When ChatGPT was released in November 2022, I — like many people whose livelihood depends on “herding words” as acclaimed Texas author Larry McMurtry says of writing — fell into a pit of existential angst. I feared my clients would fire all the writers and replace us with a ChatGPT subscription.
While that may entirely be within the realm of possibility in some dystopian future, I decided I’d tackle my fear of generative AI writing like I’ve tackled any other fear I’ve faced in life: by learning all I can about it.
In the year and a half I’ve been experimenting with it, I’ve come to rely on it as a thought partner and tool that can help me drastically speed up my writing process.
Here, I’ll walk you through my writing process and how I bring AI into each step. But first, some caveats:
Caveat 1. The process I use below is for commercial writing, not creative writing. While the two aren’t mutually exclusive, I’m clarifying this because, uhm … writing a great literary piece? I don’t have a process for that. Because I’ve never done it. I’m not sure if or how I’d use AI.
But to write blog posts and other content that is intended to help a buyer make a buying decision — what my old journalism school professors called “service journalism” — I do follow a repeatable, reliable set of steps.
Caveat 2. Most of what you’ll read about below involves my use of ChatGPT4 and ChatGPT4o. There are far more AI tools out there. I reference a few of them, but this blog mostly covers how I use ChatGPT.
With that out of the way, let’s dive into the writing steps, and where I’ve been relying on AI. I’ll also show you where and how I used AI to draft the blog post you’re reading now.
One fun aside. The actual writing part doesn’t even happen until Step 6.
Step 1: Understanding the Buyer Persona I’m Writing to
The first question I ask a client who’s given me a writing assignment is: Who are the intended readers of this post? What buyer persona am I writing to?
This is critical because I’d write a piece intended for a decision-maker at a company differently than I’d write one intended for an influencer at the same company. An influencer is going to want content that could help them sway their boss to make a decision. On the other hand, the decision-maker is going to want content that explains the value of whatever I’m selling.
Most of the time, my clients share their buyer personas with me. Almost always, I need more detail than what they provide. This is where I bring in AI.
I’ll share the persona in ChatGPT 4o and I’ll ask questions like:
- What keeps this person up at night?
- What are their jobs to be done?
- What would make them look like a hero in their boss’s or shareholders’ eyes?
- How could they use my client’s product? How would they use different features of my client’s product?
This usually gives me a better understanding of who I’m writing for.
Here’s an example using one of m3’s personas: Marketing Mary.
Step 2: Understanding How My Product Solves the Buyer’s Pain Points
When I’m writing about a product and how it can solve a buyer’s problems, one of my goals is to do so in a way that my competitor couldn’t. In other words, I really lean into the unique value proposition. To that end, I need to understand specifically how my client differentiates their products from the competition.
If my client’s products do pretty much the same thing as their competitors, then I need to understand how my client positions their products in a unique way.
AI is less helpful at this step because product differentiation, positioning and unique value proposition details should come from the client’s product development strategy. That said, I can ask AI to evaluate the websites of my client and its competitors and offer a summary of how their messaging differs.
I can also ask AI to map out the buyer persona’s pain points and offer suggestions as to where my client’s product can solve those pain points.
While I put more trust in the client’s messaging and positioning than an AI-generated messaging and positioning, AI might surface some aspects of the buyer pain points or jobs to be done that I or my client may have overlooked.
A side note: If my client’s product positioning isn’t well defined (please tell me your positioning isn’t “we offer quality services, on time and on budget!”), then the client has a problem that a blog post — written by a human or otherwise — can’t solve.
Step 3: Identify Subject Matter Experts (SMEs) Who’ll Share Their Expertise for the Blog Post
In step 2, I noted that I want to write a blog post that a competitor can’t write.
Ways to do this include getting quotes and anecdotes from subject matter experts on my client’s team, or case studies from their customers. Infusing the blog post with the client’s insights, or those of their customers, is a surefire way to write something competitors can’t.
This means interviewing SMEs, which requires a lot of legwork, including:
- Researching the SME so I can ask questions relevant to their background. LinkedIn is a great resource for this, as are any customer-supplied bios.
- Thinking of questions to ask the SME.
- Doing the interview and recording it.
- Transcribing the interview.
- Reading the transcript and identifying major themes.
AI tools offer massive time savings with each of these tasks.
- I write as many questions as I can think of for the SME, and then I upload them to ChatGPT and ask it to suggest additional questions.
- When it’s time for the interview, I ask the SME’s permission to record. With their permission, I record the call. Zoom, Teams and WebEx have call-recording and transcription functions. Know the recording laws for your state.
- Need to do the call via phone? I like Rev.com.
- These apps offer transcription services as well; most are AI-generated. If you need something to be transcribed by a human, Rev.com offers this service.
- Most if not all of these tools will generate a summary of what was discussed and list any next steps that the participants agreed on during the call.
I won’t go into too much detail here, but when I was in journalism school and throughout my early career, I tape recorded interviews. I’d then painstakingly transcribed the recording.
Stop. Rewind. Play. Repeat … and type what the person said as fast as I could. If the interview was an hour, it’d take me double that to transcribe it. AI offers exponential time savings compared to OG methods!
Step 4: Conduct Additional Research
First-hand research — interviewing SMEs — is only part of the information-gathering process. I’m often asked to cite third-party research, like PubMed articles or other scholarly works.
To save time, I’ll ask ChatGPT to summarize the contents of these reports. I can even ask ChatGPT to explain the relationship between the article and the topic of my blog post to help create connections.
Step 5: Write an Outline
By this point, I’ve usually figured out what I want to say and how I want to organize the blog post, so it’s time to create the outline.
While I could have ChatGPT write the outline from scratch, I prefer to do my own thinking at this stage.
That said, I will share my outlines with ChatGPT and ask it to evaluate it and offer suggestions for improvement.
Here’s a comparison of my original outline, and ChatGPT’s response to my prompt:
If you’re reading carefully, you’ll see I haven’t followed ChatGPT’s recommendations to a T, but I am following many of its suggestions.
Step 6: Write the Blog Post
Again, here I prefer to use my neurons and write what I think should be written. I want it to include my voice (or my client’s), and use phrases I come up with.
Where does AI come in? Here are a few prompts I find helpful throughout the writing process:
- Refine this thought. I use this when I know what I want to say, but I’m taking too long to say it. ChatGPT will offer different approaches to my original wording. It will also offer explanations as to why it offered the suggestions it did, and how they may have improved my original writing.
- Transition from one thought to the next. Transitions can be tricky. Sometimes I’ll have ChatGPT help me think of ways to get from point A to point B.
- Call back to my lead. I spend a lot of time thinking about creative ways to lead a story. By the time I’m done with a blog post, my brain is fried. I’ll ask ChatGPT to offer conclusions to the blog post, recalling what I used in my lead.
ChatGPT doesn’t always produce results I like. It often spits out purple prose; ChatGPT seems to love superlatives and cheesy adjectives. Again, I use my own brain cells to decide if my original is better or if ChatGPT’s response is.
Sometimes there’s not a clear winner, but Chat GPT will use a word or phrase that I like, and I’ll replace a word here or there that ChatGPT offered.
Step 7: Editing and Proofreading
While I’d like to say I write flawless first drafts, I’d be lying if I did.
Why Everyone Needs an Editor (Seriously, Everyone)
I’ve found ChatGPT to be a great first editor for:
- Grammar and spell checking
- Awkward phrasing
- Simplifying complex sentences and enhancing readability
- Consistency and tone
- Content structure and readability
ChatGPT will respond with edits and explain why it is making each suggestion.
Again, this is where my experience as a writer and editor come in. I trust my instincts, skills and education more than ChatGPT, so I’ll decide if I want to apply ChatGPT’s edits.
There are so many AI tools out there for writing and editing support, including Grammarly and Hemingway. I’ve only used ChatGPT thus far and look forward to experimenting with others.
Step 8: Write H1s and H2s, AKA Headlines and Subheads
Headlines and subheads are crucial elements in blogs and other content because they grab attention, provide structure and set expectations for the reader.
In print, these elements are called headlines and subheads. For blog posts and web pages, they’re often called H1 (heading 1) and H2 (heading 2) and so forth. They’re important for:
- Readability. Make content easier to read by breaking it into clear sections.
- Search Engine Optimization (SEO). Search engines use these headings to understand the structure and main topics of your content, which can help your page rank better in search results.
In my first magazine job out of college, my editor used to get the entire staff together — sometimes four or five of us — to brainstorm headlines for all of our feature stories. We’d all have spent 30 minutes or so thinking up every headline we possibly could for a particular story and we’d share them during the meeting. Our editor would write them on a whiteboard and we’d discuss the pros and cons of each one, and come up with more throughout the process.
I cannot believe we had that much time on our hands.
ChatGPT can do this in nanoseconds.
I’ll have it read my blog posts and ask it to offer H1s and H2s geared for SEO. A caveat here is that my competitors are doing the same thing, so it’s possible the SEO recommendations coming from ChatGPT are similar to what they’d give my competitors.
Again, this is where I think human brains come into the picture and determine if the AI-generated SEO suggestions align with your strategy.
Step 9: Write Code
Sometimes the information I’m sharing in a blog post is more easily readable in a table format. Tables on a web page are created with HTML, and ChatGPT is fantastic at generating this type of code.
Check out the table in this blog post, Future-Proof Your Business: 6 Marketing Trends for 2024 That You Should Know – Digital Marketing Agency | Madison Miles Media. Head to Tip No. 3, which compares B2B social media tactics from 2019 to 2024. I used ChatGPT to create the code for the table.
| Aspect | B2B Social Media Strategy 2019 | B2B Social Media Strategy 2024 |
|---|---|---|
| Platform Usage | Primarily LinkedIn focused, with Facebook and Twitter for brand awareness. | Diversified approach including Instagram, TikTok for storytelling and company culture. |
| Content Strategy | Professional, formal communication; focus on whitepapers, case studies, industry news. | Engaging, interactive content like short videos, live streams; conversational tone. |
| Personalization and AI | Emerging personalization; basic AI and machine learning usage. | Standard AI-driven personalization; sophisticated use of predictive analytics. |
| Data-Driven Decisions | Basic data analysis tools; focus on content impact and audience reach. | Advanced data analysis tools; emphasis on ROI and engagement metrics. |
| Customer Engagement | Primarily one-way communication; basic interaction with audience. | Dynamic engagement; community building, active participation, Q&A sessions. |
| Sales Integration | Lead generation with off-platform sales conversions. | Direct sales through social platforms; integrated social selling and CRM systems. |
| Influencer Marketing | Mostly a B2C phenomenon. | Significant growth in B2B; leveraging industry leaders, niche experts, micro-influencers. |
| Ethical Practices | Less focus on a brand’s ethical stance or sustainability. | Highlighting commitment to sustainability and ethical practices. |
| Mobile Optimization | Important but not central. | Mobile-first approach; content and campaigns designed for mobile users. |
| AR and VR Integration | Mostly experimental in B2B marketing. | Common use for product demos, virtual tours, immersive experiences. |
While a developer may need to go in and update it with spacing or brand colors, it can offer a good starting point — and save them time.
Step 10: Create Images
I’m a newbie when it comes to using AI to create images, and I’m constantly experimenting. I generally leave image creation to design experts, but when I’m trying to convey a concept I have in my head, based on what I’ve written, I try to give my designers back-of-the-napkin concepts that could accompany the writing.
Below is an example from early 2024. My goal was to create a timeline or comparison of how breast cancer treatments have evolved over 25 years.
I’ve shared a portion of my prompt. I did not share the full prompt to protect the client’s identity.
Are the images creepy? Yes. Did they help me in the creative brainstorming process? Also yes.
The designer and I ended up taking a more conventional, horizontal timeline approach, but the exercise was helpful in the brainstorming process.
Step 11: Write Channel-Specific Copy to Promote the Blog Post.
Writing the blog post is only half the battle. If you want people to read it, you have to distribute it through different channels, like email, social media and influencer outreach.
Once the blog post is written, I’ll ask ChatGPT to:
- Write a first draft of email copy, including subject lines and preview text
- Adapt the email copy to appeal to different buyer personas
- Suggest some social media posts, including multiple posts for different platforms; I’ll ask it to differentiate the posts by focusing on a different point in the blog post.
- Provide some ideas for converting the blog post into video snippets
The Benefit of Using AI to Help With the Writing Process
While these little writing experiments with AI are fun, they’re more than a novelty. They translate into real value for me, as a writer, and ultimately what I’m able to deliver to my clients — particularly in time that is freed up to support more strategic aspects of their marketing efforts.
In the words of Larry McMurtry, writing is indeed about “herding words.” By leveraging AI tools like ChatGPT, I’ve become a more efficient word wrangler, freeing myself up for more creative and strategic elements of my job. 
