🤖 My AI Showdown: ChatGPT vs DeepSeek vs Perplexity vs Gemini vs Grok

Hey everyone! 👋 Vengat here.

Now, let me ask you…
When the internet first came, what did we even do with it?

Well, it wasn’t for reels, youtube videos, or meme wars. The internet came so you could search for something you needed. That’s it. Back then, you had browsers, you had email… and that was basically the whole party.

Then came Internet Explorer (RIP), followed by a bunch of browsers. People didn’t even know what exactly to do with them. they just clicked random stuff and scrolled aimlessly. Slowly, the internet transformed into a place where you searched for information.

Then, it became social media, we know that story.

And then… About two and half years ago, AI was just a sci-fi movie concept. Today, it’s in everyone phone as an app. ChatGPT, Gemini, Copilot, Grok, Perplexity, DeepSeek, etc.… it’s like a new one is born every day.

The way we use the internet has changed so much, we don’t even open a browser anymore. If you need something, you just talk to your AI and get the answer.

So, I decided to do what I do best, Test Everything.
Here’s my personal, no-nonsense breakdown of which AI engine I liked, what’s great, what’s useless, and which one I didn’t even bother with. (NOT TECHNICALLY)

💬 ChatGPT – The One That Started It All

ChatGPT was the first big name in this game.
People had no idea what AI really was.. then suddenly, “Hey.. There’s this company called OpenAI… check out their ChatGPT thing!”

We’ve all seen chatbots before (Yahoo Messenger, anyone?), but this was different.

• You could ask anything

• It answered like a human

• You didn’t need a browser

• You could even generate images, videos, etc

That mix of “chat + search” was incredibly powerful.

And of course, Sam Altman got all the fame and criticism. News headlines were like “He’s going to control the world” and “This is dangerous technology”… and yet, we all kept using it.

I still use ChatGPT until GPT-4, it was my favourite. Two days before GPT-5 launched, I tested a few things… and wow. Let’s just say we’ll talk about that in a separate blog..

The problem? The free plan is more stingy now. After 5-6 questions, it tells you to get the pro plan. That’s when my daily use dropped.

I did find a tricky little method to get GPT-4 & GPT-4 Turbo features for about a year and now even GPT-5 unlimited access without a subscription… but that’s a hidden story for another day.

🔍 Perplexity – The Madras Boy’s AI

This one is special, made by Aravind Srinivas from Madras
The best thing? When you ask a question, it also shows 4 related questions right below. Just click and go deeper. You don’t have to think “What should I ask next?” — it does it for you.

Accuracy? Pretty solid. Great for research works.

The downside? You can get lost in it. Like Wikipedia. You start on one topic, click a link, then another, and before you know it, you’re reading about how penguins sleep.

Also, in Android, the scrolling is a pain. The question-answer chain goes so deep, you have to keep scrolling like your thumb is training for a marathon. The mobile UI needs fixing.

They gave me a free 1-year Pro plan worth of ₹18k, in that I tried their “Deep Research” feature… but honestly? The free plan already had most of the good stuff.

And I respect this part — the founder himself said in an interview:

“We’re not competing with ChatGPT. We see AI as the next web search.”

This man even said he was ready to buy Chrome when Google faced a lawsuit. That’s when I realised… these AI startups have grown HUGE in just two years.

If I have to ask many questions very quickly, Perplexity beats ChatGPT for me.

🏆 DeepSeek – My Daily Driver

Ah… DeepSeek. My personal favourite.
Chinese company, open source, unlimited free text queries… basically perfect for me. I use it every day.

Why I love it:

• No question limits like ChatGPT

• Great for DeepThink mode — it actually thinks step by step detailly about your input before answering

• “Deep Research” option gives extra-detailed answers (like Gemini)

For text-based questions, DeepSeek beats everyone for me.
And it has a mode called “DeepThink (R1)”, I’m ADDICTED to it. The accuracy level for complicated topics is unreal. Perfect for people who want 99% accuracy level super-detailed answers.

Only one catch - Politics...
Ask about China, China Politics, Karl Marx, Mao… suddenly the answer disappears mid-sentence. Feels like someone hit backspace from behind the scenes. CENSORSHIP vibes.

Still, for text-heavy searches, DeepSeek > everything else in my daily use.

🌐 Gemini – Google’s AI Child

You can tell it’s a Google product instantly.

Sometimes, I throw the same question into all the AIs like ChatGPT, Gemini, DeepSeek, Perplexity just to see who wins.
- Often, ChatGPT and Gemini are tied in same place.
- DeepSeek struggles a bit and then gets there or match the same place.
- Perplexity gives mostly reddit feel, it has a browser tone. This doesn’t give you the feel of using AI like chatgpt.

But there was one day… I’d been hunting for a tiny book which I read in school for years. No luck. I asked ChatGPT and DeepSeek, but they gave me the wrong answers.

Then I asked Gemini.
By its third answer, it nailed it.. The title, details, even links to buy. That was a goosebumps moment.

And, What impressed me most about Gemini is that Google introduced something called Google LLM Notebook. (My personal favorite)
If I upload a large PDF, say 300 pages or more, it analyzes the document and answers my questions almost instantly. It completed ten days of work for me in just two minutes. This also works for YouTube videos and technical podcasts I listen to in multiple languages like English, Hindi, and Tamil, from various experts in my field of interest. These podcasts are usually three to four hours long and in languages like Hindi and Telugu, which I don't understand. YouTube doesn't have subtitles or auto-translation for these, but I'm interested in the content. The speakers are well-known in their fields, and I want to follow them, but I can't understand what they are saying. So, I upload the link to the Notebook LLM, and it provides the entire podcast translated transcript, even if it's three to four hours long, in just a few minutes.

The translation quality is on another level. Google's language models, especially when it comes to Indian languages like Tamil, Hindi, Telugu, it outperform everything else I’ve seen. Google Translate already set a high standards, but Gemini takes it further. Its fluency, contextual understanding, reasoning, and domain-specific adaptation in the Notebook environment feel almost human.

The technique behind this is flawless not just in terms of language translation, but also in understanding domain-specific vocabulary, context, tone, and intent. The translation isn’t just literal but it captures the technical meaning, the emotional tone, and the intent of the speakers with impressive precision. It feels like reading a properly written summary, not just machine output. Even domain-specific terms like in engineering, medicine, or science are translated with remarkable accuracy.

Awesome

Gemini’s big win? No free plan limits. You could ask 10,000 questions, and Google won’t complain because what it needs is your DATA. Sure, it’ll casually suggest you buy Pro, but you can ignore it.

🎭 Grok – The Meme Lord AI

Now this one… is a completely different vibe.

Grok, developed by xAI and integrated directly into X platform (formerly Twitter). Grok stands out with its distinctly sarcastic, meme-aware, and irreverent personality. Unlike traditional assistants, Grok doesn’t shy away from internet humor because of its trained datasets, which includes real-time data from X. That means it’s been trained on actual social media posts like unfiltered, raw, and often chaotic. Its heavily on X's tweet content, making it fast with real-time information and a sarcastic sense of humor. There’s also a standalone app if you want a less X platform noisy experience.

What Makes Grok Different is, unlike traditional AI models that use carefully selected text from books, news articles, or formal websites, Grok is exposure includes:

• Slang, sarcasm, and street language

• Code-switching between English and regional languages

• Internet-specific humor, trends, and cultural conflicts

• Abbreviated writing, typos, emojis, memes, and GIF culture

This gives Grok an unique edge in understanding how people actually communicate online not just what they say, but what they mean.

Where Grok really shines is especially for Indian users, is in understanding mixed-language conversations, aka "Tanglish" (Tamil + English).
For e.g. “Dey, enna da solra? Total waste-u nee😂” Grok gets it. Tone, emotion, cultural context - all of it.

What about profane language, like swear words in English, Tamil or any other languages? Yes, it knows, and not just knows them, it can use them too. Let’s be real, people don’t always use like textbook words on online platforms like X.

Grok’s training data includes vast amounts of real-time, unfiltered content from X (formerly Twitter) and that comes with the full spectrum of human expression likes sarcasm, memes, slang, regional dialects, and yes, even explicit or profane language, including in Indian languages. That doesn’t mean it’s offensive by default — but it understands the language of the internet in a way most AIs don’t.

But don’t let its casual tone fool you, Grok is technically solid under the hood and provides strong competition to its rivals.

Trained extensively on public posts from X, Grok has an edge when it comes to real-time awareness. It taps into the social media stream and augments it with information from external web sources to provide context-rich, up-to-date answers.

I tested it:

• Asked about Chennai’s weather – it analyzed 12 X posts, cross-checked with websites and official meteorological sources, gave a neat, accurate summary.

• Asked for Python code — wrote it instantly.

Extra features in the app:

• Deep Search – for detailed lookups

• Think – for more thoughtful, reasoned answers

Grok feels like that cool friend who answers your question but also cracks a joke at your expense.

❌ The Ones I Skipped

I didn't try Meta AI or Claude. Meta AI has privacy concerns, and Claude has a limit of 50 questions per day with a token-based nonsense… which isn't worth it for me when DeepSeek is available.

📌 Why I Still Use AI Daily

• Understands queries even with spelling mistakes or broken English.

• Finds in minutes what takes days in a browser.

• No ads or annoying hyperlinks.

• Fast, straight-to-the-point answers.

But… I avoid using it for emotional or psychological advice. AI will just agree with you and tell you you’re right — even if you’re wrong. That’s dangerous.

🎯 Final Verdict

My AI hierarchy right now:
DeepSeek > ChatGPT = Gemini = Perplexity = Grok > Others AI’s

Browser? Rarely. AI has replaced it for me, unless I need very local info.

AI isn’t just a fad. It’s one of the most useful things right now. But whether it makes us smarter or lazier… that’s up to us.