGo process your query, identify keywords, and output a results page that’s been sorted by relevance with the help of an algorithm.

The page results can (and usually are) highly relevant to your search query, but if you don’t use the right keywords or there’s nothing on the topic, you’ll just get a random assortment of search results with some of your keywords thrown into the mix. If you want to find the answer, you’ll have to dig deep and try a few different search queries.

In contrast, AI chatbots take only 10–15 seconds to answer your prompt. More importantly, they always stay on topic, even if they have to guess the answer by offering multiple possibilities. This is because a primary goal of LLMs is to understand the context of your prompt and attempt to interpret the prompt by any means necessary. Here’s an example of a highly specific question that I couldn’t find the answer to using Google, yet ChatGPT provided some plausible answers:

When I was planning to buy my new headphones recently, I checked lots of reviews online, but since I was doing this little experiment, I also figured it wouldn’t hurt to ask ChatGPT what it thinks of the Sennheiser HD 599. To my surprise, the mini-review was rather thorough, and I also found it to be accurate after testing the headphones myself. My only gripe is that it slightly exaggerated the “amp dependency” aspect, as these headphones work quite well without one, though an amp can still be beneficial.

In the same vein, I discovered that my new headphones’ headband is made of protein leather, a material that tends to peel off and flake after a couple of years. I didn’t want that to happen with my headphones, so I looked into ways to clean and maintain my headband. Google wasn’t of any help, but ChatGPT understood my question and gave me some tips that sounded legitimate:

Another benefit of AI chatbots and AI search engines is their ability to scour the web to find extremely specific information. You’d think that a search engine would be better at this, but they’re mostly limited to what you type in. At best, search engines give you articles and forum posts that discuss the topic, whereas a chatbot can “understand” your query and provide you with exactly what you’re looking for. Here’s an example using Copilot:

You Can Never Trust AI

The main issue with AI chatbots is that they’re highly unreliable, as they tend to hallucinate. In simple terms, this means that they occasionally provide a confident yet incorrect answer to your question.

Keen-eyed readers might have noticed in my example using SUVs above that not all vehicle lengths are accurate. While the BMW X2 and Mitsubishi Outlander Sport are SUVs that don’t exceed 185 inches in length, the dimensions that Copilot provides are incorrect. Technically, the Nissan Rogue is also off by 0.1 inches, but we can give it a pass.

I’ve noticed that the lengths that Copilot provided aren’t entirely random, either. After doing some research, I realized that Copilot drew data from tables and charts that compared these models against other SUVs. For instance, the X2’s apparent 174.2 inch length is actually taken from the Volvo XC40.

Should You Use AI for Everyday Questions?

You can use AI chatbots to ask routine or random trivia questions, such as “How to use the INDIRECT function in Excel?” or “Why do penguins have tuxedos?” For the most part, they’ll provide you with detailed answers quickly and with a high degree of accuracy, regardless of the uniqueness of your question. Maintain a healthy dose of skepticism, and if it isn’t something important, then it doesn’t really matter if the chatbot gets it wrong.

While AI chatbots can be fun tools to toy around with, you should never trust their answers, especially when it comes to critical areas like medical, legal, or financial advice. In fact, you shouldn’t trust search engines, either. Always consult educated professionals who can provide personalized guidance or use credible sources of information.