Azure AI Fundamentals (AI-900): Free Mock Test & Practice Questions

Azure AI Fundamentals (AI-900)

Practice for Microsoft's AI fundamentals cert.

Sit a full mock under the clock at least once. Nerves are half the battle.

Free practice questions

Try 3 real Azure AI Fundamentals (AI-900) questions

Have a go, then reveal the worked explanation. A taster from a bank of 325 original questions — unlock for unlimited drills and full timed mocks.

Q1. A retail company wants to identify unusual spikes in daily website traffic that could indicate a fraud attack. Which type of AI workload best fits this requirement?

A. Anomaly detection
B. Computer vision
C. Natural language processing
D. Knowledge mining

Show answer & explanation

✓ Answer: A. Anomaly detection

Anomaly detection workloads are designed to identify data points that deviate significantly from expected patterns, such as unusual traffic spikes. Computer vision and NLP work with images and text respectively, not numeric time-series patterns.

Q2. What does the term 'overfitting' describe in a machine learning model?

A. The model performs poorly on both training and new data
B. The model performs very well on training data but poorly on new, unseen data
C. The model has too few features to make accurate predictions
D. The model takes too long to train due to excessive data volume

Show answer & explanation

✓ Answer: B. The model performs very well on training data but poorly on new, unseen data

Overfitting occurs when a model learns the training data too closely, including its noise, so it fails to generalise to new data. This is different from underfitting, where the model performs poorly on both training and new data because it is too simple.

Q3. After training a Custom Vision object detection model, the evaluation shows high precision but low recall. What does this indicate?

A. Most positive predictions are correct, but many actual objects in the images were missed
B. Most positive predictions are wrong, but nearly all actual objects were found
C. The model performs equally poorly at finding and correctly labelling objects
D. The model cannot be published until precision and recall are both 100%

Show answer & explanation

✓ Answer: A. Most positive predictions are correct, but many actual objects in the images were missed

High precision means that when the model predicts an object, it is usually correct; low recall means the model is failing to detect many objects that are actually present, effectively under-predicting. Confusing precision and recall in the opposite direction is a common exam trap.

Exam format (pass mark, question count, timing) verified against Microsoft Credentials. Questions are our own, not the official paper.

Start here

Ten questions to see where you stand

Two minutes, no timer, marked the moment you finish with every answer explained. Then you'll know which topics need the work before you sit a full mock.

Exam mode

Full mock exam

50 random questions, real timing, marked against the official pass mark. Different every time.

Create a free account for a full mock

No card needed. A free account unlocks a full timed mock and 3 drills on every exam.

Try it free

Topic drills

Ten quick questions on one topic, marked instantly with explanations.

1 free drill left — then a free account unlocks 3 + a full mock.

Azure AI Fundamentals (AI-900) — common questions

Frequently asked

How many questions are in the Azure AI Fundamentals (AI-900)?

The exam has 50 questions. Every Revision Robin mock uses the same 50-question format so your practice matches the real thing.

What is the pass mark for the Azure AI Fundamentals (AI-900)?

You need 35 out of 50 correct to pass, which is about 70%. Our mocks mark you against this exact threshold.

How long do you get?

The test is timed at 45 minutes. Our full mock runs on the same clock so you can practise your pacing.

Are these the real exam questions?

No. Our questions are original and written to match the current syllabus, so they give realistic practice without copying the official paper. Every answer comes with a plain-English explanation. Always confirm current rules and content with Microsoft.