A Marketer’s Tale of Four AI Tools: What I Actually Think After Using Them All

There’s a version of this post that ranks tools from best to worst. This isn’t that. This is a working marketer’s field guide to four tools I’ve actually used—what each one does well, where each one will let you down, and why the question isn’t which tool is best, but which tool belongs where.

Every marketer experimenting with AI right now is quietly sitting with the same question: which of these tools is actually worth the subscription? And if I’m already paying for two or three of them, am I using them right or am I just paying for the same capability twice?

I’ve spent real time inside ChatGPT, Claude, Gemini, and NotebookLM over the past year. Not as a reviewer. As a working marketer building content systems, running client campaigns, and trying to make these tools actually useful—not just impressive in demos. (If you’ve been thinking about how AI is changing search visibility too, we covered that here.) Here’s what I found.

The Quick View: How Each Tool Fits

Before diving in, here’s how I think about each tool’s role at a glance:

Tool	Role	Best For	Watch Out For
ChatGPT	Thinking Partner	Brainstorming, early-stage ideation, strategy exploration	Confident hallucinations, validates instead of challenges
Claude	System Builder	Structured content workflows, team consistency, complex briefs	Can feel rigid in freeform creative work; usage limits
Gemini	Researcher	Data gathering, Google Docs/Sheets, pre-workflow inputs	Tone can be blunt; needs softening for client-facing work
NotebookLM	Synthesizer	Multi-source analysis, complex topics, client presentations	Limited flexibility; not built for full content workflows

ChatGPT: The Thinking Partner

ChatGPT is still the tool I reach for when I’m stuck and need to think out loud, though I’m getting more used to Claude.

It’s genuinely strong in early-stage work: brainstorming directions, pressure-testing a strategy angle, breaking down a problem that feels too big to start. The conversational quality is natural, and because I’ve used it for so long, it has a sense of my thinking style that makes those early conversations efficient.

The caveats are real, though. ChatGPT has a confidence problem. It will deliver an answer that sounds authoritative and be completely wrong. It tends to validate ideas more than challenge them, which feels good in the moment but isn’t always what you need. And left unsupervised, it develops habits: overusing em dashes, defaulting to “X vs. Y” frameworks, producing copy that sounds polished but still needs significant editing to feel like your actual voice.

It’s a strong thinking partner. It’s a risky execution tool if you’re not watching closely.

Claude: The System Builder

My workflow shift this year was about recognizing I needed something that could work inside a system, not just a conversation.

Claude performs differently when it has structure to work with. Organized folders, markdown files, shared knowledge sources, clear constraints (these inputs make Claude significantly more efficient than a clean prompt ever would). Once I built that infrastructure, my team could use the same workflows with minimal prompting, and the outputs stayed consistent in a way that’s actually useful at the agency level.

What Claude does better: it follows instructions and constraints reliably, handles complex structured tasks without drifting, and works well with teams because the system does the heavy lifting instead of the individual prompt. It’s less “let’s explore” and more “let’s execute.”

What I’m still working out: where to let it be flexible versus where to hold it to a tighter brief. The same structure that makes it reliable in a content system can make it feel rigid in freeform creative work. That’s a trade-off I’m navigating, not a dealbreaker.

Gemini: The Researcher

Gemini lives in a different part of my workflow. My use is less about creation, more about gathering.

Gemini’s Google ecosystem integration is genuinely useful. It works across Docs and Sheets without friction, pulls from multiple browser tabs simultaneously, and handles research and data-gathering tasks in a way that makes it a natural first step before moving into higher-cost processing elsewhere. I often use Gemini to pull together inputs that then go into Claude for structured output.

One honest critique: Gemini has a tone that sometimes needs softening before anything goes anywhere near a client. It can read as overly critical or blunt in a way that’s useful internally but requires adjustment for external-facing work. It’s a strong utility player (not the star of the workflow, but consistently in the rotation).

NotebookLM: The Synthesizer

NotebookLM is the most underrated tool in this group, and the one most marketers haven’t explored yet.

What makes NotebookLM different: it works with your data, not just general knowledge. You bring in multiple documents (research, reports, client briefs, data exports, competitor urls) and it synthesizes across all of them. The result is a coherent, grounded output that reflects the actual sources you’ve given it, not a general-knowledge approximation.

I use it most when I need to make complex information digestible. It translates a dense research stack into something I can actually present to a client without it feeling data-heavy or hard to follow. It’s also useful for developing a quick, deep understanding of a new topic before I need to write or present on it.

The limitation is flexibility. NotebookLM is purpose-built for synthesis, not execution. It’s not where I run a full content workflow. But as a complement to Claude (NotebookLM for understanding, Claude for output) it fills a gap I didn’t know I had.

The Moment That Made Validation Non-Negotiable

One experience shaped how I approach all of these tools now.

An AI-generated report showed conversions up 9,000% month-over-month. It looked clean. It looked credible. What it was actually doing was comparing a period when a campaign was running against a prior period when it wasn’t, technically accurate but completely misleading.

The number was real. The interpretation was wrong. And it would have gone to a client if I hadn’t caught it.

That’s the risk across all of these tools: confident outputs that pass a quick read but don’t survive scrutiny. It matters most in high-stakes areas like paid media reporting and SEO performance analysis where a wrong number in front of a client can do real damage. Validation isn’t optional. It’s the part of the workflow that doesn’t get to be automated.

The Actual Takeaway

The marketers getting the most from AI are the ones who’ve gotten specific about the role each tool plays. ChatGPT can be useful for thinking through problems. Claude is strong for executing within a system. Gemini works well for research and data gathering. NotebookLM is great for synthesis.

None of them replaces strategy or judgment. What they do is compress the time between an idea and something worth editing… if you know what you’re actually trying to build.

So if I had to pick one? Claude is my clear winner right now because of how well it works inside structured workflows.

I’m moving away from ChatGPT due to ongoing issues with accuracy, overconfidence, and overall company practices. And not every tool earns a place in the stack, some are intentionally left out (I won’t be trying Grok, ew).

Moving forward, my stack looks like Claude + Gemini (as a data assistant), with additional tools layered in where they make sense.

If you’re figuring out how AI tools fit into a strategy that actually drives results, or want help building AI workflows your team can actually use, we’d be glad to talk through it.