Can ChatGPT Replace Search Engines in 2025?

Can ChatGPT Replace Search Engines? A Comprehensive Review
The emergence of advanced AI language models like ChatGPT has sparked an intriguing debate about the future of search engines. With its ability to generate human-like responses, some experts speculate that ChatGPT could overshadow traditional search engines by 2025. In this article, we will explore the strengths and weaknesses of ChatGPT in comparison to conventional search engines, and examine whether AI-driven conversational search could indeed be the way forward. Join us on this journey to uncover whether ChatGPT is poised to replace traditional search engines.

The Experiment: ChatGPT as My Default Search Engine

To evaluate ChatGPT’s capabilities, I decided to integrate it into my daily workflow as my primary search engine. After several weeks of testing, my experiences yielded mixed results. While satisfactory in some instances, ChatGPT hasn’t proven itself capable of replacing established search engines completely. Here’s a detailed breakdown of my findings.

User Experience

Using ChatGPT for searches is remarkably easy due to its integration across platforms, including web, mobile, and desktop applications.

Search Functionality: ChatGPT search is enabled for the 4o and 4o-mini models, automatically triggering searches when necessary or allowing manual searches through a dedicated button.

Sourcing: You can easily explore sources; on the web, results are displayed in a sidebar, which remains open for subsequent queries until you choose to close it. This feature adds convenience by keeping relevant information accessible.

Chat History: While seamless in pulling information, starting a new chat for every query can clutter your chat history. An option to consolidate queries could significantly enhance the user experience.

Performance

ChatGPT typically delivers responses swiftly, with only occasional delays. However, I did experience one instance where the response time raised concerns about functionality. In contrast, utilizing other AI platforms often yielded faster responses for similar queries.

Real-Time Information Capability

ChatGPT shines when dealing with general queries but struggles with real-time information. For example, I tested it by inquiring about a concert that had already concluded. While ChatGPT provided details about the event’s location and future shows, it failed to mention that tickets were sold out, which hindered my planning.

Contextual Awareness: ChatGPT retains context effectively, allowing follow-up questions. However, its results can sometimes miss the mark, such as suggesting dining options that are 30 minutes away instead of local choices.

Outdated Information: In another instance, ChatGPT failed to provide real-time updates on a service outage, claiming everything was functioning normally when it wasn’t. Competing AI models delivered accurate information on their first attempt, emphasizing the need for improvement in ChatGPT’s data accuracy.

ChatGPT, with its intelligence, should have performed better at this job. Perhaps, the fact it’s relying entirely on sources is the reason for the botched job. But OpenAI definitely needs to improve this aspect.

Also, it did fail to provide up-to-date information once; Canva was recently experiencing an outage and when I searched for it, ChatGPT told me it was working fine.

I went to Canva’s website it linked to check the status and the website clearly mentioned there was an outage.

It took me multiple attempts to get ChatGPT to correct itself. Copilot and Perplexity AI, however, got it right on their first try.

Results contain mostly text

With traditional search, we’ve come to expect images and videos as part of the search results. For a lot of people, videos are a better source of information than reading huge chunks of text. But if you’re using ChatGPT Search, chunks of text are what you’ll mostly get.

Even though ChatGPT is capable of returning videos and images, it’s a rarity. And it isn’t even capable of returning other rich results like Google. If you’re searching for a restaurant nearby, for example, Google returns all the right results – the location, timings, reviews, opening hours, etc. along with the other links.

In the case of ChatGPT however, it ended up giving me information about the other chains of the restaurant in faraway cities, without any mention of the one in my own city.

I would like my search engine to present the right information on the first try as would most of us in this busy world. ChatGPT does have access to location information since it uses it for weather; it needs to put it to other use as well.

Cleaner Interface

One thing with ChatGPT’s response is that there’s virtually no clutter at all. There are no ads, no related questions, and no other information overload. For some people, that’ll be a great experience.

For others, not so much. While everyone would certainly appreciate the lack of ads, I’m not so sure about the lack of related questions. Many times, when I am trying to research something (especially when I’m not sure where to start), it’s hopping from one topic to another related topic that points me in the right direction. With ChatGPT, I hit a wall. If nothing else, ChatGPT should perhaps have some follow-up questions, as Perplexity AI does.

Limits and Restrictions

Alas, ChatGPT search, much like the chatbot, has a lot of restrictions and won’t give you a response for a lot of queries (mainly NSFW). If it wants to be a search engine in its own right, it cannot afford to engage in moral policing.

People aren’t interested in a search engine that tells them what’s right or wrong (when it’s not downright refusing to respond, that is). People want their search engine to come back with the information if it exists. Plain and simple.

Availability

One of the things that putting ChatGPT at a disadvantage right now is that it’s not available for free users. Even when it will be available for free users, there is talk of the number of queries being limited by ChatGPT 4o’s usage limits (according to your plan).

Sometimes, when you’re on a search spree, you’re putting in one query after the other. How would this work with usage limits? I haven’t yet encountered this problem, but I didn’t also go on one such “spree”. And if I ultimately decide to not pay for Plus anymore, would I still not encounter this problem on the free plan (given that Search has rolled out to free users by then)?

That’s something OpenAI needs to get on top of if it wants to become a dominant player in the search field.

💡 Meanwhile, if you’re a free user and experiencing FOMO, here’s something for you.

So I obviously had a thought of comparing ChatGPT search with Copilot, given OpenAI’s and Microsoft’s entangled history. I found a really interesting thing. ChatGPT and Copilot gave exact word-to-word responses for multiple queries.

I ran multiple tests; for some queries, the responses matched more than 90% but never less than this.

So, if any free users are experiencing FOMO on OpenAI’s delay in releasing ChatGPT search for you, head over to Copilot.

The verdict

ChatGPT Search is a move in the right direction, but for me, it’s nowhere even closer to replacing Google. It gets a lot of stuff right, and for some stuff, like deep research where you could benefit from ChatGPT’s ability to retain context, it’ll be the better choice. But to become the default search engine for most people and replace Google? A lot needs to change and improve for that, especially if it wants to become the default search engine on mobile, where on-the-go searches need to be quick and efficient.

Meanwhile, you could add ChatGPT as a custom search engine to your address bar instead of making it the default and get the best of both worlds.