ExamPark
Course menuAIF-C01

Exams / AIF-C01 / Study guide

AWS Certified AI Practitioner (AIF-C01) study guide

What the AIF-C01 exam tests, domain by domain, and how to tell the right answer from the nearly right one. Each topic links to fuller study notes and free practice questions.

Last checked against the official exam guide and AWS documentation · How we write our content

About the AIF-C01 exam

The AWS Certified AI Practitioner exam checks that you understand AI, machine learning and generative AI well enough to talk about them, choose between approaches, and use AWS AI services responsibly. You don’t need to build or train models. AWS aims it at people with up to 6 months of exposure to AI and ML on AWS who use AI solutions rather than build them: business analysts, product managers, sales and marketing people, IT support and managers.
Questions 65: 50 scored and 15 unscored, which aren’t marked
Time 90 minutes
Question types Multiple choice, multiple response, ordering (put 3 to 5 steps in order) and matching (pair 3 to 7 prompts with answers)
Credit Multiple response, ordering and matching questions need every part right to score; there is no partial credit
Score Scaled from 100 to 1,000; 700 to pass
Guessing An unanswered question counts as wrong, and there is no penalty for a wrong answer, so answer everything

AWS updated the exam guide to version 1.1 in April 2026, adding agentic AI and newer services such as Amazon Bedrock AgentCore. Check the official certification page for the current price, languages and booking options before you book.

What the exam really tests

Most questions describe a business that wants to use AI and ask which approach, service, setting or safeguard fits. You’ll see three kinds:

  • Vocabulary in context: is this classification or regression, overfitting or underfitting, a hallucination or a prompt injection?
  • Choosing an approach: prompt engineering, RAG or fine-tuning; a ready-made AI service or a custom model; on-demand or provisioned throughput.
  • Responsible and secure use: guardrails, bias, explainability, data privacy, who is responsible for what.

The cheapest, simplest approach that meets the need is usually right. If a ready-made AI service does the job, it beats training a model. If better instructions fix the problem, prompt engineering beats fine-tuning. If answers need the company’s latest documents, RAG beats retraining.

How to read a question

  1. Find what the business needs the AI to do: classify, predict a number, generate text, answer from documents, take actions.
  2. Note the constraints: cost, effort, data the company has, how current the answers must be, privacy.
  3. Pick the least complex option that meets every constraint.
  4. For ordering questions, think through the real sequence (collect data, prepare, train, evaluate, deploy, monitor) before you look at the options.
  5. For matching questions, start with the pairs you are sure of; the rest often fall into place.

How the domains add up

Applications of Foundation Models is the largest domain, and together with Fundamentals of Generative AI it makes up more than half the scored content. The responsible AI and security domains are smaller but very learnable, and their questions often come down to knowing which AWS feature handles a risk.

Domain 1: Fundamentals of AI and ML 20%

This domain covers the language of AI and machine learning: how AI, ML and deep learning relate, supervised and unsupervised learning, the kinds of problem ML solves (classification, regression, clustering, forecasting), when ML is the wrong tool, AWS’s ready-made AI services, and the ML lifecycle from data collection to monitoring, including MLOps and the metrics used to judge a model.

How AI, machine learning, deep learning and generative AI fit inside each other Four nested areas. Artificial intelligence is the outermost. Inside it is machine learning, inside that is deep learning, and inside that is generative AI. Artificial intelligence Machine learning Deep learning Generative AI foundation models, LLMs
Each field sits inside the one before. AI is any system that performs tasks that seem to need human intelligence. Machine learning learns patterns from data instead of following hand-written rules. Deep learning uses neural networks with many layers. Generative AI uses very large deep learning models, foundation models, to create new text, images or code.

Domain 2: Fundamentals of Generative AI 24%

Generative AI questions test the core ideas (tokens, embeddings, transformers, foundation models, diffusion models, context windows), the strengths and weaknesses of generative AI for a business (speed and flexibility against hallucinations and cost), AI agents and how they use tools, and the AWS services for building generative AI applications, led by Amazon Bedrock.

Domain 3: Applications of Foundation Models 28%

The largest domain is about putting foundation models to work: choosing a model and setting inference parameters, grounding answers in company data with RAG and knowledge bases, writing effective prompts, deciding when to fine-tune or customize a model, and evaluating whether the result is good enough. Many questions ask you to pick the cheapest approach that meets the need.

Ways to customize a foundation model, from cheapest to most expensive Four options in order of rising cost and effort. Prompt engineering changes only the instructions. Retrieval-augmented generation adds your documents at question time. Fine-tuning trains the model further on labeled examples. Continued pre-training trains it on large amounts of unlabeled domain text. Prompt engineering better instructions RAG adds your documents Fine-tuning labeled examples Continued pre-training unlabeled domain text Cost, effort low high
Start at the top and move down only when you need to. Prompt engineering costs nothing extra. RAG grounds answers in current company data without changing the model. Fine-tuning teaches a style or task from labeled examples, and continued pre-training teaches a whole domain’s language from unlabeled text, both at real training and hosting cost.

Domain 4: Guidelines for Responsible AI 14%

Responsible AI questions cover the features of a responsible system (fairness, explainability, privacy, safety, controllability, veracity, transparency, governance), how bias gets into models and how to detect it, the trade-off between powerful and explainable models, and the AWS tools that help: Amazon Bedrock Guardrails, SageMaker Clarify, model cards and human review with Amazon A2I.

Domain 5: Security, Compliance, and Governance for AI Solutions 14%

This domain applies familiar AWS security to AI: IAM, encryption, private networking, logging and the shared responsibility model, plus AI-specific risks such as prompt injection and data leakage. It also covers data governance (lineage, retention, residency) and the tools that record and check compliance, such as CloudTrail, Config, Audit Manager and AWS Artifact.

A 4-week study plan

This plan assumes 30 to 60 minutes on most days. If you already work with generative AI, two or three weeks may be enough.

Week Focus Do this
1 Baseline and ML fundamentals Take the free practice exam in review mode. Then learn AI vs ML vs deep learning, supervised vs unsupervised learning, the ML lifecycle, the main metrics, and AWS’s ready-made AI services.
2 Generative AI Tokens, embeddings, foundation models, hallucinations, inference parameters, Amazon Bedrock, and AI agents (Bedrock Agents, AgentCore, MCP).
3 Applying foundation models RAG and knowledge bases, vector stores, prompt engineering techniques, fine-tuning and other customizations, and model evaluation.
4 Responsible AI, security and practice Guardrails, bias, explainability, the shared responsibility model for AI, data governance. Then full timed practice exams, reviewing every wrong answer.

How to use practice questions well

  • Practice the ordering and matching formats. They have no partial credit, so get comfortable with them before the exam.
  • Read every explanation. The wrong options often describe real techniques that suit a different problem.
  • Keep a “which approach” list. Prompt engineering, RAG, fine-tuning, continued pre-training and distillation each solve a different problem; write one line for each.
  • Aim for 80% or more on a few full timed practice exams before you book.

Try it yourself

Ten minutes in the Amazon Bedrock console teaches more than an hour of reading. In a personal account with a budget alert set, try the chat playground, change the temperature and compare answers, and create a small knowledge base from a few documents. Delete what you create.

Frequently asked questions

Do I need to know how to code or do maths?

No. The exam tests concepts and choices, not code or formulas. Knowing what a metric means (for example, that recall measures how many real cases a model finds) is enough.

How is AI Practitioner different from Cloud Practitioner?

Cloud Practitioner covers AWS as a whole. AI Practitioner goes deeper on AI, ML and generative AI, and only touches general AWS topics such as security where they apply to AI. Neither is a prerequisite for the other.

What comes after AI Practitioner?

For hands-on ML work, the AWS Certified Machine Learning Engineer – Associate. For building generative AI applications, AWS also offers the AWS Certified Generative AI Developer – Professional.

Are the practice questions the real exam questions?

No. Every question on this site is original, written from the public exam guide and checked against AWS documentation. Using real exam questions breaks the AWS Certification Program Agreement.

Test yourself

The free AIF-C01 practice exam has 65 questions with an explanation for every option. No sign-up needed.

Take the free practice examReview questions untimed

Want to be sure you're ready? The 5 full-length practice sets add 325 new questions, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets