The AI search tools are improving, but they still have a long way to go before they get the concept of a search engine and how people use them. Search engines will soon be impacted by AI. This is what we have been informed. It appears that we are rapidly approaching a novel approach to information discovery and consumption on the internet, as Google keeps getting worse while tools such as ChatGPT, Google Gemini, and Microsoft Copilot continue to improve. Search giants Google and Bing are placing large bets on artificial intelligence (AI) as the search industry’s next big thing, while startups like You.com and Perplexity are promoting themself as next-gen search services. We don’t need multiple links now. We need direct answers.
Now, here’s the thing: a search engine is multi-functional. Many more people use Google to access their email, go to Walmart’s website, or know who was the last president of China than for important and difficult-to-find scientific knowledge. The most fascinating statistic is the huge number of annual users that enter “Google” into Google’s search bar. Although Google is primarily discussed in terms of its research capabilities, it actually receives billions of requests per day for almost any task imaginable.
That being said, the true test for all these potential Google competitors is not their information-finding abilities. It’s the ease with which they can mimic Google’s every move. As a result, we decided to do an experiment with three cutting-edge AI products: we used a number of AI techniques after retrieving the most recent list of top-Googled queries and questions compiled by SEO research firm Ahrefs. Compared to the contents of Google results, these bots based on language models are really more helpful in some cases. However, in the majority of instances, we realized how difficult it will be for any other contender, whether AI or not, to replace Google as the web’s dominant player.
Search engine optimization specialists will tell you that there are essentially three categories of queries. People most often use navigation, which consists of simply typing a website’s name into a search bar to access that website. All of the most popular Google searches, including “YouTube,” “Wordle,” and “Yahoo Mail,” are for navigation. The fundamental function of a search engine is, in fact, to direct users to specific websites.
None of the AI search engines can compare to Google when it comes to navigational questions. You will never get the wrong result from a navigational Google search; given that it’s strange that Google shows you all those results when it should just send you straight to Amazon or something, it’s quick and usually correct. However, the AI bots prefer to ponder for a brief moment before providing an overload of vaguely relevant information about the business, even if all you really need is a link. Some even failed to include a link to Amazon.com.
More than the extra data, what annoys people is the amount of time it takes for these AI systems to get the results they require. People prefer a direct connection to Home Depot rather than having to wait ten seconds for a paragraph of auto-generated content about the company. In every instance, Google comes out on top.
The second most common type of search is an information query, in which the user is looking for precise, one-size-fits-all answers. A few examples of extremely common information queries include “NFL scores,” “What time is it,” and “weather.” The score, the time, and the temperature are all information that you must have, regardless of who provides them to you.
In this case, the outcomes are entirely unpredictable. You can’t rely on AI for things like real-time sports scores: While Copilot was usually spot on, You.com and Perplexity both provides you with out-of-date information on a regular basis. Google is superior to the rest because it not just gets it properly but also typically displays a widget with additional statistics and information. The same goes for tasks that necessitate knowledge of your precise location or context; while Google likely has this data, AI bots typically do not.
Answers to questions with a longer shelf life, such as “How many weeks in a year?” or “When is Mother’s Day?” were perfectly accurate across the board. The AI responses also include some useful background information in several instances. To be honest, they cannot be trusted very often. The number of weeks in a year is actually 52 plus one day, according to You.com, even though Google will tell you that there are 52.1429 weeks. Compared to 52.1429, that is more helpful.
Later on, though, Perplexity contradicted itself head-on by saying that a normal year consists of 52 weeks and a leap year of 53 weeks and one day. That doesn’t make much sense.
A leap year is 52 weeks long, with an extra day added on. A regular year has around 52 weeks. More specifically, there is one extra day in a typical year since 52.143 weeks make up a typical year. In contrast, 52 weeks and 2 days make up a leap year, which happens every four years with a few exceptions. The variation in the number of weeks can be explained by the fact that both common and leap years have an extra day in February.
More investigation has given me the conviction that what You.com says is correct. It took too long, and now I have to double-check my searches, which kind of ruins the point of having them summarize everything for me. If speed is the only criterion, then Google will continue to win in this market.
That being said, there is a specific kind of information query where the inverse is actually correct. They are called Buried Information Queries. “How to screenshot on Mac” is one of the most frequently asked questions as an example. The answer is probably there on the internet somewhere, but you’ll have to navigate through a lot of advertisements and SEO nonsense to get it. For example, to capture the entire screen, press Cmd-Shift-3. It seems like every AI tool, including Google’s own Search Generative Experience, just grabbed that data and handed it to you. Great job!
Does it raise any difficult questions that might destabilize the web’s business model and infrastructure? Yep! On the other hand, it’s better when used for searching only. Similar outcomes have been experienced when seeking out information that is both common knowledge and often hard to come by, such as ingredient alternatives, coffee ratios, earphone waterproofing ratings, and similar topics.
The exploration query is the third type of Google search. There is no one correct response to these inquiries; rather, they mark the start of a journey of discovery. “What is TikTok?” “Why were chainsaws invented?” and “How do you tie a tie?” are all examples of popular explorational queries. Searching for topics like “things to do in Helena Montana” or “NASA history” on Google, or even just looking up the name of a musician that you heard about, is an example of exploration. The rankings show that these are not the most popular searches on Google. However, in these situations, AI search engines truly excel.
Let’s get this chainsaw thing straight. Upon testing, before detailing their technical development and eventually adoption by lumberjacks, Copilot gave a multi-part response regarding their medical origins. On top of that, it provided with eight links to excellent resources where you could learn more. The answer from Perplexity was considerably shorter, but it came with some neat pictures of vintage chainsaws and a YouTube video that explained everything. Although Google returned many of the same links, it did not perform any kind of evaluation on our behalf. Its generative search still only provided with the bare minimum.
Citations are an outstanding feature of AI engines. You can now easily go to the original source if you come across any interesting fact on sites like Perplexity and You.com because they are gradually improving their practice of including links to their respective sources, often inline. This is a positive and useful trend, though they don’t necessarily provide enough sources or place them correctly.
Do you know that the most common query on Google is straightforward: “What to watch?” Google has a dedicated page layout for this, with rows of posters showcasing “Top picks”, “For you”, and finally, popular titles plus genre-sorted options. AI search engines all failed at this, though Copilot provided a rundown of five hit films, Perplexity seemed to have a jumble of choices, and You.com provided an abundance of outdated data while recommending “the 14 best Netflix original movies” without specifying which ones they were.
Artificial intelligence (AI) is the way to go here; however, a chatbot would be an inappropriate user interface because all people want sometime is a response to their question, not a list of links. A page containing search results is, in fact, the same thing! The fact that this is the platform’s most-asked question suggests that Google has taken note and come up with a solution that is significantly more effective.
That pretty much sums up the current situation. In comparison to search technology from decades ago, generative AI may prove to be a superior tool for certain online searches. However, contemporary search engines are more than simply linked lists. The best way to describe them is as mini operating systems. With the built-in calculators, converters, flight pickers, and an array of other tools, they can respond to questions directly and get you where you need to go with a couple of clicks. The majority of people don’t set out to become information enthusiasts. The objective is to find a solution or a link and then exit. These LLM-based systems simply can’t compete at the moment.
So, in our opinion, the product, rather than the technology, is the central concern. Google isn’t the only one who thinks AI can improve search engines’ ability to comprehend queries and process data. Everyone in the industry knows that already. However, will AI companies be able to develop more sophisticated chatbots more quickly than Google can revamp its business model, information presentation, and surface speed? A search bar with ten blue links isn’t the solution, but neither is a generic text box. If you want to know anything, you have to search for it. So it is safe to say that AI search tools are not de-throning Google Search Engine anytime soon.