Free Azure AI-900 Practice Questions
Original practice questions with a full explanation for every answer — see why the right choice is right and why each wrong one is wrong.
⏱️ 12 questions
Click an answer to reveal the reasoning. These are original questions written to test the same concepts as the exam — not copied exam items — so you learn the <em>why</em>.
In short: These are free, original Azure AI Fundamentals (AI-900) practice questions covering all five skill areas. Each question includes a full explanation of the correct answer and why the other options are wrong. No signup is required.
Azure AI-900 practice set
Twelve original questions across the five skill areas. Tap an option to check it, or use “Show all answers” to review the reasoning.
Area 1 · AI workloads and responsible AI
1. A hiring tool is checked to make sure it does not disadvantage applicants based on gender or ethnicity. Which responsible-AI principle does this address?
Correct answer: Fairness.
Ensuring a system treats all groups equitably and does not create unfair bias is the principle of fairness. Transparency is about being understandable, reliability and safety is about consistent safe performance, and inclusiveness is about serving people of all abilities.
Area 1 · AI workloads and responsible AI
2. Reading the text printed on a photographed receipt is an example of which AI workload?
Correct answer: Computer vision (OCR).
Extracting text from an image is optical character recognition (OCR), a computer-vision workload. NLP works on text that is already digital, while regression and clustering are machine-learning task types, not image reading.
Area 1 · AI workloads and responsible AI
3. Which responsible-AI principle is most directly supported by clearly telling users how an AI system reached a decision?
Correct answer: Transparency.
Making a system understandable and explaining how it works supports transparency. Privacy and security protects data, fairness is about equitable treatment, and accountability is about people being answerable for the system's behaviour.
Area 2 · Machine learning fundamentals
4. A model is trained to predict the exact selling price of a house from its size and location. Which type of machine learning is this?
Correct answer: Regression.
Predicting a continuous numeric value such as a price is regression. Classification predicts a category, clustering groups unlabeled data, and reinforcement learning trains an agent through rewards — none predicts a specific number here.
Area 2 · Machine learning fundamentals
5. In a machine-learning dataset, the columns used as inputs to make a prediction are called the:
Correct answer: Features.
The input columns a model learns from are the features; the value it predicts is the label. Clusters are groups found in unlabeled data, and hyperparameters are settings that control training, not the input data itself.
Area 2 · Machine learning fundamentals
6. A retailer wants to group customers into similar segments without any predefined categories. Which machine-learning approach fits BEST?
Correct answer: Clustering.
Grouping unlabeled data into similar segments with no predefined classes is clustering (unsupervised learning). Regression predicts numbers and classification predicts known categories — both need labels — and object detection is a computer-vision task.
Area 3 · Computer vision
7. An application needs to locate every car in a photo and draw a box around each one. Which computer-vision capability does this require?
Correct answer: Object detection.
Finding multiple items and returning their locations (bounding boxes) is object detection. Image classification only labels the whole image, OCR reads text, and sentiment analysis is an NLP task — none locates individual objects.
Area 3 · Computer vision
8. Which Azure service would you use to analyse images and read printed or handwritten text from them?
Correct answer: Azure AI Vision.
Azure AI Vision provides image analysis and OCR. Azure AI Language works on text, Azure AI Speech handles audio, and the Machine Learning designer is for building custom models — none is the ready-made image-and-OCR service.
Area 4 · Natural language processing
9. A company analyses product reviews to decide whether each one is positive, negative, or neutral. Which NLP capability is this?
Correct answer: Sentiment analysis.
Judging the emotional tone of text is sentiment analysis. Key phrase extraction pulls out main points, entity recognition finds named things like people or places, and translation converts between languages — none scores positivity or negativity.
Area 4 · Natural language processing
10. Which Azure service converts spoken audio into written text?
Correct answer: Azure AI Speech.
Azure AI Speech provides speech-to-text (and text-to-speech). Translator converts text between languages, Vision works on images, and Content Safety detects harmful content — none transcribes audio.
Area 5 · Generative AI
11. A large language model generates a paragraph of text one piece at a time by repeatedly predicting the next what?
Correct answer: Token.
An LLM works by predicting the next token — a chunk of text such as a word or word-part — given everything before it. Pixels belong to images, and clusters and labels are machine-learning concepts, not how text is generated.
Area 5 · Generative AI
12. Which Azure service provides access to large language models for building generative-AI applications such as chat and text generation?
Correct answer: Azure OpenAI Service.
The Azure OpenAI Service gives access to large language models for generative tasks. Azure AI Vision is for images, the Machine Learning designer builds traditional models, and Translator converts languages — none serves generative language models.
🔑 How to read your score
Your result here reflects how you did on these practice questions — it is not a prediction of your real exam score. Use it to spot which skill areas to review next.
Frequently asked questions
Are these real Azure AI-900 exam questions?
No. Every question here is original, written to test the same concepts as the exam. Reproducing real exam items would breach the Microsoft certification agreement and copyright, and it wouldn't help you understand the material.
Do I need to sign up to see the answers?
No. The answer and full explanation for every question are free and shown on this page.
Ready to test yourself under exam-style conditions?
Go hands-on next: explore the AI Architect Academy →
Independent study resource. Not affiliated with, authorized, endorsed by, or sponsored by CompTIA, Amazon Web Services, Microsoft, or ISC2. All trademarks are the property of their respective owners and are used here for identification only. All practice questions are original.