Writing Better AI Prompts for a Vancouver Cannabis Delivery Menu

Many delivery shop owners in Vancouver have tried AI image and text tools, gotten results that look nothing like their actual products, and given up. In most cases the problem is not the tool. It is the prompt. Owners who start from a tested starting point, such as a collection of midjourney prompts, usually spend less time fighting the software and more time deciding what their customers actually need to see.

Why delivery shops struggle with AI tools

A cannabis delivery business has a narrow set of visual and written needs. The menu has to be clear, the product photos have to look consistent across dozens of SKUs, and every piece of copy has to respect the rules that govern cannabis promotion in Canada and British Columbia. A generic prompt like “cool weed product photo, neon, dramatic” produces images that are off-brand, hard to reuse and sometimes misleading about what is being sold.

Good prompts describe the subject, the framing, the lighting, the surface and the mood in plain terms. They also state what should not appear. For a delivery shop, that usually means no people consuming product, no suggestion of health or medical outcomes, and no imagery aimed at young audiences.

What to look for in a prompt library

Before paying for prompts, check whether the seller shows what each prompt produces and whether the prompt has been reused by others with reported results. A useful listing will include:

• The exact prompt text, not just a vague description of the style.
• Notes on which generator it was written for, since syntax differs between tools.
• Example outputs so you can judge the look before buying.
• Clear licensing terms that say whether you can use the output commercially.
• Easy ways to adjust variables such as colour palette, camera angle or background.

Be cautious with any listing that promises guaranteed sales or traffic from a prompt. A prompt is a creative input, not a marketing strategy, and no prompt can guarantee how customers will respond.

Building a consistent product image set

Consistency matters more than any single striking image. A customer browsing your menu on a phone should be able to tell at a glance that every product belongs to the same shop. Start by writing one master prompt that defines your house style, then vary only the product-specific details.

A practical structure looks like this:

1. Name the object clearly, such as a glass jar, a pre-roll tin or a vape cartridge, and avoid implying any effect.
2. Specify the background, for example matte charcoal or pale birch wood.
3. Set the lighting, such as soft overhead diffusion from the left.
4. Fix the camera angle and distance so every product is shot from the same height.
5. Add negative instructions to exclude people, text overlays, brand logos you do not own and any hint of consumption.

Keep a spreadsheet of the prompts you used for each product. When a supplier changes packaging, you can rerun the same prompt with the new details and get a matching image.

Writing menu copy that stays compliant

Text is where delivery shops most often run into trouble. Federal and provincial rules restrict how cannabis can be promoted, including limits on appeal to young people and restrictions on making health or therapeutic claims. Licensing bodies and regulators update their guidance, so verify current requirements directly rather than relying on an AI tool’s assumptions about what is allowed.

When you prompt a text model for menu descriptions, give it the constraints up front: To go deeper, explore The marketplace for AI prompts that actually work.

• Describe the product category, THC and CBD ranges as listed on your verified lab results, and the format or serving size.
• Ask for plain, factual language with no claims about relief, sleep, anxiety or any other condition.
• Ask for no slang or imagery that appeals to minors.
• Request two or three variants so you can choose the most accurate one.

Then read every output against your licence conditions and your current lab data. The AI does not know your shop’s compliance history, and you are responsible for what appears on your menu and in your delivery app.

Social posts without the risk

Social platforms add their own restrictions on cannabis content, and those rules differ from regulator rules. Many platforms restrict paid promotion of cannabis entirely. Use AI to draft evergreen content instead: explanations of how delivery works, delivery zone updates, store hours, packaging and privacy information, and seasonal notices about holidays or road closures in Vancouver neighbourhoods.

A simple prompt for these posts might read: write a short, informational update about our delivery window in the Kitsilano area this weekend, include the order cutoff time, state that customers must be 19 or older with valid ID, and avoid any product claims. Output in plain text, under 60 words.

Keeping a human in the loop

AI speeds up drafting, but it does not replace review. Build a short checklist that someone on your team signs off before anything goes live:

• Does the image show only products, with no people or consumption?
• Does the copy make any health, medical or performance claim?
• Do the THC and CBD figures match the current lab certificate?
• Is the age requirement stated where it must be?
• Would this post be acceptable on the platform where it will appear?

Store approved prompts and approved outputs together. Over time you will build an internal library that reflects your own compliance decisions, which is more valuable than any generic collection.

Testing before you commit

Try any new prompt on a single product first. Generate five or six versions, note which ones match your house style, and adjust only one variable at a time. This approach makes it obvious which change caused an improvement and saves time compared with rewriting the whole prompt after every disappointing result.

Track how customers respond to new imagery through your normal analytics, such as menu click-throughs or order conversion from specific pages. Treat results as directional. Small shops rarely have enough traffic for definitive conclusions from a single test.

The bottom line

For a Vancouver cannabis delivery business, AI tools are most useful when they are given clear, constrained instructions and checked carefully by a person who knows the rules. Start with a consistent house style, write compliance requirements into every text prompt, and keep a record of what you approved. A good prompt library can shorten the learning curve, but your licence, your lab data and your judgment remain the final authority on what goes in front of customers.