Since OpenAI released ChatGPT, the adoption of AI has rapidly expanded, from writing blogs, to generating images of the Pope in a stylish puffer to generating constituent response drafts for congresspeople. As the technology gains traction, the opportunities it presents for teams large and small become increasingly apparent.
In this series, I'll outline how teams running advocacy campaigns can leverage generative AI, specifically ChatGPT, to scale content creation, outbound strategies, and document drafting. I'll go over where ChatGPT can provide the most value and how to work with its output to produce the best results. Throughout the article, I'll also provide ChatGPT prompts for advocacy uses and explain what makes them effective.
In this first blog post of this series, I'll show you how to unlock the potential of ChatGPT for content creation with a guide to prompt wiring.
Creating an effective prompt in ChatGPT is easier than it may sound. Even a prompt as simple as “Write me a blog about x” can produce an impressive response. But Large Language Models (LLM) like ChatGPT are using content from the entire internet to generate the response, so the more specific a prompt is, the better a prompt will be.
The golden rule of prompting: the more detail you add to a prompt, the better your results will be.
This is the result I got when I prompted ChatGPT, “Write me a blog about Carbon Capture.” The response is good, written in a formal tone, but ultimately it’s bland.
It’s impressive AI can produce such a response with so little to work off, but that impression shouldn’t be confused with thinking the output is good by the standard at which a writer can write. Here it's important to note that writing with AI is an iterative process. When an initial result is not satisfactory prompt again.
But with the output there are two ways to go:
a. Copy what you have into a new document and use it as a baseline
or
b. Refine your prompt to get a response closer to what you were expecting.
Refining a prompt is the process of adding context to a prompt. I'll provide some advice, but there's no set rules to this, a lot of prompt writing comes down to experimenting and finding what works.
Here are a couple examples of context you can provide to make your post better:
So now instead of “Write me a blog about Carbon Capture.” our prompt looks like:
I want you to act like a marketing professional with 10+ years of experience working on advocacy campaigns. You are tasked with writing a blog post about carbon capture. The goal of writing and publishing this post is to raise awareness about environmental innovation and attract people to sign a petition on our website. The target audience of the blog post is college students in the United States. Use a tone and style that will appeal to this audience. Eliminate any fluff in your response, avoid cliches like “game-changing.” Write the blog post in well formatted markdown.
The result is already much better:
The real power of ChatGPT is that it remembers the previous messages you sent in the conversation. With this in mind, writing pillar content should be an iterative process in the chat.
A great technique for creating pillar content like a blog post in ChatGPT is to make the first message a prompt with dense context asking for an outline.
With the outline in hand you can edit it yourself to better match your objectives. Then in the next message to ChatGPT, give it the revised outline and ask it to write the first section and so on, giving feedback between each section. As you’ve already given context in the first message, there’s no need to give context again.
The same prompt-writing techniques shown here can be applied to social media posts, emails, petitions, position papers, even slogans to put on a sign. With this tool, even small teams to produce resources at a rate not otherwise possible.
Now you know how to generate great one off items, but the power of ChatGPT doesn’t end there. In the next section we’ll talk about how ChatGPT can help plan, strategize, and schedule advocacy efforts.