Build Your First Google AI Studio App: Brief and Review
Google AI Studio can build a web app prototype from a description without requiring users to write every part themselves. Calculations, data storage, and errors still need checking before sharing it. This article uses a small task with independently calculable answers to explain how to specify and review the work.

Set acceptance criteria before generating a prototype
Google AI Studio Build mode lets users describe an app in natural language, then generates code and a working preview. Users can request revisions in chat or open the Code tab to inspect and edit code. Current documentation says web apps default to React with a Node.js server runtime available, but this article's example asks for recording and totaling logic to run in the browser without further API calls. Read Google's Build mode documentation
You can start prototyping without writing every line yourself. App owners still must define rules, check anomalies, and take responsibility for security. Money or personal-data apps require stricter review than experimental examples.
Start with a small task you can verify yourself
Hypothetical example: build an income-and-expense tracker in baht without bank connections, a database, or additional AI services. Open Google AI Studio then select Build mode and enter this brief:
สร้างเว็บแอปบันทึกรายรับรายจ่ายภาษาไทยสำหรับทดลองใช้
- มีช่องวันที่ รายการ ประเภท (รายรับ/รายจ่าย) และจำนวนเงินเป็นบาท
- แสดงยอดรายรับรวม รายจ่ายรวม และคงเหลือ
- รับจำนวนเงินมากกว่า 0 และทศนิยมไม่เกิน 2 ตำแหน่ง
- คำนวณเป็นจำนวนเต็มหน่วยสตางค์ แล้วแสดงผลเป็นบาททศนิยม 2 ตำแหน่ง
- ปฏิเสธช่องว่าง ข้อความ จำนวนศูนย์ ค่าติดลบ และทศนิยมเกิน 2 ตำแหน่ง
- ใช้ข้อมูลตัวอย่างสมมติ ห้ามเชื่อมบัญชีหรือขอข้อมูลส่วนตัว
- ให้ข้อมูลรายการและการรวมยอดอยู่ในเบราว์เซอร์ของผู้ทดลอง
- ยังไม่ต้องมีระบบสมาชิก ฐานข้อมูล หรือการส่งรายการไปบริการภายนอก
- เพิ่มปุ่มล้างข้อมูลทดลองพร้อมถามยืนยัน
- อธิบายว่าเมื่อรีเฟรชหน้า ข้อมูลจะคงอยู่หรือหายไปตามโค้ดจริง
State exclusions and conditions early because “money-tracking app” does not specify storage, access, or what negative values mean. Leaving these choices to the system may produce unwanted behavior.
Compare the app’s output with independently calculated answers
When the prototype is ready, enter the hypothetical inputs below and compare the results with your own calculations. These are proposed tests for readers, not results from tests conducted by KrobClick. A series of tests offered to readers not results from tests conducted by KrobClick
| What you enter | The result you should get | Review points |
|---|---|---|
| Income 1,200 baht and 350 baht | Total income 1,550 baht | Addition and number formats |
| Add an expense of 500 baht | Remaining 1,050 baht | Subtraction and balance |
| Enter the amount as text. | Report an error; do not add an item | Checking data types |
| Enter the amount −10 baht | Rejects inputs under defined rules | Testing for negative values |
| Start a new set: income of 0.10 and 0.20 baht | Total income 0.30 baht | Handling satang and decimal display |
| Blank, 0, or 1.005 baht | Report an error; do not add an item | Values outside the rules |
| Press to clear data. | Ask for confirmation before deleting | Protection against accidental deletion |
The first two cases use consecutive data; start with an empty list for the decimal case. A negative balance caused by expenses exceeding income should be displayable even though each entry must exceed 0. These rules belong to this example; they are not ready-made Google requirements. If an app accepts decimal currency amounts, check its actual calculations, rather than merely formatting results with .00.
Also check what happens after refreshing, closing the tab, or opening another browser. Data held only in memory may disappear on refresh; browser storage may retain it on the device. Do not claim there is a backup until you have checked the storage system.
Fix one issue at a time and recheck
- Identify symptoms: Tell the entered information, the results, and the expected results, such as “enter 1,200 and 350, but the total is not 1,550.”
- Request a targeted edit: Please fix the totaling logic without changing storage format or adding external services
- Review the code: Open Code to check API calls and data storage, particularly secrets and user data
- Run the original test: Solving one problem may break another. Therefore, every case in the table should be tried again.
Google documentation says code can be edited in AI Studio, downloaded as ZIP, or synced with GitHub for further development. Automatic generation and editing do not certify code as bug-free. Learn how to edit and export your project.
Check data, costs, and access before sharing a prototype
Use only hypothetical data in this example app, not real account details, passwords, or personal documents. Google says Build mode stores keys server-side by default for new apps, but developers must still inspect code and Secret settings if adding Gemini API later. Read the guidance on managing secrets
Sharing a prototype and publishing a public service require different checks. Google says people with shared access may view code and copy the app. Before Cloud Run deployment, check account eligibility and deployment options. The documentation distinguishes Starter Tier for eligible accounts from Standard deployment, which requires billing. Do not assume every account needs a card or receives the same allowance. See deployment conditions To estimate cost per user, read Google AI Studio cost breakdown article of KrobClick which covers costs rather than prototype validation.
“Do not send entries outside the browser” must be verified in the resulting app; it is not AI Studio's data-policy guarantee. Your prompts still go to Google. If you cannot review code, keep using hypothetical data and ask a qualified reviewer to check network calls and permissions before real data. For broader acceptance principles, read How to check an AI agent's work which is broader than this sample app.
A good first app has clear passing and failing criteria
Before adding features, state what the app does, what it excludes, and what data it uses. Prepare at least three test cases with answers you can calculate independently. Fix mismatches before sharing. This gives AI-assisted development clear acceptance criteria.
Sources
- Google AI for Developers: Build apps in Google AI Studio
- Google AI for Developers: Develop full-stack apps in Google AI Studio
- Google AI for Developers: Deploying from Google AI Studio
- Google AI for Developers
Reviewed Google documentation on October 15, 2026, Thailand time. This article is based on documentation; we did not build or test an app on a real account. Prompts and expected results are hypothetical KrobClick examples, not product screens or experimental results. Features and account access may change.