A Shortcut Hidden in Your Own Writing
Most organizations are already in the publishing business, whether they think of it that way or not. There’s always another message to explain, update to share or audience to reach. That work might be a newsletter, a board report, a social media post, or blog.
It’d be great if you could give AI your notes and source material, explain what you’d like to say, and get back a polished piece that sounds like everything else you’ve written for this particular platform. While we’re not quite there yet, we’re closer than ever. The posts you’ve already finished are now the key to stronger AI-assisted drafts, for blogs, reports, social media, and almost any other writing project as well.
All Structure, No Signature
As is, AI’s first drafts are better used as scaffolding than finished copy. The information may be sound and neatly assembled, but the writing can feel generic, detached from the people and purpose behind it. Strip away the logo, and the piece could have come from almost anywhere.
What gives? Routine organizational writing should be a natural fit for AI. In theory, a robot should be the perfect tool for maintaining standards of tone, rhythm, word choice and more across a whole collection. A blog, for example, builds a library of niche content over time, providing useful information in a style and format readers recognize. The more you publish, the more important that consistency becomes.
But authentic writing still requires a good deal of human input. Each revision adds another directive to your prompt, and over successive drafts the conversation can become cluttered enough that commands blur, compete or cancel one another out. Prompting becomes an exercise of diminishing returns.
The good news is, if you’ve been writing with AI for some time now, you’re likely sitting on a gold mine of editorial clues that can speed the process by telling ChatGPT more than another page of instructions ever could.
Context is King
The trick is to show AI the difference between its work and yours. Start with a strong sample that represents your organization’s voice and spirit, ideally a piece that required a lot of back-and-forth with AI to produce. Load the finished copy into a new chat window along with the original draft, and a prompt like:
Compare these two versions: “<Article Name> AI First Draft” and “<Article Name> Final Edit.” Identify the key differences, explain the editorial choices they reveal, and suggest rules that could improve future drafts.
Repeat with another first-and-final pair, and a third one, too, if you have it. AI will analyze them all and identify distinguishing patterns, like: preferences around lists, first-person narration, casual wording, historical context, and other recurring choices that give your work its character.
Written communications reflect hundreds of editorial judgments, big and small. By comparing the first and final drafts, AI can infer which adjustments brought the article closer to completion. Then it puts together a playbook of practical guidelines for stronger first drafts, faster. These rules become a set of standing instructions it can follow every time you write similar content.
The Robot Writes the Rulebook
The easiest way to put those criteria to work is to create a Project for your blog (or other comparable task). Add the guidelines under Project settings, then upload finished examples, research, and other useful material to Sources.
Every chat inside the Project can draw on the same instructions and reference files, so each new draft starts with that context already in place. Projects isn’t necessary, however, you can also keep the rules in a master document and attach it whenever you open a new chat to draft a post.
Now the fun part: a test drive! Your first writing attempt will probably expose some gaps, so ask ChatGPT to keep a running list of any new preferences or recurring problems that emerge as you edit. Try a prompt like:
As we revise this draft, track any new writing rules or preferences we uncover. Don’t update the guidelines yet. At the end, give me a copy-ready list, then help me decide which ones belong in the permanent instructions.
This way, the test draft becomes useful evidence for improving the playbook—not just another article to finish. Even better: the next one comes together a little quicker, a little easier. Before long, you’ll have a custom shortcut that moves a good idea toward finished, original content with less friction.
Trial by Edit
While AI’s capabilities are impressive, they’re no substitute for real-world experience and judgment. This comparison method helps easily translate your particular expertise into clear, effective outreach. It can reduce the time and effort of composition, so it’s just a matter of uploading your notes and research, then summarizing your article idea in a stream of consciousness (as a document or recording).
The rules you’ve developed help ChatGPT turn that raw thinking into a readable draft. You’re still the writer, but now with better instructions the AI should be more effective at shaping content to your liking. Ideally, you’ll spend less time making the same tweaks again and again.
This article itself is a real-life demonstration. The version published here is about 29 percent shorter than the original AI draft, largely through cuts to repetition, detours and explanatory padding. By comparison, in the three earlier articles used to build the playbook, much more had to be created or substantially rebuilt during editing.
While ChatGPT organized ideas and produced a logical framework, it habitually ignored several explicit instructions. To its credit, this AI caught these errors and added an extra self-check to scan future drafts specifically for rule-breaks before returning them.
Perhaps the most improvement came during our final “red flag speed round,” when ChatGPT highlights one potential problem at a time for review. That stage usually requires several exchanges to settle each revision, but this time the suggested fixes often landed on the first try. Polishing moved much faster, with far less back-and-forth.
This remains an ongoing experiment aimed at leveling up EFM’s blog production. For now, early evidence shows how comparing first and final drafts helps ChatGPT understand the editorial choices behind the finished work—and apply them more effectively to the next one.
Editor’s Note: Throughout this article, “AI” and “ChatGPT” are used almost interchangeably, since that’s the tool we currently subscribe to. The technique isn’t platform-specific, though — comparing first drafts against final edits to build standing instructions works with any AI tool that supports saved instructions and reference files, including Claude’s Projects and Gemini’s Gems. Use whichever you prefer.
What do you think? Everyone uses AI differently, new perspectives are always appreciated. Please consider sharing your thoughts below in the comments.
✨ Keep Exploring: If this kind of AI-assisted workflow intrigues you, we’ve got more where that came from — our series kicks off with Boost Your Workflow with ChatGPT Memory, where we show how to save time, stay organized, and keep your ideas at your fingertips.
Thanks for reading! This blog is powered by East Falls Media, where we help small businesses, nonprofits, and local governments connect with clarity and purpose.
