Blog
I tested Perplexity vs Google AI overview with 7 prompts — the results were shocking
Search is undergoing a profound change. For decades, Google has dominated the web search world, with some 90% of all searches funnelled through the massive Google machine. But suddenly, with the arrival of artificial intelligence, things are starting to change, and seriously so.
Not only are people increasingly using AI products like ChatGPT as their default search tool, but companies like Perplexity are also building businesses around search services. The idea is to combine the power of AI analysis with the huge amount of conventional search data available at the end of a cursor.
But Google is fighting back. The company has recently released an advanced search function called AI Overviews, which aims to bridge the two disciplines and deliver the kind of informed search results the market demands. It’s a new kind of search on steroids.
So how do the two approaches compare in everyday use? We take a look at Google’s new AI Overviews and compare the results to Perplexity AI, to see which gives a better bang for the buck.
Prompt One – Fathers of AI
Prompt: Summarize the key contributions of John McCarthy, Geoffrey Hinton and Noam Shazeer to the development of artificial intelligence.
We thought we’d start with something close to home – a look at the architects of AI from the past. First impressions are that Google delivers a competent but fairly traditional results page from this request. Its answer of 238 words covers all the basic points and gives a good overview of the points as you’d expect.
Perplexity delivers over 400 words, but it’s more than just the quantity that’s impressive. It’s the fact that the results are laid out in a much more engaging manner, with the user being encouraged to explore additional information in a variety of different ways. They can explore related data, look directly at the sources and even regenerate the results to get a different perspective. Where Google seems to do the bare minimum, Perplexity really seems to add user value.
4/10 | Row 0 – Cell 2 | |
Perplexity | 7/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Two – Trip to Tokyo
Prompt: Create a 3-day itinerary for a first-time visitor to Tokyo on a modest budget. The user is interested in Japanese culture and food, but wants to avoid tourist traps.
This is a real kicker, a stark example of the old versus the new. Google completely fails to deliver any meaningful response, but instead retreats to a standard Google search. The answer merely features a selection of third-party websites offering tour advice. This is obviously beneficial to Google as it will no doubt allow it to earn ad revenue.
Perplexity, on the other hand, delivers a glorious 1000 words of real down-to-earth itinerary. This includes gorgeous photos, maps and itemized costings, which will be more than enough for the user to get a great idea of the proposed experience.
2/10 | Row 0 – Cell 2 | |
Perplexity | 9/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Three – Noise cancelling headphones?
Prompt: What is the technology behind noise cancelling headphones?
Once again we can see the difference between traditional search results and new style AI analysis, although in this case the difference is not so great. Google’s results are very credible with 186 words of explanation, along with a useful YouTube video.
Perplexity, however, takes it to a more advanced level. The explanation is roughly the same, but the prose is much more accessible to a layperson. Instead of using the word ‘inverse’, for example, the app chooses to use simpler English to explain how sound is cancelled. It’s a subtle but important use of 500 words to achieve a more understandable answer.
6/10 | Row 0 – Cell 2 | |
Perplexity | 8/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Four – Farming and garden prep
Prompt: Explain the ‘double-dig’ method of garden bed preparation and contrast it to no-till farming
This is quite an obscure test, involving little-known agricultural techniques. But for gardeners it’s a very important topic. Google’s response is very workmanlike and informative, and uses 260 words to deliver a good answer to the question. The use of Reddit and the Royal Horticultural Society (RHS) are also great sources, which add authority to the answer.
Unfortunately for Google, Perplexity once again matches and exceeds in response quality. The RHS and Reddit are also mentioned, as is YouTube. But two things really make this answer stand out. First, the use of a great table to explain the differences at a glance and, most importantly, a conclusion, which gives a clear indication as to why no-dig is increasingly considered the better solution.
7/10 | Row 0 – Cell 2 | |
Perplexity | 8/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Five – Compliance changes
Prompt: What are the primary compliance challenges for a US-based tech startup under the EU’s AI Act?
This request pushes search to the limits of topicality and obtuse legal documents. Surprisingly, Google’s results are very lackluster. The search engine offers up a 57 word March 2025 ‘featured snippet’ from an obscure third-party publication, and that’s it. It shows no interest in digging deeper into the topic for the user.
Yet again Perplexity tries harder. We’re talking 600 words set in a beautiful bullet point format, running through the main challenges and issues surrounding compliance. Along with 9 easily accessible sources and a handful of related subject matter links. Masterful.
2/10 | Row 0 – Cell 2 | |
Perplexity | 8/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Six – Cryptocurrency history
Prompt: What does the history and potential future of blockchain and cryptocurrency look like?
This prompt clearly demonstrates why Google’s AI Overview is unfortunately not really ready for prime time. The original prompt was something like ‘explain cryptocurrency to a fifth grader’, but when tested Google couldn’t handle it and served up a lame Quora snippet. It’s only when we changed the prompt to this one that AI Overview kicked into action, and delivered a reasonable result.
It’s obvious there’s not that much AI involved in AI Overview yet. Interestingly though, this was probably Google’s best result. We got 400 words of densely packed information covering the topic clearly and succinctly. Perplexity was also good, providing 600 words and a nice table. Not much to choose between the two then.
8/10 | Row 0 – Cell 2 | |
Perplexity | 8/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Prompt Seven – Cat discovery
Prompt: What kind of cat is this?
For the final prompt we thought we’d go with something a little more exotic. Both search platforms support image upload, so what better than to upload a friendly looking cat to get some more information?
Google takes the uploaded image as a prompt to display a page full of similar images, which aligns with its original image matching search. But a re-prompt of ‘what kind of cat is this’ then delivered a very short four line answer which, although correct, was not super helpful.
Perplexity’s response was 246 words, with bullets points, covering coat pattern, fur, and the breed. Even a fun fact (calico cats are almost always female). Engaging and informative.
3/10 | Row 0 – Cell 2 | |
Perplexity | 7/10 | Row 1 – Cell 2 |
| Row 2 – Cell 0 | Row 2 – Cell 1 | Row 2 – Cell 2 |
Final thoughts
The king is dead, long live the king? Based on this showing, the rumors could indeed be true. The mighty Google may in fact be on the way to losing its grip on the world’s search traffic. Is this the end of an era? Time will tell.
However if there’s one thing we’ve learned over the years, it’s never to discount the ability of the Google empire to strike back. Uniquely in the world, the company has the compute power, the data and the legendary AI pedigree to surprise us all.
Test Notes. It should be noted that we did not use any of the advanced Perplexity functions, but kept to the basic default service. Which make the results even more impressive. It’s also important to recognize the fact that AI can get things wrong. Both services feature disclaimers which stress that users should not assume AI search responses are factually correct.
This is an early technology finding its feet, users should take care.
More from Tom’s Guide
Back to Laptops
