Block by Block: Making ChatGPT Work for You

ChatGPT is fast—remarkably fast. It works so quickly, however, that it is easy to lose track of its changes. To keep up, I break down my prompts into steps and then refine the results.

In this example, I’m working with references that need Chicago Manual of Style (CMOS) formatting. The references are missing the authors’ first names, are not formatted correctly (or in alphabetical order), and contain spelling errors. I could ask ChatGPT to format them according to CMOS style in one generic prompt, but I’d rather do it step by step so I can keep track of the changes made.

Setting the Stage

I start by giving ChatGPT instructions for what I want it to do. First, I give ChatGPT a role (in this case, ChatGPT is now my reference editor), specific instructions, and examples of how I want to see the output.

Here’s my original prompt to ChatGPT. I’ve bolded key points and highlighted some errors so you can see them quickly (I didn’t highlight the errors for ChatGPT; I wanted it to figure that out on its own):

My First Prompt to ChatGPT

Note that I use Markdown syntax for italics so ChatGPT returns its results to me with italicized titles. (See Markdown Cheat Sheet for a quick reference to Markdown syntax.)

Building It Block by Block

And now ChatGPT does its work. Let’s recall my first step:

Step 1: First, please use the internet to find the first names of all authors and replace the initials with the first names. If you cannot verify the authors’ first names, please note that.

Here is ChatGPT’s answer:

ChatGPT’s First Step: Finding Authors’ First Names

Note that ChatGPT searched six sites to find the names. I double-checked the accuracy, and ChatGPT is correct with all first names—no hallucinations!

ChatGPT then asks if I want it to proceed to the next step, which is to format the sources according to CMOS. I confirm that I would like it to move to the next step:

ChatGPT’s Second Step: Formatting References to CMOS Style

Refining the Output

ChatGPT put in the first names, formatted the references, and even corrected spelling errors. Note that I asked ChatGPT to follow the newest edition of CMOS (the 18th edition) and provided examples of reference entries following the 18th edition. The newest edition of CMOS does NOT require the location of the publisher in reference entries; ChatGPT didn’t catch that. So I let it know:

ChatGPT’s Third Step: Refining the Output

The final step was to put the list in alphabetical order, which it already did. I’m happy with my results, so I’ll copy what ChatGPT and insert into my document.

Using Iteration to Get (Almost) Perfection

Using an iterative process with ChatGPT allows you to see incremental results and refine them as you go. By breaking tasks into smaller steps, you maintain control over the output, making ChatGPT a valuable—although not always perfect—assistant.

If you enjoyed this post, consider signing up for my blog (see the Editing with AI subscriber bar at the bottom of the page). You’ll be notified when the next post is up and of tips and classes I think are useful. I promise to never misuse your information.