Local delivery brands in King County live on visuals. A clean product shot, a seasonal menu banner, or a weekend promo graphic can decide whether someone opens your menu or scrolls past. Many small teams now experiment with AI image generators to fill that gap, and a lot of them start by searching for midjourney prompts that might save them hours of trial and error. The catch is that a prompt that looks great in someone else’s feed may produce something off-brand, misleading, or unusable for a regulated product. This guide walks through how to think about AI prompts as a working tool for a Seattle-area cannabis delivery business, where the output has to be useful and compliant at the same time.
Why most AI prompts fail for product marketing
Generic prompts like “a beautiful cannabis leaf on a green background, high quality” produce pretty images that do nothing for a menu. They lack the specifics a marketing team actually needs: the format, the aspect ratio, the lighting, the brand palette, the space reserved for text, and the mood that fits your audience. When the image lands in a banner, it either crops awkwardly or competes with the headline.
Working prompts tend to share a few traits. They name a subject, a setting, a camera angle, a color direction, and an output size. They also say what to avoid. For a delivery service, that might mean no people consuming products, no imagery that looks like it targets young people, and no claims about effects. Writing those constraints into the prompt saves you from throwing away half of your generations.
Stay inside Washington’s rules before you generate anything
Cannabis marketing in Washington is governed by rules set by the Washington State Liquor and Cannabis Board. Those rules cover what kinds of content can appear in advertising, including restrictions on material that appeals to minors and requirements tied to how and where ads run. The specifics change over time, so check the current WSLCB guidance before you publish, and ask your compliance advisor or attorney if you are unsure about a particular image.
AI tools add a twist. A generator will happily produce cartoon characters, bright candy-style packaging, or playful animals if your prompt suggests them, and none of those are safe choices for a cannabis brand. Build a short internal checklist and run every generated image through it before it goes anywhere near a social post or a delivery app listing. A reasonable checklist might ask whether the image shows anyone who looks under 25, whether it uses imagery associated with children or pets, whether it makes health or effect claims, and whether it resembles packaging for a non-cannabis product.
Build a prompt library instead of starting from scratch
The most useful habit a small team can build is a shared prompt library. Each entry should include the prompt itself, the intended use (menu header, Instagram square, email banner), the model or tool version that produced the best result, and notes on what went wrong. Over a few months, that library becomes more valuable than any single clever prompt because it captures your brand’s specific taste.
Structure each prompt with consistent slots. For example:
- Subject: the product type or scene, such as a tray of pre-rolls, a glass jar on a wooden shelf, or a rainy Capitol Hill street at dusk
- Style: photographic, flat illustration, or muted editorial
- Palette: two or three colors that match your logo
- Composition: where the empty space for text should sit
- Constraints: no people, no packaging logos from other brands, no text in the image
- Format: aspect ratio matched to the placement
Once the slots are stable, you can swap the subject for each new campaign without rewriting everything. That consistency also makes it easier for a second person on your team to pick up the work.
Make your visuals reflect the King County neighborhood
Generic cannabis imagery looks the same whether it comes from Denver, Los Angeles, or a stock library. A King County delivery brand can stand out by grounding visuals in place without making them literal. Think about the textures and light of the Pacific Northwest: wet pavement reflecting neon, evergreen silhouettes over Lake Washington, ferry lights on Elliott Bay, or the soft gray of a typical Seattle morning. These cues feel local and do not require showing any specific landmark in a way that could be confusing for customers. To go deeper, explore The marketplace for AI prompts that actually work.
Be careful with recognizable public places and people. AI tools sometimes produce figures that look like real individuals or invent signage that reads as an official sign. Keep scenes generic, avoid readable text, and double-check any image that includes architecture or street names.
Use AI for drafts, then add human judgment
AI output is a starting point. Plan a review pass where someone on your team checks color accuracy against your actual product, verifies that the image does not misrepresent what a customer will receive, and confirms that the crop works on every platform you use. If your menu shows a flower strain, the photo should not suggest a different look or density. Customers who order based on an image and receive something visibly different will lose trust quickly, and in a delivery business repeat orders are everything.
Pair AI visuals with real photography where it counts. Use generated images for banners, seasonal backgrounds, blog headers, and abstract graphics. Keep actual photos of your packaging, your drivers, and your current inventory for product listings. That mix keeps the brand honest and gives customers a clear sense of what they are buying.
Measure what works and retire what doesn’t
Track which visuals drive clicks to your menu, which ones lead to add-to-cart actions, and which ones get reported or ignored. Your own analytics will tell you more than any generic advice. When a prompt family consistently performs, document it in the library and create variations. When one underperforms, note the reason and move on. Treat the prompt library as a living asset that improves with every campaign.
A practical starting plan
If your team is new to AI visuals, start small. Choose one recurring need, such as a weekly menu banner, and write three prompts with the slot structure above. Generate a batch, run each result through your compliance checklist, and pick the best. Post it, measure the response for two weeks, and adjust. Only after that loop works should you expand to social campaigns or paid placements.
The goal is not to replace your creative judgment with a tool. It is to give a lean King County delivery team a faster way to produce clean, on-brand, compliant visuals so that your time goes into service, product quality, and customer experience, which are the things that keep people ordering.
Leave a Reply