What I use, when I use it, and why the "which AI is best?" debate is the wrong conversation.
I get asked some version of this question almost every week: "Isaac, which AI should I be using?"
And every time, my answer is the same: that's the wrong question.
The right question is: what are you trying to accomplish, and which tool is purpose-built for that specific job?
A year ago, "using AI" meant one thing — typing into a chatbot and going back and forth until you got something usable. That was the entire workflow. And in that world, "which chatbot is best?" was a reasonable question.
That world is over.
Today, the AI platforms have evolved into full work environments. They can execute code, build spreadsheets, create slide decks, conduct research with real citations, work inside your Excel files, operate on your desktop, and — increasingly — carry out multi-step projects with minimal supervision. The chatbot is still there, but it's now the smallest piece of a much larger system.
So instead of ranking the platforms, I want to walk you through how I actually think about AI as a working marketer — organized by the work itself.
The One Setting That Changes Everything
This applies to every platform, and it's the single highest-leverage thing you can do.
Every AI platform defaults to a fast, lightweight model. It's optimized for casual conversation — quick, friendly, and often wrong on the details. If you're doing real marketing work with the default settings, you're leaving most of the capability on the table.
Here's what to do:
- ChatGPT: Manually select GPT-5.2 Thinking Extended. The default "auto" mode will route you to a weaker model more often than you'd expect.
- Claude: Select Opus 4.6 and turn on extended thinking.
- Gemini: Select Gemini 3 Pro or Thinking mode.
All three require a $20/month subscription to access the frontier models. It's the best $20 you'll spend this month. The gap between the free tier and the paid tier is enormous — we're talking about a fundamentally different level of reasoning, accuracy, and output quality. Everything I'm about to recommend assumes you've made this switch.
Writing Copy and Creative Strategy
This is where I spend a significant portion of my AI time — drafting ad copy, email sequences, landing page messaging, content strategies, creative briefs, captions. Work where voice, tone, and precision matter.
My go-to: Claude Opus 4.6.
Here's why. Claude consistently produces copy that sounds like it was written by someone who understands the brief. When I give it a brand voice guide, it internalizes it faster than the other platforms. When I ask it to write for a specific audience, the output reflects genuine understanding of what that audience cares about, rather than generic marketing language stitched together.
The practical result: I spend less time editing Claude's output than I do with the other two. For a marketer, that's the metric that matters. The best AI writing tool is the one that gets you closest to "done" on the first pass.
ChatGPT is solid here too — especially for high-volume iteration where you want a lot of variants fast. Gemini is capable but tends to default to a more neutral, informational tone that requires more steering for persuasive copy.
Data, Analysis, and Reporting
Campaign performance analysis. Budget modeling. Media mix calculations. Customer segmentation. Anything where you need the AI to think carefully about numbers and draw accurate conclusions.
My go-to: ChatGPT with GPT-5.2 Thinking Extended.
GPT-5.2 in Thinking mode has the strongest quantitative reasoning of the three. When I upload a dataset and ask it to find patterns, test hypotheses, or build a model, it approaches the problem with a rigor that feels more like working with an analyst than a chatbot. It writes and executes code to validate its own conclusions, which means you're getting tested outputs rather than confident guesses.
One thing I particularly value: both ChatGPT and Claude can generate actual, downloadable spreadsheets and formatted reports from your data. This means you go from raw data to a polished deliverable inside a single session. Gemini's model is equally capable of the analysis itself, but its interface can't produce those downloadable files yet — so you'd need to do the formatting work yourself.
For marketers who live in spreadsheets specifically, Claude for Excel deserves its own mention. It's a plugin that operates directly inside Excel as what I'd describe as an embedded junior analyst. You tell it what you want — "build a pivot table by channel and quarter," "calculate blended ROAS excluding branded search," "clean up the formatting on this media plan" — and it does it, right there in your spreadsheet. The results are in Excel, so they're immediately verifiable and shareable. If you spend meaningful hours in Excel every week, this will change how you work.
Research and Strategic Thinking
Competitive analysis. Market research. Synthesizing customer interviews. Building strategic frameworks from multiple sources. The kind of thinking work that requires pulling insights from many places and connecting them.
My go-to: A combination, depending on the task.
For live research (what's happening right now, what competitors are doing, what's trending in a market), all three chatbots now have web search built in. Claude and ChatGPT both provide citations you can verify, which matters when you're building strategy on top of the findings.
For deep synthesis — where you have a stack of documents, reports, transcripts, or videos and you need to extract the data — Google's NotebookLM is the tool I'd recommend every marketer set up this week. It's free. You upload your materials, and it builds an interactive knowledge base you can query, turn into slide presentations, generate mind maps from, or even produce AI podcast episodes where two hosts discuss your material. For anyone who regularly needs to make sense of a pile of information (which is every strategist, every planner, every account lead), NotebookLM is genuinely useful.
Each platform also offers a Deep Research mode — you give it a complex question, it goes away, conducts extensive multi-step research, and comes back with a comprehensive report. These are worth experimenting with for any research-heavy project.
Building Things: Landing Pages, Tools, Dashboards
This is the category that has changed the most dramatically, and the one I think most marketers are sleeping on.
A year ago, if you wanted a landing page, a calculator, a client-facing dashboard, or an internal tool, you needed a developer. Or you needed to learn to code. Or you cobbled something together in a no-code tool and accepted the limitations.
Today, the coding agents — Claude Code, OpenAI Codex, and Google Antigravity — let you describe what you want in plain language, and the AI builds it. A working, functional, tested thing. Landing pages. ROI calculators. Reporting dashboards. Interactive tools. You describe the outcome. The AI writes the code, runs it, tests it, fixes the errors, and delivers a finished product.
You don't need to understand the code. You need to understand the outcome you want — clearly enough to brief it the way you'd brief a developer.
For marketers who aren't technical, Claude Cowork is the entry point I'd recommend. It's a desktop application that can work with your local files and browser, and it handles multi-step tasks: "take these 12 campaign reports and consolidate the key metrics into a single spreadsheet," "organize this folder of creative assets by platform and campaign," "draft a competitive brief from these three websites." You describe the outcome, it plans the steps, and it executes.
It's still early — it's a research preview, and it uses up your usage allocation quickly. But the capability it represents is where all of this is heading: AI that carries out the work rather than advising you on how to carry it out yourself.
Visual Content: Images and Video
My go-to: Gemini.
This is the one area where Gemini has a clear lead, because it bundles the best creative generation tools directly into the platform. Nano banana (the image generation tool) is, as of right now, the most capable AI image creator available. Veo 3.1 handles video. Both are accessible directly from the Gemini chatbot.
ChatGPT's image generation is decent — and if you're already in the ChatGPT ecosystem, it's more than capable for most marketing needs.
Claude currently lacks image and video generation entirely, which is worth knowing if visual content is a core part of your workflow.
How I Think About This
The mental model I'd encourage every marketer to adopt: you're assembling a team, not picking a favorite.
Each of these tools has a role. Claude is my writer. ChatGPT is my analyst. NotebookLM is my researcher. Claude for Excel is my spreadsheet operator. Gemini is my creative studio. Claude Code is my developer.
No single platform is the best at everything. The marketer who tries to force one tool to do every job will consistently get outperformed by the marketer who reaches for the right tool for the task at hand.
And here's the shift that underpins all of this: the skill that matters most is no longer prompt writing. It's briefing. It's the same skill that makes someone good at managing people — clarity of intent, quality of context, and the judgment to know whether the output is right.
The marketers who learn to delegate to AI the way they'd delegate to a talented junior hire will find that the leverage is extraordinary. And it compounds. Every week you spend learning to work this way, the gap between you and the marketers who are still just chatting widens.
That's the opportunity in front of us. It's worth taking seriously.
If this was helpful, forward it to your team. The sooner everyone's on the same page about which tools to use, and more importantly, how to use them, the sooner the whole team starts compounding that advantage together.
Until next week,
- Isaac Rudansky
Founder of the Modern Marketing Institute
Responses