Since its release, people have been discussing whether ChatGPT, a chatbot developed by OpenAI, will replace search engines like Google [1, 2]. Because of its advances compared to common chatbots, ChatGPT has become a popular topic of discussion.
It can remember what was said earlier in a conversation, write code, and answer complex questions. For the latter, virtual assistants like Apple’s Siri or Amazon’s Alexa need the help of a search engine, which ChatGPT doesn’t. Thus, people have begun using ChatGPT for information retrieval in place of search engines.
The short answer to whether ChatGPT will replace search engines like Google is “probably not.” It is unlikely that ChatGPT will replace search engines in their entirety anytime soon because of two main reasons:
- They were designed to serve different purposes. While search engines are designed to search for websites, chatbots are designed for conversation.
- ChatGPT, in its current state, has limitations like occasionally producing wrong information or not having access to recent events [3].
Thus, it is unlikely that ChatGPT will replace search engines soon.
The short answer to whether ChatGPT will replace search engines like Google is “probably not.”
Although search engines and ChatGPT were designed for different purposes, they have one shared use case: The retrieval of timeless information. For this article, we define timeless information as information that doesn’t have to be updated (at all or at least regularly). For example, the answers to questions like “What is the highest mountain in the world?” or “How do you make sugar cookies?” are timeless information.
Information that is not timeless would be the answers to the questions like “Is my favorite soccer team winning the current match?” or “Will I need an umbrella tomorrow?”. Since ChatGPT has limited knowledge of events after 2021 [3], the latter is a use case only search engines can handle. But for the retrieval of timeless information, both are suitable.
Thus, for this article, we refine the initial question: Will ChatGPT replace search engines for retrieving timeless information like looking up a sugar cookie recipe.
For this purpose, we will compare search engines and ChatGPT in the following aspects:
- Inputs and outputs: Questions & Answers vs. Keywords & Search Results,
- Integrity of Information,
- Relevance: Supporting vs. Distracting Information
- Customizing Information.
Questions and Answers vs. Keywords and Search Results
If you don’t ask a question, don’t expect to get an answer. Since search engines and ChatGPT are designed to serve different purposes, you will interact differently with them to retrieve information. Specifically, what you input and what it outputs will be different for the same use case.
The way we request information from ChatGPT is already more intuitive than for a search engine. Although you can type a question into the search bar of a search engine, you will likely achieve better results if you don’t. Instead, most of us have gotten used to converting our questions to keywords like “sugar cookie recipe” to type into the search bar.
However, with ChatGPT, you can ask it directly for a recipe, just like you would ask your Grandmother for hers. This conversational interaction with ChatGPT is much more intuitive and user-friendly.
How we receive the requested information from ChatGPT is also more user-friendly than for a search engine. Since you don’t ask a search engine a question, you will not receive an answer but a list of websites that are likely to contain your answer.
If you ask ChatGPT for a recipe, it will give you exactly that: Just one recipe. Thus, how ChatGPT outputs the information you are trying to retrieve is more user-friendly than when using a search engine.
ChatGPT is a more user-friendly way to retrieve information. You can ask it a question and will receive the answer, just like you would in a conversation. With search engines, you first have to convert your question to a few keywords and then find your answer among a list of websites.
Integrity of Information
How we receive the requested information determines how we decide to trust its integrity. For search engines, you have to decide which one you want to trust. For ChatGPT, you get only one answer, and you must decide whether you trust it.
Although the search engine helps us to find the most relevant website, it is not transparent how it does so. Because the search result is a long list of websites, a search engine ranks them by relevance.
Anyone trying to direct your attention to their website can either search engine optimize their content or advertise it to you. Thus, it may not always be clear whether a website contains the best information or its relevance is manipulated. The lack of transparency is unfortunate because users heavily rely on the proposed ranking.
But you can try to validate which search results you want to trust. You can either look at the source and decide if you think it is trustworthy. Or, you can compare a few recipes to one another. Also, you can see if you are the only one that thinks the ratio between dry and wet ingredients is off, based on the recipe’s ratings and comments.
In contrast, ChatGPT provides only one answer, but it is up to you to decide whether you take that information with a grain of salt. OpenAI warns that ChatGPT may generate incorrect information on occasion [3]. A neural network, like ChatGPT, is only as good as the data it was fed – pun intended.
But since food is a popular blogging niche, we can assume that the amount and quality of recipe data enable ChatGPT to produce quality recipes. It can even generate suitable recipes for different dietary preferences, e.g., vegan or gluten-free.
Now, what do these sugar cookies taste like? – They are good. They are not the best sugar cookies I have ever had, but their are also not horrible.
Although you don’t have to decide which source of information to trust with ChatGPT, you can’t immediately validate its information. Although we can assume that most recipes will be edible, you must decide whether you trust a recipe. Currently, you can’t validate any ChatGPT recipes based on their source or ratings because it doesn’t provide this information – you just have to try the recipe.
Supporting vs. Distracting Information
While ChatGPT only answers your question, you will receive additional information when retrieving information with a search engine. Once you have picked a trustworthy website, you have to find the relevant information among its other content. Depending on its type, additional information can be helpful but also distracting. ChatGPT will only give you the answer to your question. If you want any further information, you have to ask for it.
Typical search results for recipes are posts on food blogs, which often contain filler content, like a personal story or a close-up shot of the ingredients. Once you have picked a website, the information you were looking for can be buried among other information, which can be distracting.
But not all additional information has to be distracting. Some information can be supporting information. Because “you eat with your eyes,” we expect an image of the final food. However, ChatGPT is a chatbot and only provides information in text form. That means it cannot provide information in other media formats like images, sounds, or videos.
Blog posts often contain additional tips and tricks to avoid common pitfalls of the recipe. Because ChatGPT doesn’t tell you anything you didn’t ask for, you will have to ask for specific tips and tricks if you want them.
Although ChatGPT answers don’t contain distracting information, they don’t provide any supplementary information.
Customizing Information
Just like we have learned to refine keywords to receive the information we want, we must re-learn the art of asking questions.
If you have something precise in mind, you can ask ChatGPT directly. For example, if you want to receive the ingredients in the metric system. Because many English recipes are created by people living in North America, they use the imperial system to measure the ingredients.
If you don’t have measuring cups available, you also have to convert the units of measure to the metric system, which adds another step to the information retrieval process. With ChatGPT, you can specify that you want the unit of measure of the recipe in the metric system.
But does this mean that everyone will make the same sugar cookies for Christmas from now on? Not necessarily. If, hypothetically, everyone asked for the same recipe, all sugar cookies from now on would taste similarly.
But, with slight tweaks in our questions, we can get different results. E.g., you can ask for a specific variation (“How do you make sugar cookies in Germany traditionally?”) or you can ask for the recipe to use a particular ingredient (“How do you make sugar cookies using Matcha powder?”).
Similarly to how we need to study prompt engineering for AI-generated images, we must also learn how to ask ChatGPT the right questions to get the results we want. But is a recipe a piece of information that should be the same for everyone or is it a piece of art?
Conclusion
We have already discussed that ChatGPT, in its current state, will unlikely replace search engines like Google entirely for two reasons. First, because of their different scopes of use cases. Second, because of ChatGPT’s limitations, like wrong information and no access to recent events [3].
However, ChatGPT and search engines have a shared use case of timeless information retrieval. We have explored this hypothesis with the use case of looking up a sugar cookie recipe:
- You will only get one answer because you can ask ChatGPT a question.
- While you won’t be overwhelmed with choosing a source of information, you have to decide whether you want to trust the information provided by ChatGPT. In contrast to a search engine, you don’t have the option to validate the results based on, e.g., ratings.
- Although ChatGPT will not provide any distracting information, it will also not provide any supplementary information.
- If you want any further information, you have to specifically ask for it. But with intentionally phrased questions, you might get some great results.
I made Christmas cookies with a recipe generated by ChatGPT this year and this time next year, you might, too. To conclude, while ChatGPT might not entirely replace search engines like Google, it has the potential to take away some of their search traffic.
I made Christmas cookies with a recipe generated by ChatGPT this year and this time next year, you might, too.
If ChatGPT could potentially disrupt the way we look up recipes, does that mean it could replace food bloggers? Another big problem with AI is that it uses existing content to create new content. The neural network is only as good as the data that feeds it. If you keep mixing all the colors in the rainbow together, you will end up with brown.
If ChatGPT could potentially disrupt the way we look up recipes, does that mean it could replace food bloggers?
Thus, providing high-quality input data will be essential in the future. How this should be incentivized is another topic. But we can say that food bloggers and content creators should not worry. Instead, you could try ChatGPT and use it as inspiration.
References
[1] Business Insider (2022) Could ChatGPT challenge Google? Morgan Stanley says the search giant has nothing to worry about (accessed December 16th, 2022).
[2] Coindesk (2022) ChatGPT Will Kill Search and Open a Path to Web3 (accessed December 16th, 2022).
[3] OpenAI (2022) ChatGPT Dec 15 Version https://chat.openai.com/chat (accessed December 16th, 2022).
This blog was originally published on Towards Data Science on Dec 22, 2022 and moved to this site on Feb 1, 2026.