How To Write AI Prompts To Build an App: A Beginner’s Guide

Vibe coding is a way to develop apps without any technical experience. Instead of writing code in a programming language (like Python or JavaScript), you prompt an AI tool to create an app that fits your requirements, by describing the look and functionality you want in natural language.

Most vibe coders use AI app builders like Base44, Lovable, and Hostinger Horizons. These platforms enable you to use prompts to build the entirety of your app. For example, you can:

• Control your app’s design
• Fix any bugs that you find
• Add app features like payment processing to take payments online

The Benefits of Vibe Coding

Is vibe coding worth your time? Absolutely, since knowing how to write AI prompts to build an app can improve your business’ efficiency over time. Here’s why:

• You’ll cut costs: Building an app with a development team is an expensive feat, with a small app potentially costing you $5,000 to $50,000. Meanwhile, AI app builders can cost as little as $6.99 per month (billed annually).
• You’ll save time: If you’re an entrepreneur and your goal is to create a minimum viable product (MVP), vibe coding is the way to go. You could create your first app prototype in an afternoon.
• You’re in the driver’s seat: Because you’re directing the AI yourself, you have much more control over your app and what you’d like it to feature.
• You’re not intimidated anymore: Even if you’re a complete beginner who doesn’t know how to code, vibe coding opens a new path for you. You could create your own internal tool or prototype without relying on a developer’s help at first.

Knowing how to prompt well is key to AI app building for beginners, so let me show you how to construct a prompt that works. Clear communication is extremely important, since your AI app builder won’t be able to create the app you’ve envisaged if it has vague, meaningless prompts to follow.

Here’s how I talk to the AI for best results.

Give the AI Enough Context To Work With

Unfortunately, AI isn’t a mind reader – and I’ve learned this the hard way. If you don’t give the AI enough context about your project’s purpose, audience, and features, it won’t be able to produce an app that’s tailored to your business needs.

To show you what I mean, I generated two versions of an inventory tracker app on Base44, using one vague prompt and one detailed prompt.

Prompt 1:

“Create an app to track store inventory.”

Prompt 2:

“Create an inventory tracker app for my resin jewelry business. Create separate categories for rings, bracelets, and necklaces, and let me mark items when they’re out of stock. The app should have a user-friendly interface that my team can easily pick up.”

Notice the difference? The second prompt provided me with better results, since I had specified:

• The app’s purpose: To track my store’s inventory, including rings, bracelets, and necklaces
• Features I wanted: Inventory categories and “out of stock” markers
• Who it’s for (the target audience): For the team to use

So, as you write your first prompt, make sure that you include these key details for best results.

Clearly Describe Your App’s Look and Feel

When you’re adjusting your app’s design, it’s important to include rich, descriptive words that define the way you want it to look and feel. For inspiration, take a look at my own example – I used Bolt to create a landing page for a graphic designer’s business.

Prompt:

“Create a vibrant landing page for a graphic designer in a “pop art” style. Use clashing, high-saturation colors (e.g. neon blue against deep purple). Use thick, black outlines around buttons and images, too.”

My prompt worked because I included descriptive words – like “pop art” and “high-saturation” – so the AI had a clear idea of which design elements to include.

If you’re not sure which words to include in your prompt to generate a design that reflects your brand identity, here are a few examples of terms you could use:

Attach Reference Information and Examples for Better Results

You can also attach data or visual references to not just tell, but show the AI what you’d like to achieve.

For example, when I tested Replit, using it to create an interactive website quiz for a skincare brand, I attached a list of products for the AI to include. The AI used my product list as the building blocks for the quiz, creating questions that would lead customers to the right recommendations based on their answers.

Other references you could attach include:

• Your brand’s style guide
• A wireframe depicting your preferred app layout
• Screenshots of your website and social media profiles for brand consistency
• Past brand copy to help the AI match your tone of voice

However, not all AI app builders let you upload attachments. I could upload reference images to Base44 and Lovable, but not Hostinger Horizons.

Break Large Prompts Down Into Smaller Prompts

Don’t overload the AI with a long list of commands in one prompt. If you’re asking it to do too much at once, it may struggle to follow through with every action – leading to a messy, unfinished app!

For better clarity, create a new prompt for each new action you’d like the AI to take, be it a design change, new feature, or bug fix.

For example, when I used Lovable to create an order tracking tool for a florist, I used a separate prompt for each key stage of building my app:

Initial app creation:

“Create an internal productivity app named “BloomBoard” for my floristry business in a playful, pastel color scheme. It should have a database in the back end where my team can track orders, and they must be able to label each order depending on its progress.”

Design change:

“The design is washed out and lacks personality. Keep pastel colors for the backgrounds and buttons only. Change all text to a deep, dark charcoal grey for readability.”

New feature:

“For each order card, create a mood board section that allows app users to upload multiple reference images. Use a Pinterest-style image grid layout and enable users to expand each image when they click on it.”

If you want to write the best AI prompts for app development, you also need to pay attention to the biggest mistakes to avoid. I’ve listed them below:

1. Sharing private information – Be mindful of what information you’re giving the AI. AI tools often use your data to improve their models, so research the platform you’re using first and avoid sharing sensitive information. You don’t want to risk your personal or business information, or your customers’ data, from being used or stored without your knowledge, so be sure to check the tool’s privacy policy.
2. Not giving the AI feedback – AI tools are constantly improving, and your feedback is key. If you don’t like the design the AI has generated, tell it, but be clear about what exactly you’d like it to do differently.
3. Treating the AI as gospel – AI isn’t a perfect tool. It can get facts wrong, so make sure that you’re consistently verifying the information it generates for your app. If you’re confused by something it has generated, ask it to explain its reasoning, or say where it retrieved the data from.
4. Forgetting to mention what you don’t want – If there’s something you don’t want to see on your app, tell the AI. For example, if you don’t want your app’s copy to be overlong, you could write, “summarize each product description in bullet points, with less than 20 words per bullet point.”
5. Adding too many app features at once – When prompting, you may get carried away and want to add 10 different features to your app. However, don’t overwhelm the AI. Start by focusing on one to three key features instead.
6. Attaching low-quality data to your prompts – Assess the quality of your data before you attach it to your prompts. For best results, your data should be clean, up to date, and free from errors. The AI will use this data to create your app, so the more accurate it is, the better.
7. Not understanding the AI’s limitations – If you tell the AI to “create an app like Duolingo,” you won’t get very far. Though AI app builders are impressive, they still don’t have the capability to create highly complex apps. Be realistic when coming up with your app idea and writing your prompts.

Next, I’ll show you how I’ve put these prompting tips and tricks into practice. On Hostinger Horizons, I started prompt engineering for app creation by creating my initial blueprint prompt first. This was to create an app for a hypothetical restaurant business that would help the team respond to Google reviews.

1. Setting the Blueprint

For my first prompt, I clearly outlined what kind of app I wanted. I included important contextual information such as the app’s purpose and who it’s intended for:

“Build an app for my small restaurant business that generates responses to our Google reviews. It should help my team respond to positive, neutral, and negative responses, so I’d like a generator form to input reviews and create a suitable response.”

2. Changing the App’s Look

I then changed my app’s design to make it look more welcoming to a restaurant owner and their team. For this step, I made sure to include specific design elements in my prompt, such as its overall style, font, and color scheme:

“Switch my app’s design from dark mode to a light theme. I’d like an approachable, minimalist aesthetic with a white background. Update the typography to “Helvetica” and change the “generate response” button color to a welcoming blue-gray (#7393B3).”

3. Adding a New Feature

Once I was happy with my design, I decided to add a new feature to my app. I told the AI to build an “escalation flagging” tool that could help the restaurant to identify high-risk, negative reviews requiring an urgent response. I specified what the tool should do and how it should look in my next prompt. 

“Add a new “escalation flagging” feature to my app that identifies high-risk keywords like “poisoning” or “injury.” When a word is found, don’t generate a response, but add a “NEEDS HUMAN REVIEW” label under “Generated Response” in bold, red text (#FF0000).”

4. Fixing an Issue

During testing, you’ll likely spot aspects of your app that don’t work so well. So it’s important to know how to construct an effective prompt to fix errors.

When testing my app, I noticed that there was a star rating feature that wasn’t interactive and didn’t let me mark what rating a reviewer had given the restaurant. I used a simple prompt to fix this and create a dropdown menu instead.

“Update the “star rating” component so it’s an interactive element, not a static design feature. Replace it with a star rating dropdown menu (1-5 stars). The star rating must influence the response’s final outcome.”

Depending on the complexity of your app and your needs, you’ll likely need to continue prompting until you’re happy with the final result. Additionally, it’s important to test your app between every prompt. This is to ensure that your app is running smoothly, its features work as intended, and it provides a smooth user journey for its intended audience.

With AI app builders, you use up credits whenever you submit a new prompt. Some prompts also cost more credits than others, which can be annoying if you have a limited number of credits to work with per day.

For example, when I tested Lovable’s free plan, I only had five credits per day to play with. I burned through these credits fast, and on day one, I was only able to:

• Generate my initial app
• Enable cloud storage for my data
• Make one design change
• Add a new feature
• Fix one error

After that, I had to wait until the following day to make more changes.

To conserve your credits, check how many you have at your disposal before you start. The amount of credits you have to spend will typically increase on paid plans, so it may be worthwhile investing in a more expensive plan if you’re going to be using your AI app builder frequently. As an example, Lovable offers 30 monthly credits on its free plan, raising to 100 monthly credits on the Pro plan.

With this guide under your belt, you now know how to write AI prompts to build an app for your business. However, you won’t become an expert overnight, so it’s important to get testing now and get used to the different prompting tips and techniques I’ve listed. Once you’re more familiar with prompting, you could even save your own prompt templates for future projects.

By venturing into the world of small business app development with AI, you’ll feel empowered to build custom tools that will save your company precious time, money, and resources in the process.