How to prepare for a moving target

What you will be able to do

• List the seven objectives added in the April 2026 revision
• Judge whether a piece of study material is current enough to trust
• Describe what AWS's own preparation path offers and where to check it
• Use this guide and the exam simulator as tools with separate jobs
• State the limits of an independent question bank and study accordingly

The AIF-C01 exam guide now lives as AWS documentation rather than as a marketing PDF, and one consequence of that move is unusually useful: it is versioned, and it has a published revision history. You can read exactly what changed, when, and in which objective. Very few certifications give you that.

What v1.1 added#

Seven objectives were added, and they cluster around a theme:

Seven services joined the in-scope list at the same time: Amazon Aurora, Amazon Bedrock AgentCore, Kiro, Strands Agents, Amazon Q, Amazon SageMaker JumpStart and AWS Transform. Amazon MemoryDB was removed.

Read the two lists together and the intent is clear. Four of the seven new objectives and five of the seven new services are about agents and the economics of running them. The exam has absorbed the shift from "call a model and read the answer" to "run a system that calls models, uses tools, remembers things and costs money per token".

Why older material is stale, and how stale#

"Stale" is doing precise work here. Material written before 2026 is not wrong about everything, and throwing it out would be an overreaction. It is wrong in two specific ways.

It is missing topics. A 2025 course covers no agentic AI, no Model Context Protocol, no context engineering, no token pricing model, no prompt management, no grounding techniques and no business alignment metrics, because none of those were in the blueprint when it was written. Those seven objectives sit inside domains worth 20%, 24%, 28% and 14% of the paper respectively.

It uses superseded names. The previous lesson has the list. A course that teaches Amazon Bedrock Agents as the way to build an agent, or Amazon QuickSight as a standalone product, or plain "Amazon SageMaker" as the ML service, is describing a surface that has moved.

AWS's own preparation path#

AWS publishes a preparation route for this exam, and it is worth using whatever else you do. The pieces, as AWS describes them:

• The exam guide itself. The authoritative document, with the domains, the task statements, the objectives, the in-scope and out-of-scope service lists and the revision history. Everything else, including this guide, is a commentary on it.
• An exam prep plan on AWS Skill Builder, linked from the certification page and structured as a small number of steps.
• A set of digital courses that map closely onto the five domains, covering machine learning and AI fundamentals, AI use cases, responsible AI, developing ML solutions, developing generative AI solutions, prompt engineering, foundation model optimisation, and security, compliance and governance for AI solutions. Most are around an hour.
• An official practice question set, built by AWS, with feedback on each answer and pointers to related resources.
• An official pretest and an official practice exam. The practice exam is full length with scaled scoring.
• A one-day instructor-led exam preparation course.
• Hands-on options: AWS Builder Labs, AWS Cloud Quest, AWS Jam, AWS SimuLearn and AWS Escape Room.

The strongest reason to sit AWS's own practice material is not the questions. It is calibration to AWS's phrasing. Every certification vendor has habits: how a scenario is set up, how much of it is relevant, which distractors it likes. That is worth more than a score.

How to use this guide#

The five modules after this one map one to one onto the five domains, in the blueprint's order, and their sizes follow the published weights. Reading them in order gives you proportional coverage without having to think about allocation.

A workable sequence for a working week or three:

1. Read a module through once, in short sessions, and use its check questions at the end of each lesson to catch what did not land.
2. Review the flashcards for that module. The service surface is recall material and there is no way round that: matching items give no partial credit, so a service you half know is worth nothing.
3. Only once you have covered all five domains, sit a full simulated paper. Sitting one early tells you what you already knew, which is that you have not studied yet.
4. Work backwards from the wrong answers. Every item links to the lesson that teaches it.

What the exam simulator does and does not do#

The simulator in this guide is configured to match the real paper's shape: 65 items in 90 minutes, reported on a scale of 100 to 1,000 with 700 as the pass mark. Three differences are deliberate and you should hold all three.

It scores all 65 items. The real exam scores 50 and trials 15 without telling you which. Hiding some of our items from the score would teach you nothing and add noise, so we score everything. Our 65 does not map onto AWS's 50.

It cannot simulate two of the four formats. The runner supports single-answer and multiple-answer items. Ordering and matching, both published formats on the real exam, are not simulated, because faking an ordering question as a multiple choice item about ordering is not the same exercise and pretending otherwise would be dishonest. Practise those two formats against the service tables and the lifecycle sequences directly.

Its scaled score is an approximation. A real scaled score equates your form against other forms of the same exam. Ours maps percent correct onto the scale in a straight line. Read it as a rough signal of readiness and nothing more, and remember from the first lesson that 700 was never a percentage in the first place.

The next module starts on domain 1, which is the smallest of the five and the one everything else rests on.