How I actually use ChatGPT and Claude in my work as a performance marketer. Not for writing ads. For the thing nobody talks about.

Everyone online is showing you how to prompt AI to write ad copy. That is the least interesting way to use it. The real value is somewhere else entirely, and almost nobody is talking about it. Here is what I actually use these tools for, every single day.

Let me start with a confession. For the first six months after ChatGPT blew up, I used it exactly the way everyone online said to use it. “Write me ten Facebook ad headlines for a dental clinic.” “Give me five email subject lines for a Shopify brand.” “Rewrite this landing page in a more persuasive tone.”

The output was always technically fine. Always usable. And always bad in a specific way that took me a while to put my finger on. It was bland. Generic. The kind of copy that looks like marketing instead of actually feeling like a human talking. I would use the AI output as a starting point, then rewrite 80% of it before anything shipped. Which meant the AI was saving me maybe ten minutes of blank-page staring. Not nothing. But not the “ten times faster” revolution everyone was promising.

Then slowly, over the next year, I figured out that I was using it for the wrong thing.

Writing ad copy is not the hard part of my job. Writing good ad copy quickly is a skill you build with reps, not with a tool. If you cannot write a decent ad in twenty minutes after five years in this field, no AI is going to fix that for you. It will just help you produce bad ads faster.

The hard part of my job, the part that eats most of my hours and most of my mental energy, is thinking. Thinking about the account. Thinking about the audience. Thinking about why something is working or not working. Thinking about what to test next. Thinking about how to explain a complex thing to a client in plain language. That is the real work. And that is where AI, used properly, has quietly become indispensable to me.

Here is how I actually use ChatGPT and Claude every day. None of it involves generating final creative. All of it involves thinking better and faster.

1. As a sparring partner, not a writer.

The single biggest shift for me was treating AI as a smart colleague to think out loud with, not as a tool to produce finished output.

Here is what that looks like in practice. A client’s lead quality dropped last week. I have a theory about why. Before I act on it, I open Claude and type something like: “I am seeing a 30% drop in lead quality on a dental clinic’s Meta Ads account over the last two weeks. The spend is the same. The creative is the same. The audience targeting is the same. My theory is that the drop is coming from a change in how the Advantage audience expansion is finding people. Walk me through three other possibilities I might be missing, and tell me which one you would investigate first and why.”

What comes back is not an answer. It is a conversation. It makes me defend my theory. It surfaces things I had not considered. It gives me a second brain to pressure test against, one that does not have the same biases I do, and one that will not get tired or distracted the way a colleague might at 11pm.

I probably have this kind of conversation ten times a day. Sometimes the AI is wrong. Often the AI misses context I have about the specific client. But even when it is wrong, the act of articulating my theory out loud, in writing, to something that will push back, makes my own thinking sharper. I end those conversations with a clearer idea of what to do than when I started.

This is the use case that nobody talks about because it does not produce a shareable output. There is no carousel to post. There is just better thinking. But if I had to pick one thing AI has changed about my work, it is this. I think out loud with it, and I think more clearly because of it.

2. To interrogate my own data before the client does.

The second biggest use for me is as a sense-check on data I am about to show a client.

I pull a report from Google Ads or Meta. Before I send it, I paste the raw numbers into Claude and ask it to tell me what story the data is telling. Not what to write in the report. What the data actually says. I want to know if there are patterns I missed, contradictions I did not notice, numbers that look fine on the surface but hint at a problem underneath.

For example, I will paste in a 90-day performance table and prompt something like: “Here is 90 days of campaign data from a client account. Spend is up 20%, revenue is up 25%, but new customer acquisition is down 10%. What is actually happening in this account? What should I be worried about? What should I ask more questions about before I present this?”

Half the time, Claude catches something I missed. The other half, it confirms what I already saw, which is useful because now I know I am not fooling myself. Either way, when I walk into the client meeting, I have already stress-tested the story. The client cannot ask me a question I have not already thought about.

This has quietly made my client meetings much better. I used to walk in with the data and the explanation. Now I walk in with the data, the explanation, three questions the client might ask, and my honest answers to all of them. Clients notice this even if they cannot articulate why. It builds trust fast.

3. To turn complexity into plain language.

A huge part of my job is explaining complicated things to people who do not have time to understand them. Business owners do not want to hear about attribution models, bid strategies, or incrementality tests. They want to know whether their money is working.

I use AI constantly to help me translate. I will write out a technical explanation the way I would say it to another marketer, then paste it into ChatGPT and say: “Rewrite this for a business owner who runs a dental clinic. Assume they are smart but do not know marketing. Keep it under 150 words. No jargon. No marketing cliches. Make it feel like a real person talking, not a brochure.”

The output usually needs a pass from me, but it gives me a much better starting point than my own first draft. Because the problem with explaining your own work is that you are too close to it. You cannot see which words are jargon anymore. You cannot feel which sentences would lose a non-expert. An AI, prompted properly, can.

Over the last year, this one use case has probably doubled the quality of my client communication. Clients understand what I am doing more clearly. They have fewer questions. They trust the work more because they can actually follow it. That shows up in retention, in referrals, and in how often I have to re-explain the same thing in month six.

4. To speed up research without replacing judgement.

When I take on a new client in a vertical I have not worked in before, there is always a research phase. Understanding the industry, the common buyer objections, the competitive landscape, the regulatory stuff, the language the audience actually uses.

Used to be, I would spend two or three days reading forums, Reddit threads, industry blogs, customer reviews. I still do a lot of that. But now I use AI as a fast first pass. I will prompt something like: “I am starting work for a boutique physiotherapy clinic in Bangalore. Walk me through the typical buyer journey for a patient looking for physiotherapy. What are their biggest objections? What questions do they ask before booking? What makes them trust a clinic versus walk away? What is the typical price range they expect? What claims should I avoid making for regulatory reasons?”

The output is a starting map. Not the final map. It gets a lot right, misses some things, and occasionally gets stuff wrong that only domain expertise can catch. But it saves me the first two days of blind research. I can then go verify, correct, and deepen what the AI gave me, instead of starting from nothing.

The mistake people make here is thinking the AI output is the research. It is not. It is the scaffolding you build the real research on top of. Treated that way, it is incredibly useful. Treated as the final word, it will make you confidently wrong in front of a client who knows more than you.

5. To write the first draft of anything administrative.

This is the least glamorous use case and probably the one that saves me the most hours every week.

Client emails. Proposal outlines. Internal SOPs. Report summaries. Agendas. Follow-up notes. All the writing that is not strategic work but still has to be done, and still eats time if I do it from scratch.

I give AI a rough bullet-point version of what I want to say and ask it to produce a first draft. Then I edit the draft to sound like me. The edit always takes a fraction of the time the original writing would have taken. The output is not creative. It is not voicey. It is not meant to be. It is meant to be clear, functional, and off my plate.

If you are a solo operator or running a small team, this stuff is the quiet killer of your week. The deep work does not get done because the admin writing filled every gap in the day. Using AI to handle the admin writing has given me back probably eight to ten hours a week of real thinking time. That is the actual compound. Not better ads. More hours to think.

What I do not use AI for.

Just as important as where I use it is where I do not.

I do not use AI to write final ad copy that goes live. The output is too generic, too safe, and does not have the specific angle that comes from understanding the client and the market in a way the AI cannot. I use it to brainstorm angles, to pressure test headlines, to sanity check hooks. But the actual copy that spends money, I still write myself.

I do not use AI to make strategic decisions. It can help me think. It cannot decide. The decision of what to bid, what budget to allocate, what campaign to kill, what client to walk away from, is mine. The responsibility is mine. The AI is a thinking partner, not a decision maker, and treating it like one is how people make expensive mistakes.

I do not use AI to replace talking to clients. There is a trend of using AI to summarize calls, generate talking points, auto-respond to emails. I do not do any of that. The relationship with the client is the actual asset. Automating the relationship away is a short-term convenience and a long-term disaster.

The pattern underneath all of this.

If you notice, almost everything I use AI for is about thinking faster and more clearly, not about producing more output.

That is the pattern. The people who are getting the most out of AI right now are not the ones producing ten times more work. They are the ones producing the same amount of work with ten times more thinking behind it. The AI is not the writer. The AI is the second brain.

The marketers who are falling behind are the ones using AI to replace thinking. They are generating more content, more ads, more copy, with less thought per output. The volume is up. The quality is down. Clients eventually notice, even if they cannot explain why the work feels off.

The marketers who are pulling ahead are quietly using AI to think better. To stress test their logic. To translate between audiences. To interrogate their own data. To speed up research. To handle the admin so the deep work gets more hours. None of that shows up on LinkedIn as a flashy use case. All of it compounds.

What I would tell someone starting with these tools.

If you are just starting to use ChatGPT or Claude in your marketing work, I would say this.

Stop asking it to write things for you. Start asking it to think with you. Change your prompts from “write me X” to “help me think through X”. The outputs you get back will be different, more useful, and much harder to replace with the next person who learned the same prompts you did.

Treat it as a smart colleague you can interrupt at 2am. Not a replacement for your own skill. Your skill is still the moat. The AI is the scaffolding that lets you deploy your skill faster and more clearly across more work.

And remember, the AI does not know your client. It does not know your market. It does not know the specific context that only you have. Everything it gives you is a starting point, not an ending point. The marketers who forget this produce confident garbage at scale. The ones who remember it quietly do better work than they ever could before.

That is the whole shift. Not writing faster. Thinking better. If you can hold that distinction, these tools become the biggest advantage you have ever had. If you cannot, they will quietly make you average, faster.

If you want to swap notes on how you are using AI in your own work, my inbox is always open. I am genuinely curious to hear what other people are figuring out.

hello@hukumyadav.com