AIF-C01 · AWS
AIF-C01 added a topic this year. Is it in your study plan?
Version 1.1 of the AIF-C01 guide added a topic that older study material does not cover. What it is, when it reached the exam, what the five domains weigh, and how scoring works.
AWS Certified AI Practitioner (AIF-C01) is a foundational exam on artificial intelligence, machine learning and generative AI, and on the AWS services that deliver them. AWS aims it at anyone exploring AI/ML on AWS, including people in cloud, development, data, IT, AI/ML or line-of-business roles. It asks you to recognise concepts and pick the right tool for a use case, not to build models.
What the exam measures
| Domain | Weight of scored content |
|---|---|
| 1. Fundamentals of AI and ML | 20% |
| 2. Fundamentals of GenAI | 24% |
| 3. Applications of Foundation Models | 28% |
| 4. Guidelines for Responsible AI | 14% |
| 5. Security, Compliance, and Governance for AI Solutions | 14% |
Domains 2 and 3 together are more than half of the exam: generative AI is the centre of gravity, not an add-on to classic machine learning.
Which version you will sit
The exam measures version 1.1 of the exam guide, published on 30 April 2026. Version 1.0 dates from 26 March 2026. AWS keeps a revisions table for this guide and prints the old and new wording of every objective it changed; we went through it in AWS put agentic AI on the AI Practitioner exam.
AWS also says guide updates are published about one month before they are reflected in the exam, so version 1.1 reached the question pool around the end of May. No later version has been announced.
AWS announced that the Italian and German language versions of the exam would be retired after 15 October 2026 (details). The certification and exam objectives are unaffected.
How the exam is scored and delivered
- Format: 65 questions in 90 minutes. 50 count towards your score; 15 are unscored and not identified.
- Question types: multiple choice, multiple response, ordering (put 3 to 5 responses in order) and matching (match responses to 3 to 7 prompts).
- Guessing: unanswered questions are scored as incorrect, and there is no penalty for guessing.
- Scoring: a scaled score from 100 to 1,000, with 700 as the minimum passing score. The model is compensatory: you are assessed on the exam as a whole, not domain by domain.
- Delivery: at a Pearson VUE testing center or online proctored.
- Validity: 3 years. You recertify by passing the latest version of the exam or by earning AWS Certified Machine Learning Engineer Associate.
Distinctions the objectives insist on
- AI, ML, GenAI, deep learning and agentic AI. Objective 1.1.2 asks for the similarities and differences between all five. Since version 1.1, agentic AI is on that list.
- Inference types. Objective 1.1.3 lists batch, real-time, asynchronous and serverless as examples. Study when each fits a use case.
- Traditional ML or a foundation model. Objective 1.2.6 asks when each is appropriate, naming regulatory concerns, explainability requirements and operational constraints as factors.
- Technical and business metrics. Objective 1.3.6 pairs accuracy, precision, recall and F1 with business metrics such as cost per user, development costs, customer feedback and return on investment. Version 1.1 removed Area Under the Curve from the technical list.
- Prompt engineering techniques. Chain-of-thought, zero-shot, single-shot, few-shot and prompt templates, plus the practices that improve responses.
- Limitations of GenAI. Hallucinations, interpretability, inaccuracy and nondeterminism appear by name.
A study plan that follows the guide
AWS guides break each domain into task statements, and each task statement into objectives. Turn those objectives into a checklist and work through the domains by weight:
- Applications of foundation models (28%). Design considerations, prompt engineering, Retrieval Augmented Generation and its business uses, the AWS services that store embeddings in vector databases, inference parameters such as temperature, and how model performance is evaluated.
- Fundamentals of GenAI (24%). Tokens, chunking, embeddings, vectors and transformers; capabilities and limitations; the AWS services for building GenAI and their cost trade-offs.
- Fundamentals of AI and ML (20%). Terminology, use cases, the ML development lifecycle, inference types and metrics.
- Responsible AI (14%). Features of responsible AI, tools such as Amazon Bedrock Guardrails, and responsible model selection.
- Security, compliance and governance (14%). Securing AI systems (IAM, encryption, Amazon Macie, AWS PrivateLink, the shared responsibility model), threats such as prompt injection and data leakage, and compliance services such as AWS Config, Amazon Inspector, AWS Artifact, AWS CloudTrail and AWS Trusted Advisor.
The last two domains are small individually but add up to 28%, as much as the largest domain.
Every date we track for this and the other exams we cover is in the exam calendar.