AI prompts for product design, 24 to copy (2026)

A good AI prompt for product design defines the problem, the user, the constraints and the deliverable, not just how the object should look. This library has 24 prompts to copy into ChatGPT, Claude, Copilot or Gemini: physical products, interfaces and hypotheses. To take it into your work, see the Online Program in AI for Creatives.

Add Founderz as a preferred source

What you’ll take from this

This is what you will find in the library, sorted so you reach the prompt you need quickly:

• 24 AI product design prompts ready to copy, with brackets for your own case.
• Prompts for physical products (render, packaging, catalog photo) and digital products (mockups, onboarding, hypothesis generation), with the output format defined in each one.
• Prompts for ChatGPT, Claude, Copilot and Gemini that use what each tool does best.
• The four prompt levels, from asking for one image to testing a product hypothesis.
• A method for writing your own prompts, with a fields table and a before and after example.

ChatGPT prompts for product design

These are the examples of effective prompts I use to design products with AI, sorted by what you want to achieve. Each one is written to copy and paste into ChatGPT and works in other tools with small changes. The text in brackets is my example: replace it with your case and keep the rest. Library reviewed in October 2026.

Physical product prompts: render, packaging and catalog photo

For a physical object, the prompt needs shape, material, finish, human scale and context of use. The more concrete it is, the less generic the answer.

Product render: minimalist smartphone

Act as an industrial designer specialized in consumer electronics.
Context: I am designing a [minimalist smartphone] for [users who value simplicity and want to use their phone less].
Task: generate a concept render.
Constraints: [a single camera, aluminum frame, no visible physical buttons, one-handed ergonomics].
Style: clean, monochrome, matte finish.
Format: studio render, front and back views in the same image, [neutral gray] background, 16:9 aspect ratio.
Avoid: real logos, unreadable text, harsh reflections.

Specialty coffee packaging

Act as a packaging designer for food brands.
Context: specialty coffee brand with [medium] roast, [young urban] audience, sold [online], premium but accessible perception.
Task: generate a front packaging concept.
Constraints: bag with a valve, [two-color] palette, no complex illustrations, space for [origin name and tasting notes].
Style: minimalist, bold typography, high contrast.
Format: front mockup on a neutral background, studio photography, square aspect ratio.
Avoid: real brands, invented and unreadable text.

Catalog image: Nordic design lamp

Create a catalog image of [a Nordic-design table lamp] for the online store of [a furniture brand].
Material and finish: [light oak wood, white linen shade, warm light on].
Scene: [minimalist living room, wooden side table], natural side light, medium shot at eye level.
Realistic catalog photo, 50mm lens, slightly blurred background, 4:5 aspect ratio.
The product is the hero: shape and material must read clearly.
Avoid: decor objects that compete with the lamp, text, logos.

Reusable bottle: studio photo on a neutral background

Realistic product photo of [a 20 oz stainless steel reusable bottle with a bamboo lid], [matte sage green] finish.
Neutral [light gray] background, soft studio lighting with a subtle shadow under the product.
Eye-level close-up, 100mm lens, everything in focus, 1:1 aspect ratio for an online store listing.
Real steel texture, no harsh reflections.
Avoid: logos or text that are not on the product, extra objects, hands.

Digital product prompts: mockups and interfaces

For an interface, the screen, the components, the state and the platform matter. Ask for one screen per prompt and describe the visual hierarchy, the target users and the expected output.

Habit tracker app mockup

Act as a digital product designer with mobile app experience.
Context: habit tracker app for [people who want to build a daily routine], platform [iOS].
Task: design the main screen.
Components: [habit list, daily progress in a ring, highlighted add button].
Style: minimalist, light mode, accents in [a single color].
Format: phone mockup, a single screen, legible text in English.
Avoid: more than four elements per block, decorative icons that add nothing.

Three-step onboarding screen

Design an onboarding screen for [a personal finance app] aimed at [users aged 25 to 40 opening it for the first time].
Structure: three steps with a progress indicator, a short title per step and [a highlighted primary button and a secondary link to skip].
Clean style, light mode, sans serif typography, generous spacing.
Format: phone mockup with the three steps in a row, legible text in English.
Constraint: ask only for [the name and the savings goal] in the first step.

Level 3 prompts: product hypotheses and variations

Here the prompt stops starting with “design” and starts with “I have this problem”. The AI does not give you a pretty screen: it gives you hypotheses you can test. The judgment is still yours.

This approach lets you use AI across the whole discovery process: large language models sort user feedback, run hypothesis generation and draft testable hypotheses, and you prioritize them for the roadmap. Use these prompt templates as a starting point, not as final answers, and refine them with your own data. Validate each AI-generated hypothesis with real users before any product decision. A product manager or UX researcher should check root causes, and share the table with stakeholders so the team agrees on the evaluation criteria.

Three hypotheses to reduce onboarding drop-off

Act as a senior product designer.
Context: users of [a personal finance app] drop off at first sign-up because we ask for [too much data]. Today [45%] leave before finishing.
Task: propose [three] onboarding redesign hypotheses that reduce initial friction.
For each hypothesis, state: the change, the assumption it tests, the metric to measure it and the main risk.
Format: table with one row per hypothesis. Then recommend which to test first and why in two sentences.
Constraint: no hypothesis may need more than [two weeks] of development.

Variations of one concept to compare

I have this concept: [a 20 oz stainless steel reusable bottle with a bamboo lid].
Generate [four] variations to compare, changing only the material and finish: [matte steel, polished steel, anodized aluminum, glazed ceramic].
Keep the shape, scale, camera angle, light and background the same.
Format: one image per variation, same framing, 1:1 aspect ratio, [neutral gray] background.
Add one line to each image about what brand perception that finish conveys.

Logo for a product brand

Create [four] logo proposals for [a specialty coffee brand] aimed at [a young urban audience].
Style: [typographic, minimalist], [two colors], no complex icons.
Format: white background, each proposal in its own box, exact spelling of the name "[NAME]".
Avoid: symbols from existing brands, extra or misspelled letters.
Then explain in one sentence what each proposal conveys.

Claude prompts for product design

Claude does not generate images. It is the tool I use when there is a lot of text to read and organize: interviews, reports, feedback and briefs. Paste the whole document into the chat or attach it, and ask for the output in a specific format.

From interview transcripts to product opportunities

Act as a product researcher.
I am attaching [the transcripts of 8 interviews with users of our delivery app].
Task: identify the [five] most repeated problems. For each one, state how many interviews mention it, a short verbatim quote and a product opportunity.
Format: table with the columns Problem, Interviews, Quote and Opportunity. Then three questions I still cannot answer with this data.
Constraint: use only what the transcripts say and flag when something is your own inference.

Critical review of a concept before prototyping

Act as a skeptical product reviewer.
Here is the concept: [a 300-word description of a travel coffee maker with reusable pods], aimed at [people who travel for work].
Task: run a pre-mortem. Imagine the product failed at 12 months and explain the [five] most likely causes (use, manufacturing, price, channel, competition).
For each cause, propose a cheap test I can run in [two weeks].
Format: numbered list, one cause and one test per item. No praise and no closing summary.

Turn a brief into a visual prompt

Act as an art director.
Here is the product brief: [paste the brief here: audience, problem, constraints, brand tone].
Task: write a prompt to generate a concept render in an image generator.
Structure the prompt in these fields: product, material and finish, scene, light, camera, format and avoid list.
Format: the prompt in a text block ready to copy, and below it two lines with the words I could change to explore variations.
Constraint: do not invent features that are not in the brief.

Prioritized hypothesis backlog

Act as a product manager.
I am attaching [the customer feedback report from last quarter].
Task: extract [ten] product improvement hypotheses in the format "We believe [change] for [user] will achieve [result]".
Score each from 1 to 5 on impact, confidence and effort, and sort them by impact times confidence divided by effort.
Format: sorted table with a one-sentence justification column.
Constraint: mark as "weak data" any hypothesis that rests on a single complaint.

Copilot prompts for product design

Copilot works inside Word, Excel, Teams and Outlook with your own files and meetings. It is useful when the product material is already in Microsoft 365. File references depend on your license and on what Copilot can see in your account.

Word: product brief from your notes

Use the open document [Product ideation notes] as the source.
Write a one-page product brief for the design team.
Sections: user problem, audience, goal, constraints (materials, cost, deadline), must-have requirements, out of scope and success criteria.
Direct tone, short sentences, no jargon. If data is missing, write "pending" instead of inventing it.

Excel: prioritize product hypotheses

I have a table with the columns Hypothesis, Impact (1 to 5), Confidence (1 to 5) and Effort (1 to 5).
Add a column with the ICE score, calculated as impact times confidence divided by effort, and sort the table from highest to lowest.
Highlight the top three in green and explain the formula you used in one sentence.
Then tell me which two hypotheses you would test first and why, given the team has [two weeks] and [one designer].

Teams: from the review meeting to design requirements

Summarize the meeting [Prototype v2 design review] from [this week].
Extract: decisions made, design changes requested (and who asked), open questions and tasks with owner and date.
Format: four lists with those titles, one line per item.
Do not include opinions that did not become a decision. If there is no date or owner, write "unassigned".

PowerPoint: concept presentation for the client

Create a [8]-slide presentation from the document [Product brief v3] to present the concept to [the product committee].
Structure: problem, user, concept, three variations, risks, proposed test and the decision we need.
One message per slide in the title, at most [three] bullets, speaker notes with the argument.
Professional, direct tone. Do not add data that is not in the document.

Gemini prompts for product design

Gemini works well with reference images and with Google Workspace (Docs, Sheets, Drive). If your team works there, these four prompts use that integration.

Analyze a prototype photo

I am attaching a photo of my prototype of [a wooden desk organizer].
Act as a product designer and review what you see: shape, proportions, visible materials, ergonomics and possible manufacturing problems.
Format: three lists (what works, what I would change, what I would ask the manufacturer), five points maximum each.
Constraint: describe only what is visible in the photo and say when something cannot be judged from an image.

Google Docs: product requirements document

Use my Drive document [Q4 product ideas] as the source.
Write a requirements document for [version 1 of a connected kitchen scale].
Sections: context, user, problem, measurable goals, functional requirements, non-functional requirements, out of scope, risks and open questions.
Format: clear headings, short lists, a requirements table with priority (must, should, could).
Do not invent figures: use "to be defined" when data is missing.

Edit a product image without changing the product

Use the attached photo of [my table lamp] as the reference. Keep the shape, color, proportions and materials of the product exactly the same.
Change only the environment: [a light wood desk next to a window, with a cup and a notebook].
Match the direction and color of light on the product to the new scene so it does not look like a cutout.
4:5 aspect ratio, catalog photo quality.
Avoid: changing the product, adding logos or text.

Google Sheets: compare concepts against criteria

In my sheet [Concept comparison] I have three concepts in the rows and these criteria in the columns: manufacturing cost, ease of use, differentiation and timeline.
Propose a 1 to 5 scale for each criterion and fill in a first score from the notes in column F.
Add a column with the weighted average (cost 30%, use 30%, differentiation 25%, timeline 15%) and one sentence justifying each score.
Mark any score without support in my notes as "assumption".

More prompts for product photos and catalogs

Many product searches ask for online store photos, backgrounds and catalogs. These three prompts work in ChatGPT and also in Gemini and Copilot.

Scene background for a real product

Use the attached photo of my product [moisturizer in an amber glass jar] as the reference. Keep the shape, color, label and proportions exactly the same.
Place it on [a white marble countertop in a bright bathroom, with a eucalyptus sprig and a folded towel].
Morning light from the window, 50mm lens, slightly blurred background, 4:5 aspect ratio for social media.
Avoid: changing the product, text or logos that are not on the label.

Advertising image of a product

Create an advertising image of [white running shoes with a lime green sole] for a campaign aimed at [urban runners aged 25 to 40].
Scene: [wet asphalt at dawn, the shoe in the foreground], warm side light, 35mm lens, low angle.
Composition with free space on the left for a headline, 16:9 aspect ratio.
Realistic photographic style, vivid but natural colors.
Avoid: real brand logos, text in the image, deformed hands or legs.

Product catalog: consistent listings

Act as the catalog manager of [an online homeware store].
Here is the list of [12 products] with name, material and dimensions: [paste the list here].
Task: for each product, write a listing with a title of at most 60 characters, three benefit bullets and a catalog image prompt ([white] background, studio light, [three-quarter] angle, 1:1 aspect ratio).
Format: table with the columns Product, Title, Bullets and Image prompt.
Use the same background and light style in every prompt so the catalog is consistent.
Do not invent features that are not in my list.

How I built these prompts for product design

There is no magic in a good prompt: it is prompt engineering applied to design, a well-written brief that lets you create effective prompts and design with artificial intelligence in a repeatable way. If you know how to write a commission for a designer, you already have half of it. The AI does not guess what you do not give it, which is why every prompt on this page follows the same fields, in this order:

Refine each prompt: iterate on the output, validate it against the prototype and the user feedback, and prioritize what to test next. Use AI as a tool for product discovery, not as a replacement for UX research.

Five practical tips for writing your own:

• Start with the problem, not the object. “Users abandon sign-up” gives more material than “design a pretty app”.
• Change one variable at a time. That way you isolate which word controls each part of the result.
• Replace vague adjectives with references. “Modern” says nothing; “matte steel, bold typography, side light” does.
• Ask for the output format. A table, a mockup or a numbered list avoids long, unusable answers.
• Add an avoid list. It is the line that fixes the most defects: invented logos, unreadable text, deformed hands.

A weak prompt made strong

This is the classic example. The weak prompt asks for “something modern” and gets exactly that:

Design a modern, nice bottle.

The same request with the seven fields:

Act as an industrial designer of sustainable homeware.
Context: bottle for [people who bike to the gym] who want to stop using single-use plastic.
Task: generate a concept render.
Constraints: [20 oz steel, bamboo lid, fits a bike bottle cage].
Style: [minimalist, matte sage green finish, soft studio light].
Format: studio photo on a [neutral gray] background, three-quarter view, 1:1 aspect ratio.
Avoid: real logos, unreadable text, harsh reflections.

The second leaves nothing to chance: you know who it is for, which limits it respects and how to check whether it works.

What are the 4 levels of AI prompts for product design?

Not all prompts aim at the same thing. There is a maturity ladder, from asking for a single image to guiding product strategy. The Hero Camp popularized a four-level model that fits how real teams work.

Most people stay at level 1: they ask for an image and use it or discard it. The jump in value is at level 3, where you stop asking for images and use artificial intelligence to think through the product from a real problem. This ladder opens up exploration that used to take days of manual work.

Which AI tool should you use for product design?

No single AI tool does everything well: each one stands out in a phase, whether you are a graphic designer, a product team or a catalog manager. Text models shine at briefing and hypotheses; image tools turn that text into visuals. For prompts on the visual side of a product, see AI prompts for art direction. Plans change often, so check the current one before deciding.

The practical takeaway: start the flow in a text tool, generate images in the one that best fits your style and finish the deliverable where you can present it clearly.

Use the same templates with any LLM, then bring the output into Figma or another prototyping tool to iterate on a testable prototype. Large language models are good at drafting; you refine, validate and decide.

How the prompt changes in ChatGPT, Gemini, Copilot, Claude, Midjourney, Stable Diffusion and DALL·E

• ChatGPT: understands full sentences, accepts reference photos and edits the image in the same conversation. Write the aspect ratio in words and ask for one change at a time. This library works best here as is.
• Gemini: accepts reference images and conversational editing, so the same prompts work unchanged. Repeat the instruction to keep the product if you make several edits in a row.
• Copilot: creates images from the chat and, inside Microsoft 365, works with your documents, sheets and meetings. Name the file it should use as the source.
• Claude: does not generate images. Use it to read long documents, write the brief and improve the prompt, then paste the result into an image tool.
• Midjourney: prefers short visual phrases separated by commas and parameters at the end: –ar for aspect ratio and –no for what to avoid.
• Stable Diffusion: supports weights in parentheses and its own negative prompt field, so the avoid list goes in that field.
• DALL·E: responds well to natural language; describe the scene as if telling a photographer.

This is the bottle prompt adapted to Midjourney:

studio product photo of a 20 oz matte sage green stainless steel reusable bottle with bamboo lid, three-quarter view, neutral gray background, soft studio lighting, subtle shadow, 100mm lens, photorealistic --ar 1:1 --no logos, text, harsh reflections

How to fit AI prompts into your design workflow

AI adds the most when it enters at a specific point of the workflow, not when you try to make it do everything. The natural moment is exploration: between the brief and the concept, where you need many quick, cheap options to discard. A realistic workflow has three moments:

1. Briefing. Use ChatGPT or Claude to structure the request and sharpen the problem before thinking about images.
2. Generation. Turn the brief into visual prompts and produce concepts in your preferred image tool.
3. Presentation. Select, adjust and lay out the project to defend it in front of the team or the client.

Automating repetitive tasks is the clearest gain: ten color variations, test mockups or catalog listings take minutes. But the decision about which concept moves forward is still yours.

Best practices for product work: ask the AI to generate hypotheses from real-world data and qualitative research questions, write each hypothesis statement in a testable format and run a root cause analysis before prioritization. Specificity makes the output actionable: for a SaaS dashboard, an end-to-end flow beats a single screen. Generative, AI-driven tools can simulate users, but treat unvalidated AI-generated hypotheses as drafts.

Limits, usage rights and human oversight in AI-assisted design

Before using an image made with AI in a real product, check three things: usage rights, model bias and human oversight. Each tool’s terms define what you can do with what you generate, and they vary between platforms and between free and paid plans.

Models also carry biases from their training data and can repeat visual stereotypes. That is why human oversight is not optional: you decide whether a proposal is appropriate, inclusive and true to the brand. A good prompt speeds up generation, but it does not replace design judgment.

Frequently asked questions about AI product design prompts

These are the most common questions when writing AI product prompts, with the direct answer first.

How do I write a prompt for a product image?

Describe the product with material and finish, the background, the light, the angle, the aspect ratio and what you want to avoid. For example: “realistic photo of a steel bottle on a neutral gray background, soft studio light, close-up, 1:1, no logos”. If you have the real product, attach its photo and ask to keep shape, color and label.

What prompt works for product photos and catalogs?

One that fixes shot type, lighting and background, and that you repeat the same way across the catalog so the listings stay consistent. The product photos section has three ready prompts, including one that generates the listings and the image prompts at once.

How do I generate prompts for AI in product design?

Structure each prompt with role, context, task, constraints, style, output format and an avoid list. Start with the problem the product solves and iterate by changing one variable at a time. To build the foundation, training in AI for product design speeds up learning.

How do I generate product hypotheses with AI?

Start from the user problem, not the solution: tell the AI what is happening, to whom and with what data, and ask for several testable hypotheses with the assumption, the metric and the risk of each. Then validate each hypothesis with real users and iterate on the prototype before you build anything. The level 3 prompts in this library follow that order.

What are the types of AI prompts?

They fall into four levels by ambition: asking for a specific image, generating variations to compare, framing and testing a product hypothesis, and defining a coherent design system. The real value shows up from level 3.

What is the best AI for designing a product?

It depends on the phase. ChatGPT, Claude and Gemini stand out for briefing, hypotheses and prompt writing. Midjourney offers aesthetic quality in concepts, Stable Diffusion gives technical control and DALL·E generates quick images. The effective approach is to combine several in one workflow.

Can I use ChatGPT for product design?

Yes. ChatGPT is useful for the brief, for generating hypotheses and for creating concept images and mockups. For real prototypes, always validate materials and manufacturing with a technical profile.

Can I use Claude for product design?

Claude does not generate images. It helps write the prompt and work with long texts: interviews, feedback and briefs. Give it your idea and ask it to organize it by fields; then paste the prompt into ChatGPT, Gemini or Copilot.

Do Copilot or Gemini work for product design?

Yes. Copilot helps inside Word, Excel and Teams with your files and meetings; Gemini analyzes prototype images and works with Google Docs and Sheets. You have prompts for both in their sections.

When is a real photo better than AI?

When fidelity is critical: texture, exact color, precise scale or regulatory compliance in food and cosmetics. AI fits better in exploration and concepts, where speed matters more than exactness. A middle option is to photograph the real product and use AI only for the background.

Your next step with product design prompts

Pick one of the prompts in the library, change what is in brackets and run your first test. Then change a single field, such as the material or the light, and compare. Working one variable at a time teaches you what each field controls much faster than theory.

If you want to apply AI to real projects without losing professional judgment, see the Online Program in AI for Creatives at Founderz.