

How-To Guides

You wanted a clean lifestyle shot of a coffee mug. But instead? You get an absolutely cursed AI image that’s straight out of a horror movie. Your mug with three handles is being held by someone with too many knuckles. Your logo is melting like wax. And its packaging looks like it was forged in the underworld.

These are the cursed AI images of legend. They’re nightmares born not from hell, but from sloppy prompting. Unfortunately—and perhaps because 70% marketers never get any form of AI training—many people approach generative AI tools expecting an almost "mind reading" experience. People tend to offer broad or vague prompts that leave them blaming the system for irrelevant, awkward or laughable results. This pattern leads to organizational frustration with AI and stifles further adoption.

But fear not, weary ecommerce marketer. Consider this blog your “Don’t Let This Happen To You” PSA. We’re teaching you all the basics, from how to write a prompt, how to build an AI prompt library and how to use Backstroke’s new AI product photo editor for your email campaigns.
By the end of this prompt engineering guide, you’ll have:
A reusable AI prompt library to serve as your brand’s personal spell book for writing the best AI prompts.
A team-wide system for sharing and governing prompts’ quality.
An AI prompt dissection playbook to diagnose and improve outputs. Think of that like your crystal ball.
Let’s grab the salt, light the candle, queue up your favorite spooky song and let’s begin. 🔮
When & Why Prompts Go Feral (Plus How to Fix Them)
AI is powerful, but it can also act like a little gremlin who takes everything way too literally. Without guardrails, your AI-generated product images can drift into chaos.
Common things that result in weird A-generated product images include:
Ambiguity in AI Prompts: The AI fills in the blanks with its own “creativity.” This is where those things like extra limbs and floating forks come from.
Missing Brand Rules: Your “minimalist” brand suddenly has neon rave backgrounds because you haven’t trained the AI on your creative guidelines.
Vague Product Specs: You asked for “shoes” and it gave you . . . hooves. You need to be extremely specific with what your products look like.
No Negative Prompts Exist: This means you never told it what not to do (see: haunted packaging above). Negative prompts prevent elements that clash with brand guidelines (colors, styles, unwanted logos) or that could cause customer confusion.
Wrong Aspect Ratios: This can be easy to forget, but still critical, so that TikTok video you need doesn’t end up as a square.
Some quick exorcisms—AKA AI prompt quick fixes—are as follows:
Add More Constraints: Include things like “true-to-life color” or even hex codes to make your AI-generated product images as accurate as possible.
Set the Scene: Include details on the environment, time of day and even the general vibe. It all helps craft the product image or video you’re looking to create.
Specify Lens & Lighting: Want a shot in 35mm? Prefer a soft, warm box light coming from the upper-right corner. Specify that! And if you need more tips on how to include great lighting copy in your AI prompts, check out this blog 10 Stunning Lighting Techniques for AI Art.
Declare Do-Not Lists: It bears repeating. Include negative prompts for an extra bit of brand safety. “No extra fingers, no watermarks, no warped logos.”

If you’re writing a totally fresh, new prompt, we recommend taking a look at our blog 6 Marketing Toolkit Essentials That AI Experts Swear By to learn how to build the best prompts for your marketing campaigns.
The Basics of Building Your Brand’s AI Prompt Library
If you want consistent results across campaigns, your brand needs a centralized library of AI prompts that actually work for your brand. Think of this as a spell book that your whole team can use when prompting AI, without summoning chaos.

Before we dive into building your actual prompt “spell book,” let’s get real about where it should live. An AI prompt library is basically just a shared, well-organized folder of your brand’s best prompts. Think of it like a recipe book that everyone on your team can cook from.
You don’t need fancy software to start. Use a Notion page, a Google Doc, a Confluence space or even a neatly structured Google Drive folder. The key is to prioritize organization.
Set up sections for “Foundations” (brand voice, colors, rules), “Prompt Blocks” (modular templates you can reuse) and “Examples” (before/after outputs with their prompts).
Use clear names and version numbers (like Mug_Lifestyle_v2) so no one accidentally resurrects a cursed, old prompt.
And most importantly, make sure it’s shared with your whole team, because an AI prompt library isn’t useful if it’s gathering dust in one marketer’s laptop.

With that in mind, here’s how to build an AI prompt library that hits every time for your brand.
Foundational Incantations: Your One-Time Definitions
Brand Essence: Write out guidelines that teammates can follow when detailing your brand voice, vibe, audience and “taboos.”
Visual House Rules: Include notes on your brand color hexes, approved backgrounds, lighting moods, lens ranges.
Product Taxonomy: Include SKUs, finishes, hero angles and compliance notes, so each prompt refers to a unique, unambiguous product.
Do/Don’t Lexicon: Note any banned adjectives (no “quirky” if you’re luxury) or off-brand tropes. Include must-use phrases.
Prompt Blocks: Modular Templates to Pick-And-Choose From
Think of these items as Legos for prompts. List them in your prompt library, so teammates can stack them together for different use cases:
Product Blocks: name, model, material, standout feature
Scene Blocks: environment, surface, props, mood
Style Blocks: studio/lifestyle/macro, lens, lighting, post-processing
Brand Block: logo integrity, brand colors, composition guardrails
Negative Blocks: “no extra limbs, no text, no warped logos, no watermarks”
Output Block: aspect ratio, resolution, file naming schema.
Base Prompts & Seasonal Overlays: Include information on how your brand handles holidays, adapts messaging throughout the seasons, etc.
Store A/B Prompt Variants: Label any variety within your prompts by objective, such as driving CTR, PDP conversion, etc.
You also need to maintain a monthly audit ritual for your prompt library. Auditing this data is about making sure that the prompts within your library are solid, safe and trustworthy, producing dependable product images without surprises or mistakes.
Your Field Kit (Swipe Files)
We also suggest that you create a kit of quick, pre-written “swipe files” that team members can copy, paste and tweak without starting from scratch. These are ready-to-use formulas that give beginners a solid jumping-off point while keeping AI-generated product images consistent and brand-safe.
Below are a few examples of simple file prompts to describe ecommerce product photography. AI prompts like these should be easily accessed in your swipe file:
Studio PDP: Product block + “on seamless [color], softbox from camera-left, 35mm, sharp focus, true-to-life color, logo intact, no distortions, no text, 4:5, 3000×3750.”
Lifestyle Social: Product block + “sunlit kitchen counter, morning glow, light crumbs, human hand cropped at wrist, 50mm, shallow DOF, warm grade, 1080×1350, safe margins for copy, no extra fingers.”
A Negative Prompt Starter: “No extra limbs, no warped logos, no watermarks, no misspelled text, no floating objects, no unrealistic reflections, no duplicate products, no disembodied heads.”
Close the Circle With AI Content Studio
At the end of the day, great prompting isn’t sorcery. It’s systematic and specific. Build your AI prompts library, share it with your team, run every render through preflight and you’ll never again face the horror of a five-fingered spatula ever again!
You don’t have to conjure all this up alone. Enter Backstroke’s AI Content Studio.

Need gorgeous, photoshoot-quality product images? That’s exactly what this tool was built for. With AI Content Studio, email marketers can create editorial level product photos without photography or engineering experience. You can generate on-brand visuals matched to subscriber clusters preferences, test them across segments, iterate instantly and drive 31% more revenue per send.
Learn more about AI Content Studio and other other next-gen visual tools here.
And if you want to send the demons packing across your entire email program—audience targeting, generative content creation and more—talk to us.