This is part 3 of a series. Read Part One and Part Two if you’re so inclined.
I suspected after last week’s regular blog post that ChatGPT might be having trouble with the hard returns that are generated by the WordPress editor. I also considered that it might not like the section headings that I use for readability. My experiment this week will involve an older post with two headings. I’m going to try using the text straight from Word, then from the WP editor, and then from Notepad without headings.
The original headline, “I’m in London This Week,” scored a 44 at CoSchedule’s Headline Analyzer. They dinged me for word balance (too many common words; no uncommon, power, or emotional words) and neutral sentiment. Let’s see what we get from ChatGPT using the text from Word with headings.
| Traveling to London, Ohio for a Wedding: Adventures of a Diabetic, Ink-Loving Grandfather | 62 | Word balance and readability (12) |
| Life on the Road: Traveling for Five Days with Diabetes and Contemplating Tattoos | 61 | Word balance, readability (10), and neutral sentiment |
| Preparing for a 5-day trip with type 1 diabetes: Tips and personal experiences. | 71 | Word balance and neutral sentiment |
I don’t mention it in the post, but I have Type 2 diabetes, not Type 1. I’m not sure why ChatGPT assumed one over the other.
I note too that the second headline came out of the chatbot in quotation marks, but the other two didn’t. There doesn’t seem to be any rhyme or reason as to why that happens.
Now we’re going to go to the WordPress editor and copy the text from there. I’ll use the same post and the same prompt of, “write a headline for the following blog post.” The WP editor has the images in place as well as captions. The images don’t carry over but the captions do. I’ll do a fresh chat too.
| Diabetic Travel: Packing for a Five-Day Trip with Insulin Supplies and More | 59 | Word balance, readability (10), and neutral sentiment |
| Preparing for a Road Trip with Diabetes – Packing Insulin, Sensor and More | 58 | Word balance and neutral sentiment |
| Traveling with Diabetes: A Reflection on Insulin Supplies, Ohio Pit Stops, and Tattoo Ideas. | 53 | Readability (12) and neutral sentiment |
Those responses are pretty much in line with each other. Now I’ll do the text from Notepad with headings.
| Traveling with Diabetes: What I Packed for a Five-Day Trip | 56 | Word balance and neutral sentiment |
| Blog Post Headline: A Five-Day Trip to Ohio: Reflections on Traveling, Family, and Tattoos | 55 | Yes, that’s really what came out of the chatbot. Word balance, readability (12), and neutral sentiment |
| Traveling with Type 1 Diabetes: My Experience on a Five-Day Road Trip to London, Ohio | 63 | Word balance and neutral sentiment |
It’s starting to look like my theory isn’t panning out. Now I’ll try text from Notepad without headings and without line feeds. It’ll just be one long block of text.
| From Ohio to London: Adventures of Traveling with Diabetes and Planning Tattoos | 69 | Word balance and readability (11) |
| Traveling with Diabetes and Planning New Tattoos: A Journey to Ohio and Beyond | 59 | Word balance and neutral sentiment |
| Traveling with Diabetes: A Road Trip to Ohio and Tattoo Musings | 57 | Word balance and neutral sentiment |
Summary
The average Coschedule Headline Analyzer ratings for each test are 65, 57, 58, and 62. That’s just about an 11% variance, which I don’t think is that significant given the sample sizes. I think that means my ideas about how the text is fed to ChatGPT were wrong, as it doesn’t seem to matter much in the end.
Thanks for reading! Feel free to share a thought in the comments. Sign up for my infrequent newsletter here. Find some of my other writing at The Good Men Project, too. Subscribe to the blog via the link in the right sidebar so you never miss an update here, either. You can also add my RSS feed to your favorite reader.
Share your thoughts!