When I search online for the most relevant, comprehensive, and accurate information I’m looking for, I find it’s typically a hit-or-miss procedure. Traditional search engines will give you results but heavily weighted to sponsored listings or paid ads, thus making it hard to determine what’s trustworthy and worth my attention. In fact, ‘listing’ is the key word as what you get is precisely that, a listing of links that might match all or some of the keywords in your search term.

Google has dominated the search space for much of the past two decades, shaping how we discover, learn, and make decisions in the digital age. It has over 90% of the search engine market worldwide. And let’s not forget – Google Search, Microsoft Bing, and many others are advertising businesses as much as anything else.

The advent of artificial intelligence tools like chatbots, kickstarted by ChaptGPT in late 2022, moved things out into the wider world and gave everyone far greater means of searching and finding better answers to everyday questions. But given that ChatGPT and other generative AI tools and LLMs like Claude, Google Bard, Microsoft Copilot, etc, are not just search engines with a bit of added AI, there’s been a market opportunity for someone to fill the gap with a tool specifically designed for search that is genuinely innovative and definitely evolutionary.

Right now, I think that’s Perplexity AI, a tool the developers call an “answer engine“, that arrived on the scene late last year to signal a new era in search technology, challenging the status quo with its very different approach to information search, retrieval and presentation.

This weekend, I had an experience that demonstrated the difference between Perplexity and every other top search engine in use around the world. It’s the perfect illustration to see that difference.

Starting with Google and Bing

I wanted to find the answer to a simple question that came to mind following an informal message chat on Threads about hybrid cars and EVs.

My question was: Is a self-charging hybrid car a better option than a plugin hybrid, known as a PHEV? It’s deliberately a broad question as it’s typical of the kind of quick question I tend to ask Google

I started with Google Search; this screenshot shows the typical result you get with Google search.

This is what’s visible in the browser window before any scrolling. All of it is sponsored content – the organisations shown have paid for this information to be algorithmically selected in response to the keywords.

As all are paid for, you’ll know that like all paid advertising, you’ll not be getting unfiltered, neutral answers. There will always be an angle and a point of view being promoted.

Next up – the same question asked of Microsoft Bing.

Bing is similar to Google Search where the first screen of results is advertising. There is a major difference, though, as you can see in the small panel on the right in the screenshot, which is AI-generated content using Copilot, Microsoft’s own AI bet.

But when you scroll down the page, you get a sense of the value of an “answer engine” (this label isn’t exclusive to Perplexity) that starts to get really interesting with the greater breadth and depth of responses, as the next screenshot shows.

This is far more useful and helpful in giving added perspective to the results rather than simply presenting a list of URLs. It gives you more to explore there and then rather than having to perform additional searches based on the first results that may or may not get you closer to what you’re seeking.

And then there’s Perplexity

But it’s when you try Perplexity that you start to see where online search is moving to in the next stage of leveraging the power and value of AI for day-to-day utility.

The same question asked of Perplexity AI produced responses that were eminently useful, far different to a list of links that an algorithm has created based on advertsing data.

This screenshot shows what Perplexity produced in response to the same question I asked Google and Bing.

Perplexity’s approach is conversational. I find that every time I use it, I engage in a conversational manner as reflected in my search requests. With Google typically, you ask in terms of keywords. With Perplexity, you ask in a way that you’d ask a question of another person. And when you get a response, you can continue the conversation without starting over each time. It’s a natural-like process.

And what responses! Perplexity looks at everything it found related to your request and then writes a summary of it all. As the screenshot shows, it also annotates the texts with reference to each source it finds (the numbers you see sprinkled throughout the text) to give those texts context and relevance, and a sense of credibilty when seeing source citations actually noted in the text. Not only does this make your verification job a lot easier, it also makes it a lot easier to decide what is trustworthy and usable.

You can share the results from each activity directly to some social networks; or, use the share link to share it anywhere. And, if you use a Chrome extension such as Save My Chatbot, you can save a result as a markdown file, which is especially handy as you get all the links to all the sources Perplexity found. If you open up the saved markdown in an appropriate editor – I’m trying out Typora – you can then edit it or export it to HTML for further use.

This markdown text is all far more useful than a simple ad-influenced list of URLs in a browser as this screenshot shows.

(Here’s a link to the Perplexity result: https://www.perplexity.ai/search/Is-a-selfcharging-2lPAHB9UQlqPn089YyBD7w?s=c)

Essentially, Perplexity has replaced Google Search for me.

Last week, I started a monthly subscription to Perplexity Pro, the paid-for next-level service that removes many of the throttles that exist in the free basic service. I’m trying it out in parallel to ChatGPT Plus, the paid service I’ve been subscribed to since last year and which, broadly speaking, is a direct competitor to Perplexity Pro. I don’t see ChatGPT as that, though, as I think both are differently focused.

Then again, perhaps it doesn’t matter. I experience much overlap already in how I use all such tools – one for search like Google, another for general research, a bespoke GPT for specific-topic research, etc. The day will surely come when you should be able to use just one such tool to accomplish all your everyday requests. Or it will be much clearer which tools are best for what purpose. The day hasn’t come yet, though.

It’s worth mentioning as well that, like every other generative AI tool out there, Perplexity has its imperfections. And you’ll still have to do your part of the process, mostly verification of everything Perplexity tells you.

I strongly recommend you try Perplexity for search if you haven’t yet. You can use the website version, or install apps for Android and iOS devices. You don’t need the Pro account to get impressive results; the free account will be just fine to try it out.

Finally, read a very good assessment of Perplexity by Kevin Roose in the New York Times last week, with a focus on the Pro version. This sets the scene:

Many start-ups have tried and failed to challenge Google over the years. But Google seems less invincible these days. Many users have complained that their Google search results have gotten clogged with spammy, low-quality websites, and some people have started looking for answers in places like Reddit and TikTok instead.

Kevin Roose, New York Times, 1 February 2024

Opportunity knocks. Perplexity could open the doors.

[Update 28/3/24:] Also, see a new post published today that expands on the themes discussed here – The Evolution of Search in the Age of AI Chatbots

Related reading:


Footnote:

I created the AI-generated image at the top in ChatGPT Plus and DALL-E 3. My simple prompt was: “Create a square image that portrays a search for information online. Show a large computer monitor in a modern high-tech office with people gathered around the display which has the logos of Google, Bing and Perplexity AI visible.” I got the Google logo but not the others.

In response to my criticism of these missing elements, ChatGPT said, “I understand your feedback and acknowledge the oversight. The intention was to create images that closely align with your requests while adhering to guidelines regarding the representation of specific brands. I appreciate your understanding and am here to assist further with any adjustments or new requests you might have.”