Stop Letting ChatGPT Do Your CAT VARC Thinking

Stop letting ChatGPT do your CAT VARC Thinking: I am almost 100 per cent sure you are using Gen AI for your prep. Everyone is. The blog you are reading right now probably has some footprint of a Gen AI tool somewhere. I am not pretending otherwise.

And honestly, there is no problem in using these tools. Especially for VARC. A good AI tool can do things in a few seconds that used to take an afternoon.

But that is also the problem. Because it is fast, because it is confident, because it sounds right, you stop noticing where it is quietly taking you in the wrong direction.

Let me take you through five ways Gen AI actively damages your VARC prep, even when it feels like it is helping.

1. Condensing a 4500-word Aeon Essay Into 500 Words

This is the most common misuse I see. You open an Aeon essay. You scroll down. Then you see it. 4500 words.

“Arrrggghhh… 4500 words!!! Nah. Let me get AI to condense it to 500 words”.

Bad move.

Here is what you are missing. CAT passages are not just “short articles”. They are carefully edited extracts from real, messy writing. Someone has taken a longer piece and cut it down while preserving the original writer’s density, awkward clauses, embedded arguments, and occasional shifts in direction. The passage feels a certain way because a human wrote it that way, and another human preserved that texture.

Now look at what AI does when you ask it to condense.

It gives you clean topic sentences. Parallel structure across paragraphs. Tidy signposts at every transition. Even density throughout. No awkward clauses. No shifts that make you pause. Everything that makes CAT passages hard, AI removes by default. That is how it is trained to summarise.

So you end up reading 500 words of nicely behaved prose and telling yourself you are “practising”. You are not. You are practising a genre that CAT does not test. You are getting better at reading AI, not at reading what CAT actually puts in front of you.

If the essay is long, read the long version. Slowly. Doodle paragraph by paragraph. That is the training.

Suggested Readings –

2. Asking AI the Moment Something Feels Hard

You read one paragraph of a philosophy piece. It did not open up. You tried again. Still fuzzy. And then you do the thing.

“Uff… what is this guy even saying? Forget it. ChatGPT Bro, explain this to me in simple terms.”

The AI obliges. Of course it does. It gives you a neat, simplified version. You read it, nod, feel that you have “understood”, and move on.

But you have just skipped the exact struggle that was going to build your reading.

CAT does not test whether you can read something after someone has already broken it down for you. CAT tests whether YOU can break down something unfamiliar, on your own, in a 40-minute window, with no assistant sitting next to you. The only way you build that capacity is by sitting with confusion and working through it.

Every time you tap AI at the first sign of discomfort, you are training yourself to outsource the hard part. And the hard part is literally the skill.

Add Career Launcher as a Trusted Source   

3. Running to AI the Moment You Get Something Wrong

You solved a para jumble. Your answer was 1324. Correct answer is 1342. You are pissed.

“What?! How is it 1342?! 1324 makes way more sense! Let me ask AI…”

Within 30 seconds, the four sentences are pasted in, and you are asking, “Why is 1342 correct and not 1324?”

The AI answers. Cleanly. Confidently. You read the justification, feel a little better, move on.

What did you actually learn?

Nothing that will transfer. You got an explanation. You did not build reasoning.

Here is what you were supposed to do.

Go back to your 1324. Defend it. Properly. Pretend you are the one marking the paper and have to mark your answer wrong. Where does the link actually break? Does sentence 3 flow into 2, or does 3 need something before it that only 4 provides? Fight with it. Let the discomfort stay.

Only after you have genuinely wrestled with it, then check. And even then, AI is not the first stop. The official explanation is.

The moment AI becomes the resolver of every wrong answer, your analysis muscle stops growing. You feel productive because you are “clarifying doubts”. But clarity that someone else hands you is not the same as clarity you built for yourself. The exam hall will ask for the second kind.

Also Read: Common CAT Mock Mistakes To Avoid | How to Approach CAT Mocks 

4. Asking AI to create RC passages and questions for you

This one is fashionable right now.

“Claude Bhai, create me an RC passage with four CAT-style questions. And make the options tricky.”

The AI will do it. Instantly. And on the surface, it will look fine.

But here is what you cannot see. What makes CAT questions CAT-questions is the options. Specifically, the wrong options. The scope distortions, the tone shifts, the extreme language, the twisted versions, the false connections. These traps take a trained human setter a lot of effort to construct well. They are built on a deep reading of the passage and a deeper understanding of where an aspirant’s reasoning is going to slip.

AI can mimic the surface. It cannot reliably build the traps. Not at CAT level. Not yet.

Now, to be fair, I am pretty sure actual question setters also use AI in some form. I am guessing here, I cannot prove it. But it would be naive to assume nobody at that level touches these tools.

The difference is in how they use it.

A setter has already read the passage deeply. They have already worked out the author’s argument, the structure, the hinge sentences, the places where an average reader is likely to misread. They have already thought about what kind of question will expose that misreading. They have already drafted an option that looks right but quietly distorts the scope. After all that work, maybe then they turn to AI to tighten a phrase, check if an option is unambiguous, smooth the language. Fine tune.

The AI is not making the question. The setter is. The AI is polishing.

When you type “create me an RC with four CAT-style questions”, you are skipping every single step that actually produces a CAT-grade question. You are asking the polisher to be the builder. Those are not the same job.

You end up practising on questions where the wrong options are obviously wrong, and the correct one stands out. Your accuracy on that material feels great. Your real accuracy against actual CAT questions does not move.

Stick to actual CAT papers. Stick to material from serious sources. If you must use AI-generated text, use it for reading exposure, not for option-level training.

Also Read: Why solving CAT PYQs is the ultimate game changer

5. Asking AI for Prep Advice

“Gemini Yaar, my scores are just not improving. What do I do?”

You type this into AI. It gives you a confident, well-structured, kind-sounding answer. Analysis matters. Consistency matters. Read more. Build temperament. All correct sounding. All generic.

Now you get smart.

Of course it is generic. I gave it nothing. Let me feed it everything.

So you do. Last ten mock scores. Section-wise breakups. Question-level accuracy. Attempt patterns. Time spent per passage. Every number the platform will give you.

You ask again. Now the answer looks tailored. It talks about your VARC dip in mock 6. Your trouble with the third RC. Your drop in accuracy after the 25-minute mark. It sounds like it is finally seeing you.

It is not.

Scores tell you what happened. Not why. They cannot tell the AI that in mock 6 you were exhausted going in. That on the third RC, your eyes glaze over after two paragraphs. That your accuracy drops after 25 minutes, not because of stamina, but because you start second-guessing options you had already rejected. That three questions you got right last week, you actually guessed. That two you got wrong, you knew, but changed your mind in the last ten seconds.

None of that is in the data.

And honestly, most of it is not clearly visible even to you. You have to sit with your paper carefully, question by question, reconstructing your own thinking, before some of it even surfaces. That work is irreplaceable. It is also the one thing AI cannot do for you.

So the AI fills the gap with the most plausible-sounding story. That is what language models do. They produce a confident version of “probably.” You take that and build your next two weeks around it.

I wrote separately about the kinds of advisers you should watch out for. AI belongs on that list. Not because it cannot process your data. It can. The problem is that your data is a thin picture of you, and AI has no way of knowing what is missing.

So it gives you the most plausible answer for the numbers you showed. Which is still nobody’s answer in particular.

Don’t ask the AI what to do with your prep. Do the analysing yourself.

So, How Should You Use it?

Do not treat the AI as your teacher. Do not treat it as your boss. Treat it as your assistant.

An assistant does not decide your direction. An assistant does not do your thinking. An assistant checks, confirms, helps you verify. The judgment stays with you. The effort stays with you. The understanding has to stay with you.

Two examples of the right use.

You read a 2000-word article on some abstract topic. You wrote a one-line doodle for each paragraph, in your own words, describing what that paragraph was doing. Now you want to check whether your doodles actually captured what was happening. You paste the article and your doodles and ask, “Do my one-line summaries capture the main idea of each paragraph?” If it says yes, good. If it flags a paragraph, go back and see what you missed. Clean use.

You solved an RC question. Got it wrong. Went back. Defended your option. Attacked the correct one. Used the S.T.E.A.L.S. framework on your wrong choice and worked out that it was a Scope issue. You are fairly sure of your reasoning but want a second pair of eyes. You check with the AI. It confirms, or it adds a nuance you had not considered. Clean use.

Notice the pattern in both. You did the work first. AI came in at the end. You used it to verify, not to generate.

The day that order flips, the day the AI starts doing the work, and you are just copying the output into your notebook, your prep is finished. You will feel busy. You will feel informed. Your scores will not move.

Use it. But you be the one thinking. AI is brilliant at polishing your thinking, terrible at replacing it.