After a Week with Perplexity, AI Search Still Falls Short of Replacing Google

The fact that I began this experiment a week later than I had originally planned should tell you all you need to know about how entrenched Google Search is in our lives. I can’t specifically tell you why I didn’t want to switch search engines, let alone to one that was powered by artificial intelligence, but I can tell you for a fact that I tried to delay the switch for as long as possible.. until I couldn’t.

There has been a lot of talk recently about Google Search and how its results have become worse, but that’s not really the reason why I decided to switch search engines. I did it because I did not believe we had reached a point where Google Search’s utility and dominance could be challenged.

In some ways, I was proven wrong. In some ways, I was proven right.

What began as an opportunity for me to try a Google alternative turned into a deep exploration into why it was so hard to transition away from a service that has become synonymous with the internet itself and why the best-case scenario for services like Perplexity is not to upend Google, but to co-exist with it.

Here is my story.

---

---

Perplexity isn’t the first search engine seeking to entice users away from Google. It’s only one of the latest, and one of a very few that use generative artificial intelligence to power its results.

Founded in 2022 by former employees of OpenAI, Meta, Quora, and Databricks, Perplexity hit the scene just around the same time ChatGPT launched. Soon, it began making headlines for its mouth watering valuation and the millions it was raising from notable investors.

Despite quickly becoming a valuable company, Perplexity is very much an underdog. Its estimated valuation of between $1 billion and $3 billion pales in comparison to OpenAI’s $100 billion price tag and its search engine processes less than 1% of user queries in an entire year than Google does in a month, based on figures available for 2023.

Google User Queries (December 2023) vs. Perplexity User Queries (Full-year 2023)

Yet, Perplexity distinguishes itself for how it functions. Its AI chatbot scours the internet for answers, listing citations for its results, and the company emphasizes accuracy as its unique selling point.

So strong is Perplexity’s allure that Nvidia Chief Executive Jensen Huang told Wired in February that he used it almost every day. Though it’s worth noting that Nvidia invested in the company a month prior to the interview being published, so it may be in Huang’s interest to present himself as a fan.

Besides Huang, Jeff Bezos also took part in the January fundraising, solidifying Perplexity’s reputation as a start up on the radar of the who’s who of the tech world.

While I did occasionally follow developments from Perplexity, it was in fact a contributor at HackerNoon who encouraged me to to try it out for a week based on their own rave experience.

With the backing of so much star power, how could I refuse? So after delaying it for a week, I decided that I would switch search engines the following Monday, come what may.

---

---

My search engine swap began with a note to my team at HackerNoon on Monday telling them that I would exclusively be using Perplexity for a week.

Hey everyone! Starting today, I am changing my default search engine from Google to Perplexity for a week and chronicling my experience of ‘searching’ in a completely different manner. If you have any thoughts or any specific things you’d like me to check out, let me know 🙂

I really don’t know why I sent that message on Slack, but if I had to guess, it was probably to hold myself accountable. At that point in time, I wasn’t really certain whether I would be able to continue on with the experience, having never really used Perplexity before.

As one of the millions of users that use Google daily, I have never really cared much for switching search engines because I always assumed the alternatives would be worse, though not without credence.

Google vs. Bing

So it was only natural that I felt a bit nervous. Like I was about to take something that I had always taken for granted and make it a whole lot difficult.

Companies like Google rely on users’ dependence on their ecosystem to maintain a dominant position in the market. Warren Buffet calls it a business moat — the idea that some brands have competitive advantages that make it difficult for others to chip away at its market share.

In Google’s case, its dominance in search is a self-reinforcing act: more users mean more data, which leads to better search algorithms and personalized results, attracting even more users and advertisers.

This iron-clad loop is why most users, including myself, never really think about alternatives. But even if we would, Google goes the extra mile to protect its moat by making it inconvenient, if not outright difficult, to switch away from its service.

In 2020, the US Justice Department charged Google with monopolizing the search engine market by paying phone makers, computer manufacturers and browser developers to ship their products with its search engine set as default. Then, as the trial began last year, The New York Times ran a story highlighting the specific design choices Google had made to make it difficult for the average user to ever swap out its search engine from their phones or browsers.

My own experience switching search engines was no different. As a Firefox user, I had to go into the browser’s settings and find the option to change the default search engine, and even then, I was only presented with Microsoft’s Bing and DuckDuckGo.

Firefox Default Search Engine Options

Thankfully, a quick Google Search (ironic, I know) took me to a Reddit thread that showed me the exact steps I needed to take to add, and then make, Perplexity my default search engine. Chrome users also need to take a similar detour to get the same result.

While it doesn’t take very long to set up a new default search engine on your phone or browser, most users are never likely, if ever, to recreate the steps needed to do so. Without making the effort to change their default settings, the average user would either then have to keep a separate tab open on their browser or install a new app to use a service outside of Google, which, in my opinion, is not sustainable in the long run.

It’s deterrents like these that keep Google dominating the search engine market, and why the company goes out of its way to ensure it remains the default search engine on users’ devices. So much so, that it paid $20 billion in 2022 to Apple alone to be Safari’s default search engine.

---

---

Having set everything up, my first experience using Perplexity was the unfamiliarity of it all. I was so used to the Google way of doing things, that I spent the first few days being acclimated to Perplexity’s overall user interface and the results I was seeing.

Initially, my reaction was often a groan. Not because the output was bad, but because I was using Perplexity the same way I would use Google: as a way to quickly get around the internet.

While Google does not charge for its service, it extracts a hefty premium from users by collecting their data and either selling it to advertisers or designing products based on the information it has. Those data points are far more valuable than any dollar amount a user may or may not choose to spend.

In contrast, Perplexity just spews out the results a user is seeking, without necessarily tracking a users’ behavior for the explicit purposes of selling it to a third-party or training its own large language models.

So while Google acts as the internet’s middleman, Perplexity appears to be the last-mile delivery agent that leaves the information at your doorstep.

Admittedly, Perplexity’s approach does take some getting used to.

For example, while reading a news article, I came across the name of a Google executive whose LinkedIn profile I wanted to check — so I searched her name from my browser’s drop-down menu without even thinking about it.

Searching for Answers Using the Dropdown Menu

But it was only after the new tab opened that it hit me that I was no longer using Google, and that getting to her LinkedIn profile or Wikipedia page would require some additional steps.

Hema Budaraju Search Results Using Perplexity

In another instance, when I was looking to do some quick measurement conversion, Perplexity told me the math, but did not display the Google converter I had taken for granted.

Pound to Kilo (Perplexity)

Pound to Kilo (Google)

It’s little things like these that made me want to return to Google. But it was still the start of my journey, and I still had the rest of the week to go.

---

---

After getting over the initial obstacles, I began using Perplexity the way I thought it was intended. Not necessarily as a search engine, but as a research tool.

Based on my experience, a tool like Perplexity works best if a user has to deal with a lot of sources, but might not necessarily be suited to those who know what they’re looking for — which is where Google might come handy.

This scenario played out when a friend of mine, while discussing a high-profile case that involved a highly affluent and influential family’s son murdering a member of a police officer’s family, claimed that the victim’s side was related to a noteworthy  politician based on a post on Reddit.

Reddit Post Screengrab

I was instantly skeptical of the Reddit users’ claim, so decided to dig up some more information, and to Perplexity’s credit, it accurately corroborated that the politician in question was not related to the victim’s family.

Perplexity Response

But Perplexity was not always accurate in its responses.

During a later conversation with the same friend, I wanted to bring up a list of all AAA video games released in 2024 and noticed that Perplexity mentioned two titles that were yet to be released.

List of AAA Video Games Released in 2024, According to Perplexity

Dragon Age: The Veilguard, which Perplexity claimed was released on March 22, is slated for an October release, while Star Wars Outlaws, which only came out recently, was listed as having been released in April.

I wouldn’t have caught this if I already didn’t know about the games, but there might be instances where a user might not be aware they’re being fed incorrect information until they start digging further.

Running the same query at a later date, Perplexity’s response was accurate, but not necessarily exhaustive. It correctly mentioned that Star Wars Outlaws was released in August, but did not mention that Dragon Age was due for release in October.

Perplexity’s Updated Response

In yet another instance in which I was reminded that Perplexity was not a typical search engine, I asked it to visit a particular website and search for a specific query — only to be told that it could not.

Asking Perplexity to Browse a Website

It is then I learned that having real-time access to the web did not give users the ability to redirect Perplexity to specific websites for information. At least, not based on what I was able to gather.

Perplexity’s Marketing Material

Instead, users would have to use Google if they wanted to run search queries targeting specific websites.

Google Search Results

---

---

During the course of the week, there was one specific instance where Perplexity inadvertently let on more about its underlying technology than it should have.

While doing my research for this piece, I thought it would be interesting if I just asked Perplexity to tell me a little bit about itself — and the response both amused and confounded me.

Perplexity’s Response When Asked to Introduce Itself

Perplexity has made no secret that it uses other large language models besides the one it has developed in house to power its results. In fact, the company offers users the ability to switch models on its paid plan with options including GPT-4o, Sonar, and Claude.

Perplexity’s Paid Plan(s)

But being told by Perplexity that it was “developed” by OpenAI point blank was either an instance of AI hallucination or an internal AI struggle in which the company’s own LLM lost — I couldn’t tell you which.

It did, however, underscore the fact that Perplexity did not think it had an identity of its own, at least, not in that specific instance, and got me wondering whether the AI search engine was simply a reskin of existing models designed to operate within certain parameters.

Perplexity does itself no favors when it offers its proprietary LLM as the base, free model, and requires users to pay to ‘unlock’ the other LLMs. This inadvertently gives the impression that the other models are somehow superior to Perplexity’s own, which is a rather odd choice to make.

---

---

Soon after Microsoft integrated ChatGPT into its search engine Bing, there was chatter on the internet on how ranking in AI search results would be the next big thing. The idea was to have publishers and businesses produce ads or content that would then be displayed in an AI’s response.

Whether marketers are trying to rank on AI in earnest is yet to be determined, but Perplexity does an excellent job of demonstrating that, as a proof of concept, ranking on AI could work.

During multiple instances, the search engine impressed me with its results by citing a range of sources, including, at times, local or lesser known publications. In fact, I was genuinely pleased when Perplexity cited HackerNoon for some of its responses. This does open up users to new sources on the internet that they might otherwise be unaware of.

I ‘Googled’ Something

But while the jury is still out on whether ranking on AI is the “next big thing,” not every publication is on board with the idea of Perplexity using their content for its answers.

Just a few months ago, Perplexity fell afoul with some major publications for bypassing their paywalls and plagiarizing their content. The Verge called it Perplexity’s grand theft AI, and the whole debacle prompted the company to introduce a revenue-sharing model for publishers.

On paper, this sounds great, but I’m personally worried that it could heavily incentivize Perplexity to get people to sign up for its pro subscription just so it could keep paying publishers, creating a symbiotic relationship that will ultimately hurt consumers.

Perplexity hasn’t reached that stage yet, thus the need to raise even more funding so it could get more and more consumers to use its service at a cost to itself. But there might come a time when the company would have to prioritize businesses at the cost of its customers, followed by prioritizing investors at the cost of enterprises. Throughout that entire process, the platform could become worse and worse.

Canadian author Cory Doctorow calls it the “enshittification” of the internet, the slow decay of online platforms that happens in three stages: be good to users; abuse your users to be good to business customers; and then abuse business customers to be good to your investors.

And then.. the platform dies, Doctorow says.

---

---

Oftentimes during my week-long experiment, I kept wondering who Perplexity is for.

Despite marketing itself as one, Perplexity is no search engine. It is a generative AI product that is competing in a market dominated by OpenAI’s ChatGPT. Not only does it have to contend with forming its own unique identity, but also compete with companies like Google or Meta, who are far, far larger than what Perplexity can ever hope to be and are investing way more resources in carving a space for themselves in the AI market.

Getting people to switch from Google is easier said than done, especially when you consider that Google has now begun offering overlays on Android phones so users can use its AI while using another app, thus ensuring they’re locked into its ecosystem.

On top of that, what makes Perplexity’s position even more tenuous is the use of its competitor’s large language models to power its paid results. The whole thing echoes Yahoo’s use of Google in the 2000s to power its search engine, which ultimately resulted in its downfall.

Things become even harder for Perplexity when you factor in the launch of OpenAI’s upcoming SearchGPT.

In OpenAI’s case, the company has a first-mover advantage and a certain level of prestige for being the first firm to crack how to do generative AI right, which might give people a reason to be excited about its product instead.

Worryingly for Perplexity, the same investors that have injected capital in the company are now planning to do the same in OpenAI at a much, much higher valuation.

When accounting for all these factors, Perplexity may end up being relegated to the same heap as other search engines that have come before it. After the artificial intelligence hype dies down, the best Perplexity could ever hope for is to capture the market share from other Google alternatives or ultimately get bought out by a firm who might be interested in its technology.

While I do think that what Perplexity is doing is interesting, it’s not necessarily the be-all end-all one would have you believe. The tool does offer the potential to speed up the research process, but ultimately suffers from the same problems that other LLMs do.

I don’t think we will ever reach a point where “Perplexitying” something will enter the common lexicon, but that is not to say that we will continue using Google forever. It’s possible that Perplexity and other products like it pave the way for something that is far superior and engaging than what the search engine experience has become.

As for me, I’d rather go back to Google and do my own research and find things — until something truly better comes along.

---

This article was originally published by Sheharyar Khan on HackerNoon.