Writing AI Prompts That Actually Work for a Spokane Cannabis Delivery Brand

Many small cannabis delivery teams in Spokane are discovering that AI image tools can produce product mockups, seasonal banners, and social graphics without hiring a photographer for every campaign. Before you buy midjourney prompts from anyone, it helps to understand what separates a prompt that produces usable work from one that produces a folder full of images you cannot publish. This post looks at that gap from the point of view of a local delivery business that has to stay inside state advertising rules.

Why generic prompts fail for local delivery brands

A prompt like “beautiful cannabis product photo, modern style” will generate something, but it rarely generates something you can use. The output tends to be generic, the lighting does not match your brand, and the packaging looks nothing like what your customers actually receive. A delivery service also needs images that work in a small square on a phone, in a menu thumbnail, and in a banner at the top of a web page, which means composition matters as much as subject.

The gap is usually not the tool. It is the specificity of the instructions. Prompts that work tend to be written more like a creative brief than a wish.

What makes a prompt reliable

In practice, a prompt that performs consistently usually includes most of the following:

• A clear subject with physical details, such as a matte black jar on a walnut tray, rather than a vague product category.
• Lighting direction and quality, such as soft window light from the left, which keeps results consistent across a batch.
• Camera framing, including lens focal length and whether the shot is top-down or at eye level.
• A color palette tied to your existing brand, so every image feels like it came from the same shop.
• An aspect ratio set for the specific placement, such as 4:5 for social feeds or 16:9 for a website header.
• Style constraints that you test once and then keep, instead of changing words every session.

Keep notes on which phrasing produced which result. A prompt that worked in March may drift after a tool update, so treat your prompts as versioned assets rather than one-off experiments.

Compliance comes before aesthetics

Cannabis advertising is a different problem from general retail marketing. In Washington, the Liquor and Cannabis Board sets rules on cannabis advertising, and those rules restrict content that is designed to appeal to minors, that makes health or therapeutic claims, or that depicts consumption in certain ways. Advertising platforms also apply their own cannabis policies, and those can change without much notice. This is general information, not legal advice, so check the current WSLCB rules and your platform policies directly before publishing anything. To go deeper, explore The marketplace for AI prompts that actually work.

For AI-generated imagery specifically, a few habits reduce risk. Avoid cartoon characters, animals, bright candy-like packaging, and playful mascots, since these can read as appealing to children. Do not generate people who look under 21, and do not create images that show or imply consumption. Keep product images accurate to what you actually sell, because an AI tool that invents packaging, potency labels, or strain names creates a problem you will have to explain later.

A practical workflow for a small team

1. Write a one-page brand brief covering colors, fonts, tone, and the words your compliance reviewer has approved.
2. Draft a base prompt for each recurring format, such as a menu header, a delivery-zone announcement, and a product card.
3. Generate a small batch, then compare outputs against a checklist: accurate product, no minors, no consumption, correct ratio, and on-brand colors.
4. Have one person other than the creator approve each image before it goes live.
5. Save the prompt, the seed or settings if your tool supports them, and the approved output together in one folder.

This sounds like more work than typing a sentence and hitting generate, but it saves time once the library is built. Most of the effort is front-loaded into the first few formats.

Common mistakes to avoid

• Writing prompts that describe a mood without describing the object, which produces images that could be for any product.
• Changing too many variables at once, so you cannot tell which change improved or ruined a result.
• Skipping the human review step because the output looks polished at a glance.
• Using text inside images. AI tools often misspell words, and misspelled product names or warnings are a serious problem in this category.
• Assuming a prompt that worked for a different niche will transfer cleanly. Lifestyle and fashion prompts often include imagery that is not appropriate for cannabis.

Building a prompt library you can rely on

Over time, a useful prompt library becomes one of the most valuable assets a delivery brand has. It cuts the time spent on each new graphic, keeps visual identity consistent across staff who may change, and gives a reviewer a clear record of what was approved. Organize it by format rather than by tool, since tools change faster than your storefront layout does. Note the date each prompt was last tested, and retire any prompt that no longer produces compliant output.

If you are short on time, you can also look at existing prompt collections, as long as you test each one against your own brand brief and compliance checklist before using it. Whatever source you choose, the prompt is only the starting point. The approval process and the notes you keep are what turn a set of AI experiments into a dependable part of your marketing for a Spokane cannabis delivery service.