Learning to Work with the Robot: A Journey in Progress
For the past five months, I’ve used ChatGPT to support my copywriting work for CDP. In this post, I’ll be sharing what I’ve learned — or at least experienced — so far.
Some brief context about my role at CDP: For the past six years, I’ve been CDP’s digital marketing copywriter. And I’ve had copywriting duties at various other public media organizations for 30 years before that.
It's worth noting that my role doesn’t involve producing copy in a one-person vacuum. The digital creative team cultivates a collaborative environment — one where ideas are refined through teamwork, so that our digital fundraising messages can be as effective as possible. That also means welcoming specific feedback from MSB Co-op partner stations and closely watching the digital messaging of public media organizations and other nonprofits.
Here are some of my observations about using ChatGPT in my copywriting work so far (in this case, the entry-level paid version of what is now ChatGPT5):
The Big-Picture View
ChatGPT can be a helpful (if inconsistent) tool for a diligent copywriter, but it’s no substitute for human creativity and skill. Why? Because it has no — to coin a phrase — “banality filter.”
In other words, ChatGPT5 often meets the structural requirements of a fundraising appeal. It can mimic tone, insert a call to action or reference a campaign. But without a human filter, it leans toward safe, generic and emotionally flat language. It tends to fill space with what “sounds like” fundraising, often resulting in the following:
- Tired phrasing (“Your support means the world!”)
- Predictable language (“ignite change,” “make a difference”)
- Surface-level emotion without authentic context
This happens because the large language model’s training data — the sheer tonnage of what AI hoovers up from the internet — isn’t built to evaluate quality, novelty or creative tension. It’s not allergic to banality — but that’s exactly what it (re)produces unless steered otherwise.
And you don’t have to take my word for it. Every sentence in the box above was generated by ChatGPT without a single edit from me. I even asked ChatGPT to confirm the accuracy of the statements. It did.
The Prompt’s the Thing!
How did I get the AI app to critique itself? I prompted it with this question: “Is it accurate to say ChatGPT5’s copy output has no ‘banality filter’ or aversion to banality? It’s possible for it to produce acceptable fundraising copy from a functional perspective, but can it sometimes be as insipid as a sink full of used dishwater going cold?”
Now, I've also learned that ChatGPT is designed to generate responses that “tell the user what they want to hear.” Knowing this, I re-entered the original prompt and asked the app to “play devil’s advocate” and say whether what I’d written was “truly accurate.” Here's what ChatGPT shared with me:
“No banality filter” is not quite accurate. It does “know” that “Your support means the world!” is common phrasing. It just doesn’t care, unless you instruct it to care.
Banality is often a feature, not a bug. For many users, especially those without a writing background, “safe and generic” means “usable.”
The AI response was also that, “ChatGPT5 can generate serviceable fundraising copy, but without specific guidance, it defaults to safe, often generic, language. It’s not immune to banality unless you explicitly ask it not to be.”
This exchange shows why it matters to thoughtfully craft the prompts you give ChatGPT. In fact, I’ve found that it’s key to prompt, re-prompt and then prompt again! The AI app’s output can be vastly improved by doggedly interrogating it.
Don’t expect ChatGPT to deliver miracles — not on the first pass, or even the sixth.
Do work patiently and incrementally — refining your prompts along the way.
Don’t expect AI to be a magic machine that writes copy for you.
Do treat it as a tool for revision that can hold up a mirror to your writing.
The Robot Curtails Some Writerly Bad Habits
Most writers have their idiosyncrasies. When those tendencies produce strong, persuasive copy, we might call it “voice,” “style,” or “originality.” But when they interfere with crafting digital fundraising copy that drives conversion, we can consider them “writerly bad habits.”
In my case, I can sometimes find myself using lofty vocabulary phrases (i.e., “big words”) in my first drafts, when simpler ones would do. I can also be prone to building complex syntax. For MSB partners’ ESOL copy, we aim to keep the Flesch-Kincaid Grade Level below 8 — ideally below 7. Likewise, we aim for a Flesch Reading Ease score of 60 or higher. Meeting those metrics isn’t always easy, and ChatGPT has been a useful tool for helping me simplify language while still making the same point.
A Supercharged Google
Beyond helping me generate and refine copy, ChatGPT has also been useful for summarizing facts, complete with citations from reliable online sources.
For example, when legislation to rescind federal funding for public media was unfortunately moving through both chambers of Congress, the digital copy for MSB stations had to stay accurate and up to date. I found ChatGPT immensely helpful on this front. Not surprisingly, much of the most reliable and detailed content the AI served up came from public media organizations!
A caveat: If you’re using ChatGPT for research, ask for links to the online sources it’s using to arrive at answers. The interface is capable of being dramatically incorrect on first passes when providing summaries of real-world situations without backing them up.
Summing Up
I do want to note that there is philosophical and/or political opposition to using AI in creative endeavors. For this post, my focus has been the experiences I’ve had experimenting with using ChatGPT, rather than delving into the arguments for/against its overall use.
And I’m not at all inclined to use ChatGPT as a tool, for example, to work on my own off-the-clock poetry or fiction writing projects. But digital fundraising copywriting is more functional craft than artistic endeavor. More about persuading people to give than it is about creating an aesthetic experience.
I’ve also found that ChatGPT is not at all likely to help produce good copy without an attentive and persistent human at the helm. Can it adjust or amp up dimensions of copy like tone, urgency or emotional resonance? Yes. But not without a conscientious and thoughtful human being defining — and then likely further refining — what exactly those dimensions are.
At CDP, we write all our partner stations’ copy in service of our thoughtfully crafted strategies. AI may be able to churn out passable copy — keeping all the aforementioned limitations in mind — but it requires repeated human interaction to ensure alignment with strategy.
After five months of working with ChatGPT — a journey still in early days — I’ve come to see it less as a writer, and not even a co-writer. It’s more of a utility player — part critic, part mirror, part editor, part research assistant. It doesn’t create anything original from whole cloth, but it can help shape my own writing. It won’t necessarily elevate my copy, but it might show me where there's room for improvement. It won’t find the ideal phrase, but — through an iterative process — it might make me more aware of why the one I’ve used doesn’t quite land.
My takeaway? What ChatGPT gives you depends entirely on how you use it. If you interrogate it, challenge it and keep refining your prompts, it becomes a powerful tool — not because it’s creative, but because it can help you become more precise, rigorous, and clear with your own creativity.
Further reading about uses and limitations of AI in fundraising and marketing campaigns:
• “Getting AI Right: A Fundraiser’s Guide to Responsible Use” – Chronicle of Philanthropy (Sep. 16, 2025).
• “The Limitations of AI in Brand Voice: Why ChatGPT Can’t Replace Human Touch” – Riggs Creative Group (Apr. 25, 2024).
• “ChatGPT Limitations for Marketing in 2025” – Pipeline Velocity (Aug. 19, 2025).
• “AI Copywriting Risks: Good Reasons Not to Rely Solely on AI” – Jenny Lucas Copywriting (Jul. 21, 2023).
• “The good, the bad, and the ugly: generative AI in marketing in 2025” – Sixgun (Jan. 8, 2025).