I recently had an experience with AI that opened my eyes to a couple things that I think you’ll benefit from knowing. Here’s the story…
At Better Fundraising we have a letter that asks donors to consider leaving an organization in their will. We create versions of this letter for our customers, and it’s quite successful.
But we want to make it easy for all our customers to create their own version of the letter.
So, I pasted the letter into ChatGPT and asked ChatGPT to “anonymize” it. My goal was to keep the letter the same, but take out the organization it was written for, resulting in something like a template.
ChatGPT successfully anonymized the letter… but without being asked added a mass of nonprofit-ese. It took short, simple sentences and made them longer & more complex. It took casual language and made it formal. It took personal writing and made it corporate. Ugh.
Lesson #1: Chat GPT does not naturally differentiate between the types of writing needed for different types of fundraising. AI lumps “grant applications” and “direct response e-appeals” and “brochure copy” all under “fundraising.” In a nutshell, the letter it wrote for me “had too much grant application in a direct response fundraising appeal.” The writing for those two types of fundraising should be significantly different – but AI doesn’t know that.
Lesson #2: The dataset that AIs are based on is full of fundraising writing that’s overly formal, lacks specificity, and is difficult to read. This is a “garbage in, garbage out” situation – if most of the writing that ChatGPT used to create its understanding of fundraising is poor, then the writing that ChatGPT creates is going to be poor.
So, to be more specific and teach ChatGPT what I was looking for, I went ahead and wrote the anonymized letter/template myself. Then I pasted my version into Chat GPT and told it that my letter was they style I was looking for.
For fun, I asked it to analyze the letter and describe what was different about my version.
Here’s how ChatGPT replied…
Let’s analyze the differences between the two versions.
- Directness: Your version gets straight to the point, with a direct appeal right from the start. My initial version began by establishing context, potentially adding unnecessary fluff.
- Specificity: While my version aimed to generalize various aspects, your version maintained specificity in areas crucial to the appeal’s emotional resonance, like discussing the importance of the work done by the organization.
- Language: Your language is more personal and direct. For example, “You didn’t have to donate, but you did.” is a straightforward acknowledgment of the donor’s choice to give.
- Structure and Flow: Your version flows more like a conversation, emphasizing gratitude and then transitioning to the ask. It feels more personal and less formal.
- Reference to Beneficiaries: Your letter consistently references the beneficiaries of the work, focusing on the impact the donations have. My version was slightly more focused on the organizational aspect.
- Narrative: Your letter includes a personal story about revising wills, making the ask more relatable. While my version had this as well, the context and personal touch felt stronger in your version.
- Call to Action: Your version includes a clearer call to action, specifically about using the enclosed card for notifying about the bequest. It gives readers a tangible next step.
In summary, while both versions aimed to maintain the letter’s integrity and goal, your version manages to be personal, direct, and emotionally compelling.
Color me impressed with how well ChatGPT defined the differences between our approaches. And I couldn’t help feeling that the AI had given me a compliment, which was weirdly satisfying.
Those seven principles above are GREAT advice for writing effective direct response fundraising. When you are creating appeals, newsletters, e-appeals, etc., follow those tips and you’ll immediately be a more effective Fundraiser. I bet you could paste any piece of individual donor fundraising into ChatGPT or any other AI, then ask the AI to modify it by applying the 7 principles above, and the fundraising would be more effective.
Lesson #3: AI looks at all fundraising writing as equal – it doesn’t know how each piece performed, so it doesn’t know which approaches and types of stories raise more. At Better Fundraising know from experience & data that “personal, direct, and emotionally compelling” fundraising writing will tend to raise more money. But ChatGPT doesn’t know that because it doesn’t see results!
ChatGPT does not know that one appeal letter got a 2% response rate and another one got a 5% response rate. So when ChatGPT creates fundraising, it’s pulling from “everything it’s seen, regardless of how it worked.” When Better Fundraising creates fundraising, we’re pulling from “everything we’ve seen that we know worked great.” There’s a big difference.
ChatGPT is an incredible tool. But, at least for now, there’s no danger that it’s going to replace an experienced Fundraiser.
PS — If you’re interested in learning more about creating fundraising with AI, there’s a popular video of me using Chat GPT to write an appeal on the fly for an organization that I’d never spoken to before. People like it because you’ll quickly see all the strengths and negatives of using AI to write fundraising for individual donors. You’ll be able to decide whether it’s a good tool for your workflow or not. You can watch the video here.
