Everyone Can Spot AI Writing, Here's How To Fix It transcript

Today, AI can write everything for you. That's why you sound exactly like everyone else. And standing apart from others is one of the most valuable things you can do for your career, your relationships, and your future. So, how do you write absurdly well with AI but stop sounding like ChatGPT and still write like a human being? I will share a five-step AI writing system that will make everything you write clearer, more insightful, and unmistakably yours.

So, let's start with the story that changed AI forever. Move 37. It was March 2016, and that was the day when one of the strongest human players of a board game called Go was playing against AI. Lee Sedol had held the world championship titles maybe 18 times. And Go is this 3,000-year-old game. It looks very simple. Black stones, white stones, uh 19 by 19 board, but there are roughly 10 to the 170 legal board positions.

That's more than the number of atoms in the observable universe. Millions of people were watching, and midway through the game at the 37th move, Lee walked out for a break. He wanted to smoke a cigarette and come back. And when he came back, he looked at the board, and he couldn't believe it. AlphaGo had just played a move that looked like complete heresy. That was move 37. And 100 moves later, AI won the game. So, AI can do all of that.

Then, why does AI that we use today write in such a predictable generic fashion? When we understand why that happens, we can become better writers, and more importantly, we can become better thinkers. The AI that we use daily, like Claude or Gemini or ChatGPT, are called large language models or LLM. Now, those are built very differently than AlphaGo. So, there's a big gap between what AI generates and what you should generate.

And that gap creates five problems in the way we all communicate. So, let's cover those five pitfalls next, and then we'll come up with a five-step system that helps you navigate around those five problems. Problem number one. Confidently drunk. In May of 2024, uh user asked Google, search how to make sure that cheese stays stuck on a pizza. Now, Google's AI overview suggested adding non-toxic glue to the sauce to make it stickier.

It's true story. Now, AI understood words from the query. It understood pizza, cheese, sticking, and then it found a joke from Reddit and thought that was the solution. AI hallucinates, and it is so confident while it's doing it. It's like having a conversation with Albert Einstein when he's drunk. He's smart, but he will just make things up, and he won't know that he's doing it. So, that's problem number one. The second problem, a broken record.

AI learns patterns from the internet. Right? So, it predicts and remixes those patterns all the time. So, it creates structures that are familiar and comfortable. And it's stuck doing that because, well, that's the key feature. Problem number three, smart and empty. AI is extremely good at producing lines in a shape of an insight. For instance, love without structure is a liability. I don't know what that means. Again, words are fine, and it sounds profound, and it's pretty smart and deep, or at least it sounds like that.

But, what did that actually mean? Because the moment you take a closer look at it, it all falls apart. Problem number four, clarity lost in noise. You ask AI to say something simple, and it might come back with some abstraction or jargon. For instance, it will happily create this text. Professional growth requires the intentional strategy and development of capabilities that align with emerging opportunities across an increasingly dynamic career landscape.

That's a mouthful. And there is a real idea somewhere in there, but it gets lost in these abstractions that AI generates. If you read it again, the idea is short and simple. Learn things people need. That's it. Problem number five, AI is boring. AI can give you a perfectly good line, but it won't give you any reason to read the next one. AI writing has no tension in it, no intrigue. So, it's all predictable and mostly boring.

So, those are the five problems. Of course, the solution is not to stop using AI. That's inevitable nowadays. But, you got to use it in a way that eliminates these shortcomings. That requires a framework. So, let's walk through the five steps of that framework. I'll also share some example prompts you can use for each one of them or at least four of them. Together, they spell the thing you're trying to create, your unique voice.

AI gives you an accent, and here's how you get your voice back. So, let's go to the first one. V is for verified. Here, you're fighting the AI hallucinations and asking, "Is this real?" In 2023, a lawyer named Steven Schwartz used ChatGPT to research a legal brief. And of course, ChatGPT gave him cases and citations and quotations and judicial opinions, but there was only one problem. None of those cases existed. When the opposing party's lawyers couldn't find these cases in the real world, Schwartz went back to ChatGPT and asked if these cases were real.

ChatGPT assured him they were. So, it's like hallucinations on top of hallucinations. And of course, this is a 3-year-old story, and AI has gotten much better. But, when it comes to numbers and names and data and dates and research, my first instinct still is to verify what AI generates. I look for links. I look for citations. And I ask AI to verify any claim. And the move here is to treat AI like a researcher whose work requires a supervising editor, you.

So, for example, here's a prompt. You can give it a draft that you're writing and ask AI to do the following. Extract every factual claim, statistic, data, date, facts, quotes from this draft. Mark each of those items verified, unverified, or disputed. Go out on the internet and find exact links and citations that verify each claim. Give me the strongest primary source for each. Give me a list of claims and their verification links and citations in a bullet form.

That's it. And on all three platforms, there are two tricks. One, you can turn the deep research mode on. And AI will give you a plan of how it's going to go out on the internet and what sources will it research. And the second trick, cross-verify. Ask Claude to verify Gemini's work and Gemini to verify ChatGPT's work and so on. And by the way, I talk about these frameworks and systems and go deeper in my weekly newsletter.

It's every Tuesday morning. Please subscribe. I love writing it and it's totally free. And now let's go to the second step. O is for owned. In a Cornell University researchers ran an interesting experiment with 118 people, 60 in India and 58 in the US. Each person had four writing prompts. Your favorite food, your favorite public figure, your favorite holidays, and so on. Half of them were asked to type it themselves.

The other half typed with the assistance of AI suggestions. And by the end, Indians writing with AI were producing more Americanized essays than Indians writing on their own. Indians who used AI, their favorite food was pizza. Their favorite public figure starting with S was Sylvester Stallone, and not Shah Rukh Khan. Their favorite holiday was Christmas. Now, nobody made them do it. But, AI made all of it homogenized, and generic, and ordinary.

The familiar. The writers lost their cultural nuance. So, the key question here is does your writing, or any kind of communication, have your fingerprints, your unique touch? What makes you extraordinary? The right move here is to give AI the raw material from your life, because that's your fingerprint. Something that others won't have. So, a prompt would be like, "Do not write anything yet. Interview me about this topic.

Ask me one question at a time about what I have personally seen, or done, or believed, or learned, or got wrong. Push back whenever my answer is vague. Stop only when you have five specific details that are relevant to the topic, and have come from me." So, that's all for owned. Now, let's go to the third step, which is I for insightful. Is there actually a so what behind your idea? That's the question you should be asking.

Let's say you're making a LinkedIn post and AI comes up with something like "High performers succeed because they extract wisdom from failure." Okay, it sounds fine, but you should ask "Who in the world is arguing that successful people learn nothing from failure?" No one. There's no unique point of view. Everybody knows that failure matters and it teaches you lessons. So what do you do? There are three quick moves that can get you out of this mess.

First, bring your own experience. For instance, you could have said "Last week I got hired after 4 months of grueling job search." Those 4 months taught me things that were more valuable than actually getting the offer letter. It's the same message, but there's you in it. There's something unique. Another way to bring something insightful is to go deeper but inside. Before I write anything, I often force myself to finish three sentences.

One, most people believe in XYZ and whatever it is. I believe in ABC. And the difference is 1 2 3. Now, you don't always have to look for the counterintuitive insight. Until you find that unique angle, you probably have a topic, but you still don't have an insight. And the third move here is to use AI as your opponent. Lots of people use AI to manufacture an insight, but you should give it your insight and then ask AI to attack it.

You could prompt AI like this, "Act as an intelligent skeptic. Here's the point I'm trying to make." And you put your point there. "What would you disagree with and why? Attack my point of view by asking, 'So what?' until you're satisfied." All right. So, the next one is C for clear. Will they understand what you're saying? That's the question to ask. Many years ago, I was at a company where I was reporting to the CEO.

And he and I were sitting there during my um annual review. And he was such a thoughtful leader with a very high EQ. He had a PhD in economics, brilliant brilliant guy. And he gave me the most useful feedback I've ever received. He said, "Sandeep, you have this habit of hiding behind your humor. You deflect. You use sarcasm instead of candor. It's going to make people unsure of who you are and whether they can trust you." E- Even right now, as I remember that moment, I I mean, I can feel the weight of those words because they hit me like a ton of bricks.

I had no idea. I thought I was being warm and and funny and non-confrontational, but what was being seen by others was completely different. You're the last one who can judge your own mode of communication and how it gets interpreted by others is not in your control. So, if you're trying to create more clarity in today's noise-infested world, here are the three moves you can make. First, try to say it out loud when you trying to write.

Speaking is faster than writing, so you'll reach for simpler words that you know instead of the words that sound impressive when you write. Once you say the sentence out loud, then just write down exactly what you said. You'll see that it'll keep your writing style very conversational. Second, sleep on it. It's always good to come back to it the next day because a day later, you've forgotten what you meant. So, you read what's actually there.

You'll see if it's clear. And finally, for important documents, have another person read it. Now, you may not have this luxury all the time, but having a true independent review for at least important documents can be extremely helpful. Worst-case scenario, use AI as your editor. Here's the prompt for it. Act as my editor. First, tell me what I said in your own simple words as if you were explaining it to someone who is 9 years old.

Second, mark every sentence where the reader may have to guess my meaning. And third, tell me where there's room to improve my clarity. Clarity matters with people, but it matters even more when you're talking to AI. That's where our sponsor comes in. This portion of the video is sponsored by Hostinger. Let's say you want to build a personal career website or a website for your creative or coaching business. Today, it's gotten ridiculously easy to build it.

Set up your hosting and domain account on Hostinger and then connect it to your AI. For instance, Claude code. If you type this one-line command in Claude code, it'll install the connector, restart your session, and the plugin is right there. Simple as that. From here on, you build your site in plain English. Build me a clean personal website that features my bio, my projects, my experience, and a contact page. Oh, make sure it looks great on mobile.

That's it. Claude will build a site, Hostinger will get it live on your domain. All you have to do is give clear instructions. That's the power of clarity. Check out the Hostinger link below and use the code clarity26 to get 10% off. Now, finally, the fifth step, E for engaging. Now, this is the hardest because you have to ask yourself, is the next line needed at all? Steve Jobs had a phrase, "Oh, and one more thing." And he would just casually say that, but the audience waited for it because his "Oh, one more thing" was the most important thing of his presentation.

And that's what made him an engaging storyteller. Now, we mostly focus on whether our line is good or bad. We rarely ask the more potent question, "Will this line leave someone wanting the next one?" And the problem is AI loves fillers. It'll give you lines, it'll give you paragraphs, but it's your job as the writer to create a story, to create something at stake. Unfortunately, for this one, there is no prompt that can help any of us.

For this one, you're on your own because this one is all about your own taste and your judgement because this is the part that's entirely human and entirely yours. So, that's voice verified, owned, insightful, clear, and engaging. Also, you can take us with you on the go nowadays. These podcasts are available in audio and video on Spotify now. That's pretty cool. We started this video with a story of move 37. People in the AI world talk about that move a lot, but they rarely talk about something more interesting, move 78 in game four.

What happened in game four was Lee was already three games behind. He had lost three games, but in game four, he did something unconventional. Millions who watching didn't understand it at first. AlphaGo responded, but Lee won that game. Move was the machine surprising humans. Move 78 was a human surprising the machine. Writing is also about taking similar chances, about you taking a surprising turn and arriving somewhere new.

Because you don't write just to find words, you write to find out what you think. Writing is one of the best ways you meet your own mind. It's the best mirror you can have. That's why people journal. It allows them to reflect, literally. Content is infinite now. Anyone can produce words, but do you have something to say? Because when you write that, you'll get to learn something about yourself. Thank you, and I love you.