From your first prompt to a live app on your own domain — a practical, no-code-required starting point for building with Google’s AI platform in 2026.
What’s in this guide
What Is Google AI Studio (and What Isn’t It)
Getting Started — Your First Five Minutes
What You Can Actually Do with a Prompt
Build Mode — Describe an App, Get a Working App
API Keys — Your App’s ID Badge
Taking Your App Live with Google Cloud
Putting Your Own Domain on It
Keeping Costs Under Control
Understanding Gemini’s Models and Pricing
Lessons from the Field
Going Deeper — Code, CLI, and Developer Resources
What Is Google AI Studio
Google AI Studio is a free website — aistudio.google.com — where you can experiment with Google’s Gemini AI models, build working applications, and connect AI to your own projects. Everything happens in your browser. There’s nothing to install.
It helps to understand where AI Studio fits alongside Google’s other AI products:
- Gemini app (gemini.google.com) — Everyday AI assistant for chat, questions, and writing help. For everyone.
- AI Studio (aistudio.google.com) — Experiment, prototype, build apps, and create API keys. For builders and creators — that’s you.
- Vertex AI (Google Cloud) — Enterprise-scale AI deployment with compliance and security. For large organisations and dev teams.
💡 Key insight: AI Studio isn’t just a chat window. Since late 2025, it includes a Build mode that generates working applications from plain-English descriptions. You describe what you want, it writes the code and shows you a live preview — no programming required.
- Getting Started — Your First Five Minutes
There’s no waiting list, no payment, and no approval process. You just need a Google account.
Step 1: Open aistudio.google.com in your browser and sign in with your Google account.
Step 2: Accept the one-time Terms of Service pop-up for Generative AI.
Step 3: You’re in. You’ll see a sidebar on the left with Home, Playground, Build, and Dashboard. The main area shows quick-start cards.
What you’ll see
The Playground is your prompt testing area — type something, pick a model, and get a response. Think of it as the “try things” screen. Build is where you create full applications by describing them. The Dashboard shows your projects and API keys.
Try your first prompt
Click into the Playground. You’ll see a model selector at the top (start with Gemini 2.5 Flash — it’s fast and capable) and a text area to type your prompt.
Try something practical rather than generic. For example:
“I run a small landscaping business in Johannesburg. Write me a professional but friendly WhatsApp message template I can send to clients after completing a job, asking for a Google review. Keep it under 80 words.”
Click Run. That’s it — you’ve just used the Gemini API.
Now try changing the System Instruction box (above the chat area). This is where you tell the model who it should be and how it should behave. For example: “You are a marketing assistant for a landscaping company. Always write in a warm, professional South African English tone.” This is the foundation of building AI systems rather than just asking one-off questions.
- What You Can Actually Do with a Prompt
Most beginners underestimate what a single well-written prompt can do. Here’s what’s available to you right now in AI Studio, no coding required.
Write and transform text
Summarise long documents. Translate between 100+ languages. Rewrite content in a different tone. Generate marketing copy, emails, social media posts, or reports. The key is being specific — tell the model the exact format, tone, length, and audience you need.
Work with images, PDFs, audio, and video
Gemini is multimodal — you can upload a photo and ask it to describe what it sees, upload a PDF and ask it to extract key data, or even upload a video and ask for a summary. Use the attachment button in the Playground to upload files. The model can handle up to about 1 million tokens of context, which means you can work with very large documents.
Get structured data out
In the Playground settings, you can force the model to respond in JSON format. This is important when you start building applications — your app can parse structured data much more reliably than free-form text.
Search the live web
Gemini can “ground” its responses with Google Search — meaning it looks up current information before answering. This solves the problem of AI models giving outdated answers. You can enable this in the Playground settings under “Grounding.”
🔗 The bridge to building: Once you’ve written a prompt that works well, click the “Get Code” button in the top-right of the Playground. This exports your exact prompt as ready-to-use code (Python, Node.js, or a web request). You don’t need to understand code yet — but when you’re ready, this is how you move from the browser into your own applications.
- Build Mode — Describe an App, Get a Working App
This is the feature that makes AI Studio accessible to non-developers. You describe what you want in everyday language, and Gemini writes the code and shows you a live, working preview — in real time.
Step 1: Click Build in the left sidebar.
Step 2: Type a description of what you want to build. Be specific — mention what the app does, who uses it, and what it should look like.
Step 3: Watch the magic. Code files appear on the left, and a live preview appears on the right. You can interact with it immediately.
Step 4: Refine by chatting. Say things like “Make the buttons bigger,” “Add a dark mode,” or “Include a download button.” Each change updates the preview instantly.
Step 5: When you’re happy, export as a ZIP file or push to GitHub. You can also deploy directly to hosting from here.
What kind of things can you build?
People are building calculators, dashboards, data visualisers, landing pages, games, quoting tools, comparison apps, project trackers, and more. If you can describe it clearly, Build Mode can generate a working first version in minutes.
Here’s an example prompt that creates something genuinely useful:
“Build a simple tool where I can paste two lists of numbers (Week A and Week B), and it shows me which week had better averages plus a bar chart comparing them. Make it clean and mobile-friendly.”
Two minutes later you have a working comparison tool with data entry fields, automatic calculations, and a visual chart.
⚠️ Be honest about the limits: Build Mode creates excellent prototypes and internal tools. But a prototype is not the same as a production app with user logins, payment processing, and security. Think of Build Mode as the fastest way to validate an idea and create a working demo — then decide whether to invest in developing it further.
- API Keys — Your App’s ID Badge
When you want to use Gemini outside of AI Studio — in your own app, website, or automation tool — you need an API key. This is a unique string of characters that identifies your project to Google. Think of it as your app’s ID badge: it tells Google who’s making the request and tracks your usage.
How to create one (all in the browser)
Step 1: In AI Studio, look for “Get API key” in the bottom of the left sidebar. Click it.
Step 2: Click “Create API key.” If you’re a first-time user, Google creates a default project for you automatically. Otherwise, select or create a project.
Step 3: Your key appears on screen — a long string of letters and numbers. Copy it immediately and save it somewhere secure (a password manager is ideal).
🔴 Treat your API key like a password. Never paste it into a public document, a social media post, or a GitHub repository. Anyone who has your key can make API calls billed to your account. If you think a key has been exposed, delete it from the API Keys page and create a new one.
Things worth knowing
API keys don’t expire on their own — they stay active until you delete them. Each key is linked to a Google Cloud project, and rate limits (how many requests per minute you’re allowed) are applied per project, not per key. You can manage all your keys from the Dashboard → API Keys page in AI Studio.
- Taking Your App Live with Google Cloud
So you’ve built something in Build Mode and you want other people to use it. The simplest path is Google Cloud Run — a service that hosts your app on the internet and gives you a public URL. You only pay for the time someone is actually using your app, and there’s a generous free tier.
The web UI path (no terminal needed)
Step 1: Set up a Google Cloud account at console.cloud.google.com. New accounts get $300 in free credits valid for 90 days — plenty to experiment with.
Step 2: Create or select a project. Use the project dropdown at the top of the Cloud Console. If AI Studio already created a project for your API key, you can use that same one.
Step 3: Navigate to Cloud Run. Click the hamburger menu (☰) at the top-left, find Cloud Run under “Serverless” — or just type “Cloud Run” in the search bar at the top.
Step 4: Click “Create Service.” If you exported your Build Mode app to GitHub, you can connect your repo here and Cloud Run will build and deploy it automatically.
Step 5: Configure your settings. Choose a region close to your users. Under “Authentication,” select “Allow unauthenticated invocations” if you want a public website. Set Maximum instances to 1 for personal projects to keep costs down.
Step 6: Add your API key securely. Under “Variables & Secrets,” add an environment variable called GEMINI_API_KEY with your key as the value. This keeps it out of your code.
Step 7: Click “Create” and wait a few minutes. You’ll get a URL like https://my-app-abc123.a.run.app — that’s your app, live on the internet.
💰 Free tier: Cloud Run includes generous monthly free usage — 2 million requests, 180,000 vCPU-seconds, and 360,000 GiB-seconds of memory. For a personal project or small internal tool with light traffic, you may not pay anything beyond your domain registration.
- Putting Your Own Domain on It
The auto-generated .run.app URL works but doesn’t look professional. Here’s how to connect your own domain, all through the web interface.
Step 1: In the Cloud Console, open your Cloud Run service and look for the “Custom Domains” tab or “Manage Custom Domains.”
Step 2: Click “Add Mapping” and enter your domain name. Google will ask you to verify ownership — this usually means adding a TXT record at your domain registrar (GoDaddy, Namecheap, Cloudflare, etc.).
Step 3: Google shows you DNS records to add — typically four A records and four AAAA records for a root domain, or a CNAME for a www subdomain. Copy these exactly into your registrar’s DNS settings.
Step 4: Wait for SSL. Google automatically creates a free HTTPS certificate. This takes 30 minutes to a couple of hours. Browser warnings during this time are normal — just wait it out.
Once done, your app is accessible at your own domain with HTTPS. The domain mapping itself is free — you only pay for the domain registration (typically $10–15/year).
⚠️ Regional limitation: Direct domain mapping is only available in certain Cloud Run regions (us-central1, us-east1, europe-west1, asia-northeast1 and a few others). If your service is in a different region, you’ll need a Load Balancer approach — see the official docs.
- Keeping Costs Under Control
This is the section beginners skip and then regret. The tools are straightforward — but you need to know one critical fact upfront:
🔴 Setting a budget does NOT automatically stop your spending. Budgets send email alerts when thresholds are reached. They don’t turn off your services. If you need an actual hard stop, you need to set that up separately (explained below).
Setting up budget alerts (do this first — it takes 2 minutes)
Step 1: In the Cloud Console, go to Billing (left nav menu or search for it). Click your billing account.
Step 2: Click “Budgets & alerts” in the left sidebar, then “Create Budget.”
Step 3: Name it clearly (e.g., “Gemini App — Monthly”). Set scope to your project. Set a monthly amount — start low, like $10 or $20.
Step 4: Set alert thresholds — defaults are 50%, 90%, 100%. Add 25% for early warning and 150% for “something’s wrong.” Click Finish.
Five practical tips for near-zero costs
Start on the free tier. The Gemini API free tier requires no billing. Only enable billing when you need higher limits or paid-only models.
Limit Cloud Run to 1 instance. On your service page, under “Edit & Deploy New Revision,” set the Maximum Instances slider to 1. This caps your compute costs.
Check Billing → Reports weekly. This dashboard shows exactly where money is going, by service, project, and date.
Set your budget below your actual ceiling. Cost reporting can lag up to 24 hours. If your real limit is $50, set alerts at $30.
Use the Pricing Calculator to estimate costs before you launch anything.
If you want a true hard stop
Google documents a method to automatically detach a project from billing if costs exceed your limit (using Pub/Sub and a Cloud Function). This is more advanced but worth knowing about for peace of mind. See the programmatic budget notifications guide. Be aware: detaching billing stops all services and may cause data loss — treat it as a last-resort safety net.
- Understanding Gemini’s Models and Pricing
Gemini isn’t one model — it’s a family. Picking the right one matters for both quality and cost.
Gemini 3.1 Pro — Most capable. Complex reasoning, advanced coding. Slower. Paid only, no free tier.
Gemini 3 Flash — Great balance of power and speed. Available as a free preview.
Gemini 2.5 Pro — Strong reasoning and analytical work. Medium speed. Free tier at 5 requests/min, 100 requests/day.
Gemini 2.5 Flash ⭐ (start here) — General-purpose workhorse. Fast. Free tier at 10 requests/min, 250 requests/day.
Gemini 2.5 Flash-Lite — High-volume simple tasks like sorting and extraction. Fastest. Free tier at 15 requests/min, 1,000 requests/day.
What does it cost?
AI Studio is free. The Playground, Build Mode, and all the tools — no charge.
The Gemini API free tier requires no credit card. You get 5–15 requests per minute and 100–1,000 requests per day depending on the model. Enough for development, testing, and light personal use. Note: free tier data may be used by Google to improve their models.
Paid pricing starts at ~$0.10 per million input tokens (Flash-Lite) up to ~$2.00 (3.1 Pro). A million tokens is roughly 750,000 words. A small app handling a few hundred requests per day on a Flash model typically costs single-digit dollars per month.
💡 Recommendation: Use Gemini 2.5 Flash for everything at first. It handles most tasks well, it’s fast, and it has a usable free tier. Only switch to a Pro model when you can point to a specific quality gap.
- Lessons from the Field
These come from building AI systems and training others to do the same — the things the official docs don’t cover.
Systems beat one-off prompts
A question gets you an answer. A system — with a system instruction, a chosen model, structured output, and clear constraints — gives you repeatable, reliable results every time. This is the difference between using AI and operating AI.
Spend time in the Playground first
The Playground is where you learn what the model does well, where it needs guardrails, and what breaks. That understanding is what makes your Build Mode apps and system instructions effective. Don’t skip this step.
Your system instruction is the real product
Be explicit: tell the model its role, what format to use, what to do when uncertain, what to never do, and who the audience is. Test with edge cases. This single piece of text is usually the difference between a mediocre tool and a great one.
Free tier limits change without warning
In December 2025, Google reduced free tier quotas by 50–80% overnight. Use the free tier for learning and development. Enable billing for anything clients or customers rely on.
The model matters less than your prompt
A well-structured system instruction on 2.5 Flash outperforms a vague prompt on 3.1 Pro for most real-world tasks. Improve your prompt before upgrading your model.
- Going Deeper — Code, CLI, and Developer Resources
When you’re ready to move beyond the web interface, these resources pick up where this guide leaves off.
Using the API from code: The “Get Code” button in the Playground exports your prompt as Python, Node.js, or cURL. Google’s Python SDK is zero-config — store your key as an environment variable called GEMINI_API_KEY and it’s picked up automatically. Full quickstart guides: ai.google.dev/gemini-api/docs/quickstart
Official resources
- Google AI Studio — The platform itself
- Gemini API Quickstart — Setup guides for Python, Node.js, Go, and more
- Gemini API Docs — Full reference: capabilities, models, features
- Pricing & Rate Limits — Free tier details, paid rates, quotas by model
- API Key Management — Creating, securing, and restricting keys
- Build Mode Documentation — Building apps with natural language
- Gemini Cookbook (GitHub) — Hands-on tutorials and code examples
- Cloud Run Domain Mapping — Connecting your own domain
- Cloud Billing Budgets — Budget alerts and cost management
- GCP Pricing Calculator — Estimate costs before committing
The path forward
The journey goes: explore the Playground → craft system instructions → build with Build Mode → create an API key → deploy to Cloud Run → connect your domain → set budget alerts. Each step builds on the last.
The technology is genuinely accessible now. A non-developer with a clear idea can have a working prototype live on the internet within an afternoon. The gap to fill isn’t technical knowledge — it’s structured thinking: knowing what you want to build, how to describe it precisely, and where the boundaries are between a prototype and a production system.
Imbila.AI — AI Training, Consulting & Workflow Implementation
This is a living document. Google’s platform evolves rapidly — verify details against official documentation when building production systems. Last updated March 2026.