Perplexity AI: The App, For The Everyday Person

 

 Perplexity AI, according to BuiltIn, is an AI-powered search engine - and it's now valued at 18$Billion dollars, with a B.

I had it on my phone for research testing - something I do between contracts for money - and simply kept it all this time. With it making a resurgence, I can show you if it's viable for every day use cases.

 I did not use it to generate "art" or writing. 

Screen Reading and Photo Identification.

I have used Perplexity to read Chinese characters on my screen, asking to point out the radicals, tone, and meaning of unfamiliar characters. There are minor differences between what Perplexity answers with and what Duolingo and DuChinese deal with, but I know enough Chinese* to figure out the difference - though a recent study calls the accuracy into question.

For instance, below I've asked it what the radical is in 水 (shui, water)

 

 

 [alt: The character in your image is 水 , which is the Chinese character for "water."
Radical in This Character

- The radical is 水 itself, pronounced "shui"] 

Spoiler for a 5000 year old language; the entire character is the radical.   

*Slightly more than beginner level. Hooah! 

 I also asked Perplexity to identify information about e-books in my third-party e-reader on my phone; The formatting is often wonky, so if I ask it what book is on the screen, it may not always be correct. I even showed the cover of a book and it got it wrong, mainly because it focused on the largest word on the cover...which is also the name of a book in the series. 

The series is 'Redwall', the title of this book is Rakkety Tam, and the author is Brian Jacques. Perplexity called the book 'Redwall', which is the first book in the series.

 

Yep, that squirrel has a sword.

One test I did was ask Perplexity to identify the hexadecimal code of a painted wall. A hexidicimal or hex color, is a 6-character string following a hashtag that dictates a color online. The letters go from A-F and the numbers go from 0-9. White is #FFFFFF and Black is #000000.

 

[alt: the predominant hexadecimal color in the image is: #AEA9A5
RGB Value: (174,169,165)

The color is a soft, light grey with a slight warm undertone, commonly found in neutral]

This was interesting; With light and shadow, it comes out much more grey on the screen. In person, it's basically "New home in the 2020s Beige." I can't fault it for that. 

 I then asked perplexity AI to identify a character on the screen and then identify the character; the characters name was below the image.   It first described the image (“it appears to be a golden dog with a hat”) and read the words (“in large blue letters it says Pompompurin”)  

And then proceeded to tell me information about the character. If it hadn't read the name, would it have known? I bet it read the words and went to search the internet or it's training data. 

I went to MondoMascots on Instagram, framing my screen so it would see the caption of the image above (A) but not the image, and the image below but not the caption (B); It only identified the unseen image of  A. Perplexity can be fooled.

Let's try something else with a doll photo; The assistant blocks the photo, but it got it correct.

 

[alt: The  Character on screen is Jinafire Long from the Monster High doll line. She is recognizeable by her long green hair, gold dragon-inspired accessories, and her distinctive red and purple Chinese-Inspired] 

It searched Pintrest and Reddit to come up with the right answer - I can see how, you can see a plethora of images without logging in, so it's easy for software to scrape.

It did not, however, correctly recognize the 3rd generation doll of the character which released much more recently than the first doll (2025 compared to 2014). If Perplexity is a modern search engine, is it searching the living web, or older data sets? 

 

The App Itself. 

 Moving on to using the app itself, the home screen looks like this;

 

It looks much like any talking machine trained on data. The explore tab has various categories.

[alt: Rencontres d'Arles  photography festival show...
The ancient city of Arles has transformed into the global epic...] 

 

 

 [Alt: 'Kpop Demon Hunters' becomes surprise Netflix s...
Netflix's animated musical "Kpop Demon Hunters" has emerged as...]

Interestingly, artscraftsteve is an individual who curated this particular news story in the app.  When you scroll down, you see an endless carousel of links, pulling from Reddit, YouTube, TikTok, Wikipedia, Screenrant, Koreaherald...I don't think Steve manually curated all of those, but I'd love to see how it's done, done, done.

 I moseyed on to the Entertainment section

 

(Bonus; In the notification bar at the top, you can see the Perplexity AI  logo I get for news alerts)

[Alt: The text bamsmackpow is on the image. Marvel reportedly developing Moon Knight sequel According to industry]

I'm not familiar with bamsmackpow, and asked what other sites reported on the news. Perplexity AI gave me about 6 other websites, some I recognized, none an official trade like The Hollywood Reporter or Variety. When I asked what makes the sources legit, it gave me a list of ways to check the authenticity of a source. 

I appreciate LLMs implementing things like this, and it's basically "It's not our responsibility what you do with this information, we're covering our bases." 

Spaces

There is a spaces option where people (or you alone) can collate information to reference. You can already place a prompt in there. 

 

 

 

Ask if you would like to join the wonderful world of tofu recipes!

I put in something akin to "You are a world-renown chef with a specialty in plant-based foods. ask me some questions about what I want to eat, what [aromantics] I have on hand, and what I may want to eat with tofu today."

Then, I asked it about tofu recipes in New Zealand (Aotearoa).

It returned information about recpies that focus on fusion cuisine, including an oven-baked peanut butter tofu, which prompted me to ask what kind of peanut butter they have in Aotearoa. 

Perplexity AI returned a bulleted list of three brands, with sources, although the final one apparently has such issues with dryness that people mention it multiple times across the internet. 

Next, I asked about tofu recipes in Taiwan. San Bei is a dish with tofu in a 1:1:1 sauce of toasted sesame oil, cooking wine, and soy sauce flavored with aromatics. That sounds good, and I save that query to my Space.

 

You can also ask Perplexity to search various sources; It's separate from the 'out of space' search on regular Perplexity. That's an interesting feature, but I can see using social searching to be problematic. We've seen it be questionable on searching recent information, and misinformation flies around social media. How does it discern something posted by the AP (Which is also dwindling in credibility) and someone's crap-stirring, sourceless text with 'BREAKING:' ?

 

 

I then made another space and gave it a prompt to find 10 Cloud engineering roles that had been posted in the past 24 hours, mentioning AWS or Azure that are remote in the United States.

It did, in a nicely formatted list. One job was closed. It could not send it to my email, but it gave me formatting to copy....and it's not easy to parse. 

Also, based on how some job sites are formatted, it will give me a link, but the link redirects to the list of every job listing 5 out of 6 times. I adjusted the prompt and the link issue still persists. It will need more adjusting, or maybe I've hit the limit of free Perplexity AI.

 Because there is Perplexity AI Pro - missed opportunity to say 'ProPlexity' - and I did not test that because I'm not paying for this, or any company that insists it's going to put people out of work. 

 Trip Planning on Perplexity AI Desktop

 Perplexity on Desktop is a little more thorough; In part to us using a larger screen.

Prompt:

What conventions are happening in Atlanta in September 2025?

 

 

Output:

 

First, it showed its thought process, or "thought" process;

 

 

 It listed a table that included neat things like an LGBTQ film festival, DragonCon, and a toy convention, however only some of these were in a table.

  

 

It's not impossible to read, but an odd choice. I'd prefer everything in two separate tables, and can probably prompt it as such.

 

Conclusion 

 Formatting leaves something to be desired if you get slightly complex with output.

Image identification sometimes works, but it doesn't appear to have enough data on certain newer things. 

The curated topics seem to imitate a social media feed, but with less interaction.  

 Overall, with refinement (or perhaps the paid option?), Perplexity AI can work for everyday use cases.