In general, whenever you search for something in a search engine, there will be countless webpages that contain your query, but you won't look through millions of webpages to click on one, so there must be an algorithm to rank these webpages based on how relevant they are to your query. What is the best result for this particular query? That's the extremely difficult problem that search engines strive to solve generally.
Because search engines are optimized for this task, they are particularly powerful when there is a single canonical result for a particular query that is always the correct one. For example, if you search for the name of a website, e.g. Wikipedia, the first result should be the URL for the website, https://wikipedia.org/. In this way, search engines become nothing more than an over-engineered phone book for URL's. If I search for someone's name, such as a celebrity name or authority, I'd want their personal website or blog, or primary social media profile page, to appear first. If I search for the title of a game, application, or movie, I'd want their official website as well.
Intriguingly, AI chatbots can't be trusted to give the correct information for a query. You'll always need to double check it just in case it has hallucinated something.
However, there are cases an AI chatbot such as ChatGPT or Google's Gemini can prove to be more useful than a search engine ever could: when you don't want a canonical result.
Searching for Lists of Things with AI
These are actually very common queries that people have and that search engines struggle to answer.
For example, let's say that I want to find applications for digital art. If I search for this on Google, Google can't actually give me a list of applications for digital art, and if it does, it often includes random applications that don't even fit the query displayed in the most awkward way possible: just icons next to their labels. The first results for this type of query will be listicles, articles titled "5 best applications for digital art" and the sort. Some people also call this "SEO spam," but honestly there is no other way to write a title for an article that is just a list of things.
Isn't it a bit strange that the SERP is a list of results, but Google can't just give us a list of applications? Instead it gives us a list of webpages that list applications. That's because it needs to rank those results in the SERP. There must always be one result that is the first one and it's the best one, and this has to be a listicle.
By contrast, AI chatbots can answer queries like this in a way that makes sense, giving you a list of things.
This is far from perfect, and one may argue it isn't even a good thing.
In some cases, what is happening is that someone wrote the listicle, and the AI chatbot just infringed on their copyright and stole all the content without even crediting the author. In other cases, it may be aggregating information by itself that you can't find on any webpage written by a human. Sometimes the items it lists don't match your query, as the AI will often hallucinate randomly.
What sets AI apart from traditional search engines and even humanly written listicles is your ability to refine the query.
Refining Searches with AI
After searching for something AI, you can control its next answers giving it some direction toward what you want.
For example, you can usually ask for more results. And then keep asking for more. You can discover some truly obscure information using this method, that honestly I'm not sure I'd be able to discover using a traditional search engine no matter what I searched for. And I'm a black belt in Google-fu. For example, I discovered Fotki by repeatedly asking ChatGPT for more photo-sharing websites until it gave me a result I didn't know about yet.
You can usually steer AI by telling it to give you more results "similar to X," or that "aren't Y."
Negative queries are something that traditional search engines really struggle with that AI chatbots would be much better at. An LLM can easily place "not paid" next to "free" in its multidimensional AI model, but in a search engine if you searched for "-paid" that would exclude all webpages that have the word "paid" in it regardless of the usage of the word in a sentence, e.g. if Krita said it's "not paid," that would exclude the webpage, since the word "paid" is inside the phrase "not paid." On the other hand, if you search for "not paid," that might include results that say something is "paid" and also say it's "not" available for Windows, for example.
Getting Links to Articles with AI
It's possible to make an AI chatbot give you links to articles that talk about a subject instead of summarizing it. That's great because you avoid the risk of the chatbot just hallucinating half of the information. To do this just write your prompt so: "give me links to articles that talk about X."
You can in fact specify all sorts of filters in natural language. For example, instead of "articles," you can ask for "links to interviews with experts that about X" or even podcasts. Whether the "expert" is an actual expert or not is a bit random, as AI usually is. Make sure the linked article has proper author information, that the author has proper credentials, and that you can verify the validity of those credentials, e.g. if they say they're Ph.D in a field you should be able to find their dissertation somewhere, and it should be cited by other people.
Keep in mind that AI is really random with the results it gives you. For example, when I tried to use it to find Youtube channels that talk about programming, it gave me all sorts of channels, including some very old channels that are now full of videos about get rich quick schemes. Even the most relevant results weren't really what I was looking for.
Even though it doesn't work sometimes, it's always good to have this extra ability at your disposal.