Stop Treating All AI Like It’s the Same Bloody Thing: A CMO’s Guide to Rejecting ChatGPT
Which LLM do you use and what for? Time to reveal all.
Listen up, dear marketers, because I’m only going to say this once: If you’re still asking ChatGPT to do everything from writing your brand strategy to generating Instagram captions to pulling competitive intelligence, you’re doing it wrong. Deeply, spectacularly wrong.
You wouldn’t ask your therapist, your bank manager, and your hungover intern the same question and expect equally useful answers, would you? (“Should I buy Bitcoin?” gets you three very different lectures.)
So why are you treating every AI model like it’s some interchangeable commodity?
Here’s the truth: The difference between a mediocre marketer and a brilliant one isn’t whether you use AI.
It’s whether you know which AI to use, when to use it, and more importantly, when to tell it to bugger off entirely.
Here’s everything you need to reach the stars.
Why This Actually Matters (And Why Your Career Depends On It)
Picture this: You’ve just used ChatGPT to draft a press release about your pharma client’s new product.
It’s eloquent, persuasive, absolutely brilliant.
You hit send. And then:
Three days later, you’re in a meeting explaining why your release claimed the drug “cures” something it only “treats,” and why you’re now facing legal action.
Or this:
You’ve asked ChatGPT to design brand assets. What you get back is bland, formulaic slop that every instinct in you screams not to share with another living soul.
The problem isn’t AI. The problem is you used a Formula One car to do your grocery shopping, then wondered why it didn’t fit in the parking space.
Every LLM has a personality, a specialty, and a dark side.
Learn, or get left behind.
The Restaurant Analogy That’ll Save Your Bacon
Think of LLMs like restaurants. You don’t go to McDonald’s for your anniversary dinner (do you?) and you don’t book Nobu for a on-the-go breakfast.
Context matters. Quality varies. Price points differ.
ChatGPT is a 3-Star Michelin restaurant — executive-friendly, intelligent approach, notoriously bad at bulk. This is your go-to for high-stakes strategic work: business strategy, topline summaries. When it’s Friday and your boss needs something brilliant by Monday, you reach for ChatGPT. When you need email variations by the hundreds, you absolutely do not touch it.
Claude is The Ritz — Elegant, precise, pathologically brand-safe. I use Claude to deliver anything that could get me fired: PR statements, regulated copy, internal communications that’ll be forwarded to the board. It’s witty without being risky, thorough without being reckless. It’s the LLM equivalent of having a very clever lawyer review your work before it goes out. The downside? It won’t give you visually striking campaign concepts or process enormous datasets at speed. Like the Ritz, it’s too exclusive to do any kind of mass production.
Gemini is Nobu — Stylish, visual, modern as hell, and they just dropped headline-worthy new features with Gemini 3.0 Pro. Brilliant for creating presentation decks, quick assets for nurturing client relationships, anything that needs to look as good as it reads. When you’re pitching and aesthetics matter, Gemini delivers. But don’t ask it to write your 40-page content strategy. That’s not what it’s built for.
Mistral and DeepSeek are Shake Shack and McDonald’s — Fast, consistent, scalable. These are your lo-fi, reliable workhorses for bulk production: ad variations, automated email sequences, social media captions at volume. They’re reliable, efficient, and absolutely not where you go for your breakthrough creative idea.
Grok is KFC — Real-time, edgy, culturally tuned in. When you need to know what’s trending on social media right now, the last conspiracy theory, or you’re building a reactive campaign around a cultural moment, Grok is your weapon. But keep it away from anything requiring regulatory approval or brand polish. It’s quick and dirty by design.
Perplexity is Sukiyabashi Jiro (of “Jiro Dreams of Sushi” fame)— Looks plain and ordinary from the outside, but the product quality is world-class. This is your research powerhouse. When you need deep factual research, market intelligence, competitive analysis, or trend validation with actual sources, Perplexity is unmatched. It’s not going to wow you with creative brilliance or come up with your next big campaign idea, but when you need to know the Truth about your market, it’s the most reliable tool in your arsenal.
The Hallucination Problem (Or: Why That Brilliant Answer Might Be Built On Lies)
Here’s what nobody tells you in the webinars: All LLMs hallucinate. Every single one.
They make things up with the confidence of a consultant who’s never worked in your category.
ChatGPT is moderately likely to invent statistics, cite studies that don’t exist, or confidently assert trends that are complete fiction. It’s creative and persuasive, which makes its lies particularly dangerous. Always fact-check.
Claude hallucinates less because it’s been trained to be cautious. It’s the safest bet for anything where accuracy matters more than creativity. But “less likely” doesn’t mean “never.”
Perplexity has low hallucination rates because it actually cites sources, but here’s the catch: if the source is rubbish, you get well-referenced rubbish. Garbage in, garbage out, just with footnotes.
Mistral and LLaMA have higher hallucination rates for factual content. They’re brilliant for pattern-based work and repetitive tasks, but terrible for anything requiring deep domain expertise or current market data.
The visual AIs like MidJourney and DALL-E hallucinate in a different way: they’ll create products that don’t exist, people who aren’t real, brand elements that were never designed. Which paradoxically is perfect for creative work.
The rule: The more creative the LLM, the more likely it is to bullshit you. Plan accordingly.
The Playbook: What To Use When
Let me make this dead simple for you.
For strategic brilliance: ChatGPT. Campaign strategy, positioning, storytelling, anything going to the C-suite. Just fact-check the data points.
For brand-safe precision: Claude. PR, regulated copy, internal comms, anything that could blow up in your face if a word is wrong.
For visual impact: Gemini for polished presentations, MidJourney for jaw-dropping creative, DALL-E for quick professional graphics (it’s not free). Pick based on whether you need “safe corporate sleek” or “stop-scrolling wowness.”
For bulk production: Mistral or DeepSeek. Email sequences, ad variations, social captions at scale. Efficiency over artistry.
For real-time trends: Grok. Social listening, reactive campaigns, cultural moment-jacking. Speed over safety.
For data and dashboards: Cohere. Structured reporting, multi-source summaries, KPI dashboards. Built for business intelligence, not creative flair.
For research and validation: Perplexity. Market intelligence, competitive analysis, trend validation with sources. Excellent for due diligence, rubbish for ideation.
The Mistakes You’re Already Making
Mistake #1: Using ChatGPT for everything. You’re wasting time and money using a Michelin-starred chef to make toast. Scale your tools to your tasks.
Mistake #2: Trusting visual AI for accuracy. MidJourney will create a stunning image of your product that looks nothing like your actual product. Use it for concepts and mood, not for specs.
Mistake #3: Not fact-checking persuasive content. The more confident and well-written the output, the more you should verify it. Eloquence is not evidence.
Mistake #4: Ignoring brand safety. Using Grok to write your pharmaceutical company’s press release is career suicide. Match the tool to the risk level.
Mistake #5: Forgetting that humans still matter. Every output needs a human who understands the brand, the market, and the stakes. AI is a tool, not a replacement for taste, audience empathy and leadership.
The Bottom Line
Here’s what separates competent marketers from brilliant ones in 2025: Knowing which tool does what, and having the discipline to use the right one for each job.
ChatGPT for strategy. Claude for safety. Gemini for visual storytelling. Mistral for scale. Grok for trends. Cohere for data. Perplexity for research.
Match the tool to the task, fact-check everything that matters, and never forget that the human in the loop is still the most important part of the equation.
Because here’s the thing:
Most of your competition still thinks AI is just ChatGPT.
Most of them are using sledgehammers for brain surgery and wondering why the results are messy.
Most of them haven’t bothered to learn the difference between tools, let alone master when to use each one.
But you? You’re reading this. You’re learning this. You’re building a capability that most senior marketers don’t even know exists yet.
That’s not mediocre. That’s smart. And in a world where everyone has access to the same technology, being smart about how you use it is your unfair advantage.
Kate Busby is CoFounder of Quiet Edge and a Fractional CMO based in Barcelona, Spain, catch her on X and Instagram. The images are extracted from X and created by MidJourney. No names and identifying details have been changed. Subscribe to Substack to receive all articles in the “Rules of Growth” series straight to your inbox. Save this article. Share it. Use it. And for God’s sake, stop asking ChatGPT to do your data dashboards.