Leah MessengerSure, you've now made AI a big part of your marketing workflows—good start. But, are you getting the best results? Make sure your AI prompting is up to scratch.
So, AI's here to help us marketers... or is it here to take over? The truth is, it's less about what you know or who you know these days—instead, we're talking more about how you ask.
To get the good stuff from your AI assistants, you can't just shout orders; you need to learn to prompt.
When it comes to modern day marketing, it really is go AI or go home. It's officially unavoidable, but not in a steal-yo-job kinda way... if you know how to prompt properly, that is.
Prompt engineering is now another string every great marketer needs to add to their bow. Clear and well-crafted prompts are what takes any AI tool's output to the next level—AKA saving us marketers tons of time, and making our lives significantly easier.
So, it's time to stop shouting "do better" at AI, and start doing better at AI prompts yourself with this handy guide.
See the moment of you giving an AI a prompt as a collab, not a demand. The clearer and more specific your instructions, the better the AI can deliver content that matches your tone, audience, and goals. And you know, that is the kinda stuff that's going to make you look good.
❌ Poor prompt:
You: “Write a headline for our new campaign.”
Your AI: [insert something super generic and forgettable here]
🆗 Okay prompt:
You: “Write a fun headline for our new campaign for our new eco-friendly sneakers.”
Your AI: "Go green in style with our newest kicks"
✅ Better prompt:
You: “Write a bold, benefit-driven headline for a landing page promoting our new eco-friendly sneakers. Audience: Gen Z. Brand voice: playful and rebellious.”
Your AI: “Step lightly, stand loud: Our boldest eco sneaker yet”
Okay, so let's dissect an AI prompt and break it down into the different elements you want to include for the best output from your AI tool (so you stop shouting at it):
Goal or task: What do you want the AI to do?
Context: What’s this for? Is it a campaign, persona-based, or a specific funnel stage?
Audience: Who is the piece aimed at?
Tone and voice: How should it sound?
Format: What form should the output take? Is it a headline, outline, blog, caption?
Constraints: What word count, style rules, and brand language should this stick to?
Fun fact: With Optimizely Opal, you don't have to teach AI every time you prompt about your brand guidelines or specifications—this is something you can configure once so it can continuously refer to the constraints whenever generating output or content.
Treat your AI like that one colleague who takes everything literally. "Write a blog post" is an invitation for disaster. Instead, try: "Write a blog post for marketing managers on the ROI of AI-powered email campaigns, using data from the last quarter." The more specific you are, the less likely you are to question your life choices.
Don't be afraid to state the obvious... with certain tools (ahem, the ones that aren't Optimizely Opal) repeatedly. To get the output you want, you need to offer your AI explicit instructions. Prompts like:
👉 "Summarize this report in 3 concise paragraphs."
👉 "Draft a LinkedIn post that doesn't sound like it was written by a robot."
👉 "Create a subject line that doesn't scream 'spam'."
Be clear, be direct, and prepare to repeat yourself (to the generic AI tools anyway).
Whether your target audience is a room full of academics or a gaggle of Gen Z influencers, you've got to tailor your AI prompts accordingly to get the best, most relevant content. What your AI doesn't know... can absolutely hurt your AI output.
There's no real science to prompting (yet). It's more like throwing spaghetti at a wall and seeing what sticks. We're talking experimentation and iteration.
Try different prompts, tweak your wording, document what's working, and iterate them regularly. It's all part of the fun.
Disclaimer: This section is also something we like to call 'How to not to let your AI go totally rogue'.
Keeping your input a lil too short and sweet: While brevity can be a virtue, starving your AI of context is a recipe for disaster. Make sure you provide substance by specifying the platform, the topic, the desired action, and any other relevant details for a more satisfying AI response.
Being too vague (enough of the "make it sound good", please): Let's be real here—AI is not magic or a mindreader. You're going to need more concrete guidance than that. Specify on tone, key messaging, and specific keywords or phrases you want to include... or stick with Optimizely Opal which 1) remembers you brand guidelines and 2) offers up keyword data to help your ranking.
Leaving out audience or brand voice: If you forget to mention your target audience in the prompt, you are essentially asking AI to write for everyone and no one at the same time. Beef out your prompt with information on your audience's demographics, interests, pain points, and preferred communication styles.
Asking for too much at once: Trying to cream too many requests into a single prompt is a surefire way to overwhelm the system and lead to below average (or just plain confused) results. Break down any complex tasks into smaller, more management prompts.
Skipping review and refinement: AI-generated content is rarely perfect when it comes straight out the box. It's so crucial to review, edit, and refine the output to ensure it aligns with your brand voice, meets your quality standards, and achieves your desired objectives.
Long story short: Make sure you hand over as much knowledge to your AI tool (even if it is just stored in your brain) to prevent it from deviating from its intended purpose.
Role: Who should the AI pretend to be?
Action: What is the AI supposed to do?
Context: What background information or specifics does the AI need to know?
Execution: What is the expected format or output?
Example prompt: Act as a [role]. Your task is to [action]. Here is some context on what you're doing—[context]. I would like you to provide me with [execution].
Describe the goal: What are you trying to achieve?
Refine the audience: Who is this idea for?
Establish tone or style: What should it feel like?
Add references or examples: What should it draw from?
Make the request: What should the AI actually produce?
Example prompt: Your goal is to [goal]. We are targeting [audience]. The tone should be [tone and style preferences]. Here are some examples—[examples]. I would like you to [request].
Primary asset: What content are you starting with?
Audience: Who is this version for?
Style: What should the tone and format feel like?
Transformation: What format do you want this turned into?
Action: What do you want people to do after engaging?
Example prompt: Here is the link to your primary asset—[link]. We will be targeting [audience]. Keep the tone [tone and style preferences]. Transform this asset into [new content format] then suggest relevant call to actions.
As you've seen, mastering AI prompts can seriously level up our creative and strategic potential as marketers. But it is crucial to remember that AI is a tool, not a replacement for human ingenuity. Knowing when to use it—and how to use it—is key to maximizing its benefits.
AI is great for tasks like:
Ideation: Generating a wide range of ideas and concepts to spark creativity.
Drafting: Quickly creating initial drafts of content, saving time and effort.
Variations: Producing multiple versions of existing content for A/B testing or different platforms.
But some tasks are still best left to human intuition and expertise. These include:
Strategic planning: Developing overarching marketing strategies that require a deep understanding of market dynamics and customer behavior.
Emotional connection: Crafting content that resonates with audiences on an emotional level, requiring empathy and nuanced understanding.
Ethical considerations: Making decisions that involve ethical dilemmas or require a strong moral compass.
Read about the dos and don'ts of using AI in our content series, AI across your content lifecycle.