Generative AI has completely changed how I work within content. This may sound biased coming from the Head of Content at Copy.ai, but if you know me at all, you’ll know that I’m not here to pitch you.
I’m here to show you:
Let’s jump right in!
Before we get into the nitty-gritty, here are the best practices to follow when you’re working with generative AI prompts—whatever the purpose.
Generative AI is kind of like having a tiny freelance writer in your pocket. It can take a lot off your plate, but it can’t read your mind. You wouldn’t expect a bare-bones brief to summon an exceptional piece of content, would you?
The quality of the output always depends on the quality of the input.
In my experience, it takes a lot longer to wade through a flurry of inaccurate, off-base, or unusable generated content than it does to spend a few minutes fleshing out your prompt.
When writing out your prompt, make sure you:
💡 Pro tip: If the outputs keep missing the mark, stop and conduct the “sheet of paper” test:
If someone handed you a sheet of paper containing the exact same instructions as your prompt—and nothing else—would you feel you had enough information to complete the request? This is a great exercise because it forces you to stop and make sure you aren’t missing any critical details. Do the test enough, and it will become like second nature—promise.
Have you ever listened to someone try to explain a super complex topic, and your brain just couldn’t deal? Then, just as you accepted that you’ll never understand quantum physics, your teacher dropped in a Mean Girls reference, and suddenly everything clicked into place.
Turns out, you weren’t incapable at all. You just had primitive knowledge of the subject and needed to have it explained in a different way.
That’s sort of like generative AI.
It’s capable of retrieving, collecting, processing, sorting, compiling, and re-organizing vast amounts of data, but needs the right prompt to successfully determine the appropriate configuration to meet your request. Changing up your prompts can help provide more clarity and context—and get you a better result in the end!
Some of my go-to synonyms for content creation:
Causative language helps define cause-and-effect relationships in your prompt, so generative AI can use information contained in one element to modify another—like rewriting landing page copy to align with your particular brand voice, or generating a Twitter thread based on a blog post URL.
A shortlist of causative language I like to include in prompts:
Generative AI is like a dog that only responds to positive reinforcement. If you tell it not to do something, it’s likely going to do that thing. This is an easy fix—just describe it using positive language.
Let’s look at how we can make these positive language changes:
Make no mistake: as much as I love incorporating generative AI into my workflow, it isn’t foolproof. Outputs are never publish-ready, so it’s important to manually review and fact-check everything, down to the last social media caption.
Now, let’s get to the good part: my tried and true process for AI-assisted content creation, including prompts I’ve found especially useful at various stages of my workflow, from blog post ideation to distribution.
🚨Quick caveat: It’s always better to test and tweak prompts to fit your specific needs, rather than using every one verbatim.
Coming up with ideas for SEO-focused blog posts is tougher than it looks. It’s not just breaking down high-level topics in your marketing strategy or conducting exhaustive keyword research. You also have to find that interesting twist or value-add to prevent your content from vanishing into the Google desert—don’t forget, more than 60% of desktop searches result in zero clicks.
For me, the first step is usually sharing two or three target keywords in Chat, then asking it to come up with relevant blog post topics. Ideally, I’ll wind up with a solid list that I can whittle down and refine. But even if I don’t like any of the suggestions, it still helps me bypass the blank screen and get straight into “brainstorm mode,” so it’s easier to come up with alternatives.
✍️ Prompt: Suggest 15 blog post topics related to the keyword [KEYWORD]. Include a brief description of the post and how it will help readers.
💡Pro tip: If you’re in brainstorming mode, don’t get too specific. With fewer restrictions, AI will generate a broader and more diverse range of content ideas.
Another thing I like to do is nab the URL from an existing blog post that’s performing well and request a list of ideas for similar posts.
✍️ Prompt: List 10 ideas for more blog post topics, similar to this post: [URL]. Include a suggested title and a brief description of the topic.
Ironically, a lot of apprehension around generative AI stems from its biggest fans—the leaders out there thinking, Who needs content marketers when we can just fire up our handy blog post generator, crank out a ton of content, and watch the leads roll in?
The thing is, that’s not how generative AI actually works. Can it supercharge content-led growth? Absolutely! But not in the way a lot of people think. Generative AI doesn’t replace content marketers or writers.
It replaces those tedious, cumbersome tasks that slow us down and eat into the time we have to spend on high-value work. Generative AI’s biggest contribution to content engines is taking menial tasks off our plates, so we can focus more of our efforts on actually driving pipeline.
When it comes to SEO-focused blog posts, a lot of the “grunt work” happens before a single paragraph is written. It’s all the time you have to spend researching, organizing your ideas, plotting out keywords, and occasionally just staring at your screen.
Here’s how I’m currently using generative AI to expedite the process:
I usually kick things off by generating an outline directly in Chat. My prompt typically includes exactly what you’d expect:
If I’m not happy with the output, I’ll keep tweaking my prompt until I have a solid foundation to work with. Here’s a starting point for you:
✍️ Prompt: Generate an outline for a blog post about [TOPIC] for [PERSONA], highlighting [main point you want to make].
If you’ve dabbled in generative AI, you probably know it’s better at broader ideas than tiny details. Whenever I’m working with an AI-generated outline—just as I would with any AI-generated content—I go through and add to it. This can look like elaborating on any points, removing certain suggestions, and adding any additional information I want covered.
At this stage, I’ll also use generative AI to come up with a list of stats, quotes, and real-world examples, which I manually fact-check before incorporating into any brief.
💡Pro tip: Create a spreadsheet for tracking all your AI-generated stats. This makes it easy to remember which ones have already been fact-checked and also serves as a handy reference for repeated citations.
If I’m planning to hand over the outline to a freelance or in-house writer, I’ll sometimes go through and flag any opportunities to incorporate brand positioning, messaging, internal links, or pain points that are relevant to the piece we’re producing.
You can use AI to dramatically scale content creation—the team at uSERP, for example, used Copy.ai to write 50-100 blog posts per month!
Here’s how I’ve fine-tuned my process for creating a single blog post:
I take my new and improved outline (see previous section) and drop it in the chat, prompting AI to generate a blog post. I always treat this as a very rough first draft—even with all the detail I’ve added to the outline, the results are still too generic for my liking.
✍️ Prompt: Using the outline below, write a blog post up to 1, 200 words. The blog post should be written for [target audience].
Outline: [Paste outline]
Similar to my approach during the outline stage, I treat the AI-generated blog post as a starting point. Then, I comb through each section to elaborate on key points, provide more context to use cases, and weave in our messaging and unique brand perspectives. If it’s a tactical post, I sprinkle in suggestions, tips, and tricks. This is also a good time to add insights from Subject Matter Experts (SMEs).
If I get stuck, I may hit up Chat again to generate examples, headlines, or introductory paragraphs for different sections. It’s important to note that at this stage, the post is a hot mess—a discombobulated mix of generative AI and my own “sloppy copy.” My main focus is getting everything down on the page. The wordsmithing comes later.
💡Pro tip: As you copy/paste outputs into your main working document, use a different text color (or italics) to clearly indicate AI-written content. This makes it easier to take care of edits and fact-checks in one fell swoop, so you don’t have to stop in the middle of writing and modify as you go.
By this point, I have my “V2” draft that contains a fully fleshed-out blog post. I know some content marketers who prefer to stop using AI AI at this point and make all the edits themselves, as if the draft was just submitted by a freelancer. But, for me personally, I have a tendency to reread a sentence or paragraph over and over again and ask myself, “Did I word this right? Does this make sense?”
That’s where I think AI can be a great unblocker in the editing process.
For instance, you can take the draft and prompt AI to rewrite it to sound consistent with your brand voice.
One of my favorite features is Copy.ai’s Infobase, which lets you create reusable “blocks” of information about your product and brand, and then insert them in your prompts. However, you can also simply describe the appropriate tone. I also ask Chat to check for confusing jargon, clunky sentences, and grammatical errors.
✍️ Prompt: Rewrite this [CONTENT] about [TOPIC]. Identify and suggest improvements for confusing statements. Ensure the tone is friendly, engaging, and conversational, but also professional. Use active voice. Use the Oxford comma. Write in the first person.
Last but certainly not least is content distribution. Three things I like to do at this stage:
Gone are the days when you could just drop the link to the same blog post into multiple channels and watch the traffic flow in. Since users have distinct preferences around subject, style, and format depending on the platform, social media promotion can be a major time sink. One of the most helpful prompts I’ve found is asking for channel-specific repurposing suggestions.
✍️ Prompt: Suggest ideas for repurposing this BLOG POST on Facebook, LinkedIn, Twitter, and TikTok. List 5 ideas for each platform. BLOG POST: [URL]
Although generative AI’s “big picture” focus requires special attention when you’re writing a detailed blog post, it also makes it perfectly suited for crafting short and snappy social media captions. All I do is drop the URL into Chat and specify the social media platform and desired tone. It then scans the blog post and distills key points in a channel-friendly format.
✍️ Prompt: Write a Twitter thread in a snappy, fun, and conversational tone. It should contain a least 3 tweets highlighting best practices featured in this blog post: [URL]
Email newsletters are a great distribution channel. And when it comes to the content within those newsletters, it’s incredibly helpful to have a solid prompt up my sleeve. It also comes in handy for internal promotions—anytime I want to share new content with the rest of the team, I can just drop a quick summary on Slack.
✍️ Prompt: Generate email copy in a casual tone promoting this blog post: [URL]