AI-200 Practice Tests: Azure AI Cloud Developer Associate
250 questions on containers, vector data in Cosmos DB, PostgreSQL and Redis, messaging, and secure monitoring
Are you ready for AI-200? Reading the study guide tells you what is on the exam. It does not tell you whether you can pick the right answer when two options both look plausible and the clock is running. That is what these practice tests are for.
This practice set contains 250 original questions in five timed tests, written against the official AI-200 study guide (last updated October 7, 2026). Every test follows the published domain weights as closely as it can: containerized solutions, AI solutions on Azure data management services, connecting to Azure services, and securing and monitoring the result.
AI-200 replaced AZ-204, which retired on 31 July 2026. The new exam dropped API Management, Blob Storage, Microsoft Graph, MSAL, Event Hubs and Container Instances. It added vector search in Cosmos DB for NoSQL, PostgreSQL with pgvector and Azure Managed Redis, plus AKS manifests, KEDA scaling in Container Apps, OpenTelemetry tracing and KQL. Data services are now the largest domain. If you practise with old AZ-204 questions, you are preparing for an exam that no longer exists.
The five tests build up in difficulty: Test 1, Groundwork: one concept per question, your baseline. Test 2, In Practice: short scenarios that separate features people confuse. Test 3, Trade-offs: two answers look right, one constraint decides. Test 4, Under the Hood: CLI commands, SDK calls, settings and limits. Test 5, Exam Day: long multi-constraint scenarios under exam pressure.
Every answer option has its own feedback, so you learn why a wrong choice is wrong, not just which one is right. Each question also has an overall explanation that ends with a link to the Microsoft Learn page it is based on.
After each attempt PrepKiwi shows your score per domain, so you know exactly where to study next. Tests 1 to 4 require 80% to pass and test 5 requires 85%. That is deliberately stricter than the real AI-200 exam, which reports a scaled score of 700 out of 1000; that is not the same as 70% of the questions.
What this practice set is not: it is not a video course, it does not contain real exam questions, and it does not replace hands-on practice. Use it next to the free Microsoft Learn training for AI-200 to find your gaps and close them before exam day.
Groundwork: Core Services and Key Concepts
50 questions · 80 minutes · pass mark 80%
Start here, cold, before you study anything else. This first test is your baseline. It covers all four AI-200 skill areas with direct, one-concept questions: what Azure Container Registry Tasks do, how Container Apps revisions work, which Cosmos DB and PostgreSQL features store and search embeddings, how Service Bus and Event Grid differ, and which Key Vault role actually reads a secret. 50 questions, 6 of them multi-select, 80 minutes. The pass mark is 80%, higher than you strictly need on the real exam, because practising at home is easier than a proctored session. Read every explanation afterwards, including the ones you got right, and note your weakest domain in the results.
In Practice: Scenarios You Will Meet on the Job
50 questions · 80 minutes · pass mark 80%
Short, realistic scenarios in the style of the real exam. Each question puts you in a developer's seat and asks you to pick between features that are easy to mix up: a base image trigger or a commit trigger, a ConfigMap or a Secret, a correlation filter or a SQL filter, a Key Vault reference or an App Configuration reference. All four domains are weighted exactly as in the official study guide. 50 questions, 7 of them multi-select, 80 minutes, pass mark 80%. Every answer option has its own feedback, so you see why the wrong choices are wrong.
Trade-offs: When Two Answers Look Right
50 questions · 80 minutes · pass mark 80%
In this test two answers usually look right, and one stated constraint decides between them: no secrets in configuration, scale to zero, private network only, ordered processing, lowest cost. This is where most candidates lose points on the real exam, so read every stem to the end. You will choose between Container Apps and App Service, HNSW and IVFFlat, Service Bus and Event Grid, Premium and Flex Consumption. 50 questions, 7 of them multi-select, 80 minutes, pass mark 80%.
Under the Hood: Parameters, CLI and SDK Details
50 questions · 65 minutes · pass mark 80%
The implementation details: ACR Tasks YAML, az containerapp commands, scale rule authentication, workload identity labels, Cosmos DB indexing paths, pgvector settings such as ivfflat.probes and hnsw.ef_search, Redis FT.CREATE and KNN syntax, Service Bus rules, host.json settings, app setting names for identity-based connections, and KQL joins. Questions are shorter, the clock is tighter: 50 questions, 8 of them multi-select, 65 minutes, pass mark 80%. If you pass this one, you know the platform, not just the product names.
Exam Day: Full-Length Simulation
50 questions · 80 minutes · pass mark 85%
Save this test until you pass tests 1 to 4. It is the readiness gate: long company scenarios with several constraints at once, domains mixed inside one scenario, and the tightest time per question in this practice set. Take it in one uninterrupted sitting. 50 questions, 8 of them multi-select, 80 minutes. The pass mark is 85%, deliberately above what the real exam asks: the real exam uses a scaled score of 700 out of 1000, which is not a percentage, and a practice test at home is easier than a proctored exam. If you clear 85% here on a first attempt, book your exam.
What you will practice
- Answer AI-200 questions across all four skill areas: containers, Azure data services for AI, messaging and Functions, security and monitoring
- Choose between easily confused options, such as Container Apps vs App Service, HNSW vs IVFFlat, Service Bus vs Event Grid
- Store and query embeddings with Cosmos DB for NoSQL, PostgreSQL with pgvector and Azure Managed Redis vector indexes
- Secure a back end with managed identities, Key Vault references and App Configuration instead of stored secrets
- Learn from per-answer feedback and an explanation with a Microsoft Learn link for every one of the 250 questions
- Build up from core concepts to a full exam-day simulation, and find your weakest domain with PrepKiwi's domain report
- Read KQL queries over Application Insights data and trace requests across services with OpenTelemetry
- Recognise the CLI flags, SDK calls and settings that the exam uses as distractors
Exam domains covered
Who is this for
This practice set is for developers preparing for Microsoft exam AI-200: Developing AI Cloud Solutions on Azure, which leads to the Microsoft Certified: Azure AI Cloud Developer Associate certification. It suits back-end and cloud developers, DevOps and platform engineers who build the services behind AI applications, students and career changers moving into Azure development, and technical leads who need to review AI back-end designs. If you studied for AZ-204 (Developing Solutions for Microsoft Azure), note that AZ-204 was retired on 31 July 2026. AI-200 is its successor but not a rename: API Management, Blob Storage, Microsoft Graph, MSAL, Event Hubs and Container Instances are gone, while vector search in Cosmos DB, PostgreSQL and Redis, AKS, KEDA, OpenTelemetry and KQL are new. Old AZ-204 practice material prepares you for the wrong exam. This practice set is a set of practice tests with detailed explanations. It checks and deepens what you know; it is not a video course and it does not replace hands-on practice.
Requirements
- Basic Python and a general idea of REST APIs and SDKs. The AI-200 exam assumes Python; you do not need to be an expert.
- Some hands-on time in the Azure portal or Azure CLI helps, but beginners can start: every answer is explained.
- These are practice tests, meant to be used next to the free Microsoft Learn course AI-200T00, not instead of it.
Free exam guide
Not sure what to expect? Read how the exam is built, what it covers and how to prepare.