AIP-C01
AWS Certified Generative AI DeveloperProfessional
AIP-C01 Exam Guide: AWS Certified Generative AI Developer – Professional
AIP-C01 is the AWS Certified Generative AI Developer – Professional exam, not the AWS Certified AI Practitioner exam, whose code is AIF-C01. It validates the ability to integrate foundation models into applications and business workflows, implement production GenAI solutions, and apply security, governance, testing, and optimization practices. This guide helps you decide whether your experience matches the professional target, which domains deserve priority, how to structure practical study, and when to schedule the exam.
First, confirm that AIP-C01 is the exam you need
AIP-C01 is a Professional-level exam for people performing a generative-AI developer role. AIF-C01 is the separate foundational AWS Certified AI Practitioner exam. Confirming the code before studying prevents a costly mismatch between a broad introductory syllabus and an exam focused on production-ready GenAI implementation.
AWS describes AIP-C01 as validating the ability to integrate foundation models into applications and business workflows. The exam also covers vector stores, Retrieval Augmented Generation (RAG), knowledge bases, prompt engineering, agentic AI, optimization, security, governance, testing, monitoring, and evaluation of foundation models.
The distinction matters for preparation. AIF-C01 is aimed at candidates with up to 6 months of exposure who use, but do not necessarily build, AI/ML solutions. AIP-C01 instead targets candidates with 2 or more years of experience building production-grade applications on AWS or with open-source technologies, general AI/ML or data-engineering experience, and 1 year of hands-on GenAI implementation experience.
Use the official AIP-C01 exam guide as the authority for the current scope: https://docs.aws.amazon.com/aws-certification/latest/ai-professional-01/ai-professional-01.html. The AWS certification page also identifies AIP-C01 as AWS Certified Generative AI Developer – Professional: https://aws.amazon.com/certification/certified-generative-ai-developer-professional/.
What capability does the exam validate?
The exam tests production-oriented judgment rather than simple recognition of AI terminology. You should be able to connect a business or application requirement to an appropriate foundation-model architecture, data flow, security control, evaluation method, and operating approach on AWS.
The official objectives include designing and implementing solutions using vector stores, RAG, knowledge bases, and other GenAI architectures; integrating foundation models into applications and workflows; applying prompt engineering and prompt management; implementing agentic AI solutions; optimizing for cost, performance, and business value; and implementing security, governance, and Responsible AI practices.
Testing and operations are part of the capability, not an afterthought. The scope includes troubleshooting, monitoring, optimization, and evaluating foundation models for quality and responsibility. A preparation plan that studies only Amazon Bedrock API patterns will therefore leave important gaps.
The exam is not designed to assess every advanced machine-learning task. The official out-of-scope list includes model development and training, advanced ML techniques, and data engineering and feature engineering. That does not make data irrelevant: you still need to understand data validation, processing, retrieval, storage, and governance as they affect a GenAI application.
How the scored domains should shape your study time
Allocate the most deliberate study and lab time to Content Domain 1: Foundation Model Integration, Data Management, and Compliance, which represents 31% of scored content, and Content Domain 2: Implementation and Integration, which represents 26% of scored content. Together, these domains cover the central architecture and integration decisions in AIP-C01.
The five official domains and their weightings are:
Content Domain 1: Foundation Model Integration, Data Management, and Compliance (31% of scored content).
Content Domain 2: Implementation and Integration (26% of scored content).
Content Domain 3: AI Safety, Security, and Governance (20% of scored content).
Content Domain 4: Operational Efficiency and Optimization for GenAI Applications (12% of scored content).
Content Domain 5: Testing, Validation, and Troubleshooting (11% of scored content).
Do not treat the percentages as a promise about the number of questions you will see in a particular sitting. They are a prioritization tool. A candidate who studies only the largest domains can still lose points through weak security, evaluation, or operational reasoning.
A sensible sequence is to establish architecture and data foundations first, then integration, followed by safety and governance. Finish with optimization and testing, while revisiting all five domains through mixed scenario practice. This sequence reflects dependencies: it is difficult to evaluate or optimize a solution you cannot first describe accurately.
Domain 1: Foundation Model Integration, Data Management, and Compliance
This is the largest domain, representing 31% of scored content, so begin by learning how a foundation model becomes part of a governed application. Study model selection and configuration, prompt strategies, data preparation, vector stores, retrieval, knowledge bases, and compliance considerations as connected design decisions.
The official exam guide identifies tasks such as designing GenAI solutions, selecting and configuring foundation models, implementing data-validation and processing pipelines, designing vector stores, designing retrieval mechanisms, and implementing prompt-engineering strategies and governance.
Build a decision map rather than a catalogue of isolated terms. For a proposed application, ask what information must come from private data, whether retrieval is required, how documents will be prepared and represented, where embeddings or searchable representations will be stored, and how retrieved context will be passed to the model.
Then add compliance questions: What data is allowed into the workflow? Which identities can access source documents and generated outputs? What must be logged? What retention or protection requirements apply? The exam may present these issues through an architecture scenario rather than asking for a definition.
A practical exercise is to sketch a RAG workflow from source data through processing, indexing, retrieval, prompt construction, model invocation, response filtering, and application delivery. Label each boundary with the relevant data, identity, and monitoring concern. The objective is not to memorize one architecture; it is to explain why each component exists and what could fail.
Domain 2: Implementation and Integration
Content Domain 2: Implementation and Integration represents 26% of scored content. Prepare to reason about connecting foundation models to applications and business workflows, including service integration, orchestration, agents, prompt management, deployment choices, and the supporting AWS components needed to make a solution usable.
Study the difference between a model capability and an application capability. A model generates or transforms content, while the surrounding application supplies identity, business rules, tools, data access, workflow control, error handling, and observability. Scenario questions often become easier when you separate those responsibilities.
Amazon Bedrock is central to the listed service scope, alongside Amazon Bedrock AgentCore, Amazon Bedrock Knowledge Bases, Amazon Bedrock Prompt Management, and Amazon Bedrock Prompt Flows. Amazon SageMaker AI and related machine-learning services also appear in scope. Learn what role each service or feature can play, rather than memorizing product names without an architecture.
Include the ordinary application layer in your review. The in-scope list includes AWS Lambda, Amazon ECS, Amazon EKS, AWS Step Functions, Amazon API Gateway, Amazon EventBridge, Amazon SQS, Amazon SNS, Amazon S3, and several data and networking services. Practice selecting a component based on the stated requirement, such as asynchronous processing, workflow orchestration, containerized execution, object storage, or API exposure.
For agentic designs, trace the complete interaction: user request, agent reasoning or planning, tool selection, authorization, tool execution, retrieved context, model response, and final application action. Pay particular attention to permission boundaries and failure handling. An agent that can call a tool is not automatically permitted to perform every action that tool exposes.
Domain 3: AI Safety, Security, and Governance
Content Domain 3: AI Safety, Security, and Governance represents 20% of scored content. Study how to protect data, restrict access, reduce unsafe or unwanted outputs, preserve accountability, and apply Responsible AI practices throughout the GenAI application lifecycle.
The domain should be studied as a set of controls around the model and application, not as a final compliance checklist. Review identity and access management, encryption and key management, secrets protection, network boundaries, logging, data protection, output safeguards, and governance responsibilities.
The in-scope services include IAM, IAM Access Analyzer, IAM Identity Center, AWS KMS, AWS Secrets Manager, Amazon Macie, AWS CloudTrail, Amazon CloudWatch, AWS WAF, Amazon VPC, and AWS PrivateLink. Link each service to a concrete control objective. For example, distinguish authentication and authorization from encryption, secrets storage, audit evidence, and network exposure.
Responsible AI questions require a broader lens than security alone. Consider fairness, toxicity, privacy, explainability, transparency, human oversight, and the risk that model output may be inaccurate or harmful. The exact control depends on the use case, data, users, and consequences of an incorrect response.
A common mistake is to assume that placing a foundation model behind a managed AWS service removes the customer’s responsibilities. Managed infrastructure can reduce operational work, but application owners still need to control prompts, retrieved data, permissions, validation, monitoring, and business use of generated content.
Domain 4: Operational Efficiency and Optimization
Content Domain 4: Operational Efficiency and Optimization for GenAI Applications represents 12% of scored content. Prepare to balance response quality, latency, throughput, reliability, and cost instead of choosing a design solely because it produces the most elaborate response.
Review the levers that affect operating behavior: model selection, prompt size, retrieved-context volume, invocation patterns, caching where appropriate, concurrency, asynchronous workflows, scaling, and monitoring. Consider the business requirement first. A customer-support workflow, a batch summarization process, and an interactive assistant may need different trade-offs.
AWS lists cost and operational services in scope, including AWS Auto Scaling, Amazon CloudWatch, AWS Cost Explorer, AWS Cost Anomaly Detection, AWS Well-Architected Tool, and AWS Systems Manager. The broader service list also includes compute, storage, networking, and deployment tools. Study how these services support visibility, scaling, investigation, and governance.
Do not reduce optimization to lowering inference spend. A cheaper design that increases latency, causes unreliable retrieval, or produces unusable answers may have worse business value. Conversely, a higher-capability model may be justified for a high-consequence task but unnecessary for a routine transformation.
Create an optimization table for each architecture you study. Record the quality target, latency expectation, traffic pattern, cost driver, failure mode, and measurement signal. Then identify the least disruptive change that could improve the stated objective. This trains the trade-off reasoning that scenario questions demand.
Domain 5: Testing, Validation, and Troubleshooting
Content Domain 5: Testing, Validation, and Troubleshooting represents 11% of scored content. Treat it as a disciplined investigation process: define expected behavior, measure quality and safety, isolate the failing layer, change one relevant variable, and verify the result.
Review evaluation of foundation models and GenAI applications across more than factual correctness. Relevant concerns include relevance, groundedness, completeness, toxicity, bias, refusal behavior, latency, availability, and cost. The appropriate metric depends on the application’s intended outcome and risk.
Separate failure categories during study. A poor answer may come from an unsuitable model, weak prompt instructions, missing or stale source data, incorrect chunking or retrieval, insufficient permissions, an application integration defect, or an operational limit. If you cannot identify the layer responsible, changing the model may simply hide the real problem.
Use CloudWatch, CloudWatch Logs, CloudTrail, AWS X-Ray, and Amazon SageMaker Model Monitor as anchors for operational review where they fit the scenario. The in-scope services list also includes Amazon SageMaker Clarify, which is relevant to examining model behavior and responsibility-related concerns.
Write troubleshooting drills without relying on leaked questions or memorized answer keys. For each symptom, list several plausible causes, the evidence you would collect, and the next test. This is more useful than learning that one service is associated with one problem in every situation.
Which AWS services deserve hands-on attention?
Prioritize services by architectural role, beginning with Amazon Bedrock and its listed GenAI features, then adding the AWS services that provide application execution, data, identity, security, monitoring, and delivery. Hands-on work should help you explain service boundaries and trade-offs, not attempt to reproduce a secret exam environment.
The official AIP-C01 in-scope list is non-exhaustive and subject to change: https://docs.aws.amazon.com/aws-certification/latest/ai-professional-01/aip-01-in-scope-services.html. Use it as a scope-checking tool, then consult current AWS documentation for service behavior and limits.
Core GenAI services and features listed in scope include Amazon Bedrock, Amazon Bedrock AgentCore, Amazon Bedrock Knowledge Bases, Amazon Bedrock Prompt Management, Amazon Bedrock Prompt Flows, Amazon SageMaker AI, Amazon SageMaker Clarify, Amazon SageMaker Model Monitor, Amazon SageMaker Data Wrangler, Amazon SageMaker Ground Truth, Amazon SageMaker JumpStart, Amazon SageMaker Model Registry, Amazon SageMaker Processing, Amazon Titan, Amazon Q Business, Amazon Q Developer, Amazon Lex, Amazon Kendra, and Amazon Textract.
The supporting list spans analytics, application integration, compute, containers, databases, developer tools, management and governance, networking, security, and storage. Examples include Amazon S3, AWS Lambda, Amazon ECS, Amazon EKS, AWS Step Functions, Amazon API Gateway, Amazon OpenSearch Service, IAM, AWS KMS, AWS Secrets Manager, Amazon CloudWatch, AWS CloudTrail, Amazon VPC, AWS WAF, AWS CloudFormation, AWS CDK, and AWS CodePipeline.
Do not try to give every listed service equal depth. For each service, determine whether you need to know its primary use, its place in a GenAI architecture, its security or operational implication, or a more detailed implementation pattern. Reserve the deepest work for services that recur across the domain tasks and for areas where your practice results show uncertainty.
A practical six-stage study roadmap
A staged plan works better than reading the service list from top to bottom. Move from exam identity and domain mapping to architecture, implementation, safeguards, operations, and mixed validation. At every stage, produce something usable: a diagram, decision table, lab note, troubleshooting tree, or error log.
Stage 1: establish the boundary. Read the official exam guide, write the five domain names and weightings, and mark topics that are explicitly out of scope. Record your current experience with AWS application development, data work, security, and GenAI implementation. If you lack production experience, plan extra scenario analysis rather than assuming terminology familiarity is enough.
Stage 2: build the architecture base. Study foundation-model selection, prompting, embeddings and retrieval concepts, vector stores, RAG, knowledge bases, data processing, and compliance. Draw several end-to-end workflows. For each diagram, explain the data path, control path, and observability path in plain language.
Stage 3: implement integration patterns. Work through model invocation, application integration, prompt management, workflows, tools, agents, APIs, asynchronous processing, and deployment. Use a small application or structured design exercise to connect Bedrock-related components to Lambda, containers, storage, queues, or orchestration where appropriate.
Stage 4: add safety and governance. Review IAM permissions, encryption, secrets, logging, network access, data protection, output controls, human oversight, and Responsible AI risks. Revisit earlier diagrams and add controls at every boundary. This exposes designs that work functionally but fail security or governance requirements.
Stage 5: optimize and troubleshoot. Take the same designs and introduce symptoms such as irrelevant retrieval, high latency, rising cost, unsafe output, authorization failure, or intermittent tool errors. Identify evidence, isolate the layer, select a corrective action, and define how you would validate the change.
Stage 6: validate readiness. Use mixed, original scenario questions and official objectives to test all domains. Review wrong answers by cause: missing concept, confused service role, overlooked requirement, or careless reading. Schedule only after you can explain why the correct option fits and why the alternatives violate the scenario.
How to study when your background is stronger in one area
Your starting point should change the order of study. Application developers should usually begin with GenAI architecture and then close gaps in model evaluation and Responsible AI. Data or ML practitioners should emphasize AWS application integration, identity, networking, deployment, monitoring, and cost. Candidates with limited production exposure should use complete workflows before isolated feature review.
If you already build AWS applications, do not assume that ordinary cloud knowledge automatically covers foundation-model behavior. Add deliberate work on prompt design, retrieval quality, model evaluation, hallucination risk, agent tool use, and safety controls.
If your experience is mainly with notebooks or experiments, shift toward production constraints. Study how requests are authorized, how data is protected, how workflows recover from failure, how outputs are monitored, and how cost and latency are measured. A successful prototype is not the same as a governed application.
If you are coming from a business or analyst role, first confirm that AIP-C01 matches your intended level. The official target is a GenAI developer with substantial application experience. If your work is primarily consuming AI tools rather than implementing solutions, AIF-C01 may be the more appropriate AWS exam to investigate; do not substitute its foundational scope for AIP-C01 preparation.
Common preparation mistakes to avoid
The most damaging mistake is studying AIF-C01 material for an AIP-C01 objective. After that, the usual problems are memorizing product labels, ignoring operational concerns, treating RAG as a fixed recipe, and using practice answers without understanding the architecture or requirement that makes one option better.
Mistake one is overfitting to a service catalogue. Knowing that Amazon S3 stores objects or that Lambda runs code is not enough. Practice deciding when a service belongs in a design, what it must connect to, and what security or operational consequence follows.
Mistake two is treating the largest domain as the entire exam. Content Domain 3: AI Safety, Security, and Governance represents 20% of scored content, while Content Domain 4: Operational Efficiency and Optimization for GenAI Applications represents 12% of scored content and Content Domain 5: Testing, Validation, and Troubleshooting represents 11% of scored content. Each still requires deliberate preparation.
Mistake three is accepting generated output without evaluation. A response can be fluent and still be inaccurate, ungrounded, unsafe, unauthorized, or unsuitable for the business process. Build evaluation into every architecture exercise.
Mistake four is ignoring the wording of multi-response items. AWS states that multiple-response questions require all correct responses to receive credit. Unanswered questions are scored as incorrect, and AWS states there is no penalty for guessing. Read every option against every requirement rather than selecting the first plausible answer.
Mistake five is relying on dumps, leaked questions, or memorization claims. They cannot establish the production judgment the exam measures and should not be treated as a substitute for official objectives, AWS documentation, and legitimate practice.
What the exam format means for your practice
AIP-C01 contains 75 questions: 65 questions affect your score and 10 questions are unscored. The scored result is reported on a scale of 100–1,000, and the minimum passing score is 750. Use these facts to plan careful reading and complete every item rather than trying to infer a pass mark from raw practice percentages.
The exam contains multiple-choice and multiple-response items. Multiple-choice questions have one correct response and three distractors. Multiple-response questions have two or more correct responses among five or more options, and all correct responses must be selected to receive credit.
AWS states that unanswered questions are scored as incorrect and that there is no penalty for guessing. Mark uncertain items for review when the interface permits it, but do not leave an item unanswered at the end. More importantly, practice identifying the requirement, constraints, preferred AWS operating model, and reason each distractor fails.
The official guide is the best place to verify the current exam content and question information: https://docs.aws.amazon.com/aws-certification/latest/ai-professional-01/ai-professional-01.html. Be cautious when interpreting section-level feedback; AWS explicitly advises caution with that feedback, so use it to locate possible weak areas rather than treating it as a precise diagnostic score.
Do not invent a personal time allocation from the supplied facts. Before scheduling, check the current AWS certification page and appointment information for the delivery conditions that apply to you. The AWS certification page identifies AIP-C01 as available at a Pearson VUE testing center or through online proctoring, lists English, Japanese, Korean, and Simplified Chinese as exam languages, and lists a 180-minute exam duration: https://aws.amazon.com/certification/certified-generative-ai-developer-professional/.
When should you schedule AIP-C01?
Schedule after your evidence shows stable performance across all five domains and you can defend design choices in unfamiliar scenarios. Do not use one strong practice result as the decision. Your readiness should include architecture explanation, security reasoning, evaluation planning, and troubleshooting—not just recognition of service names.
Before booking, verify four things: the code is AIP-C01; your current experience fits the professional target; you have checked the current AWS page for delivery and language details; and your study notes reflect the current official guide rather than an older outline.
A useful readiness review is to choose a blank business scenario and produce an end-to-end design without notes. Include model and retrieval choices, data preparation, identity, encryption, network access, prompt or agent behavior, evaluation, monitoring, cost controls, and failure recovery. Then compare your design against the domain tasks and the in-scope service list.
If you cannot explain why a managed GenAI feature is appropriate, when a supporting AWS service is needed, or how you would measure whether the solution works responsibly, continue studying. A schedule should create a firm review deadline, not compensate for an untested foundation.
What to do in the final review
Use the final review to compress decisions, not reread everything. Revisit domain gaps, service-role confusion, architecture diagrams, security boundaries, evaluation metrics, and the reasons you missed practice items. The last review should make your reasoning faster and more precise.
Create a one-page matrix with five columns: requirement, likely architecture pattern, relevant AWS services, key risk, and validation signal. Fill it with patterns such as grounded question answering, workflow automation, agent tool use, document processing, and monitored production inference.
Review distinctions that commonly control an answer: model capability versus application orchestration; authentication versus authorization; retrieval quality versus model quality; safety control versus ordinary output formatting; cost reduction versus business-value optimization; and prototype success versus production readiness.
Check the official documentation again immediately before scheduling or sitting the exam because AWS states that the in-scope service list is non-exhaustive and subject to change. The current in-scope page is https://docs.aws.amazon.com/aws-certification/latest/ai-professional-01/aip-01-in-scope-services.html.
Finally, make a short list of unresolved questions and research only those questions in AWS documentation. Avoid expanding into unrelated services or advanced model-training material that the official AIP-C01 scope identifies as out of scope.
After certification: keep the credential current
AWS certifications are valid for three years from the date earned and require recertification to remain current and active. Treat recertification as a separate planning obligation: track the certification date, monitor AWS policy updates, and confirm the available recertification path from the official policy page.
The recertification policy is available at https://aws.amazon.com/certification/policies/recertification/. Because certification policies can change, use that page rather than relying on an old study note or an unofficial summary.
Continuing technical review is also practical for this certification because the official AIP-C01 service list is non-exhaustive and subject to change. Keep your architecture knowledge current, especially around foundation-model integration, agents, data retrieval, security, evaluation, and operational controls.
Conclusion
AIP-C01 preparation should look like production design review, not a vocabulary exercise. Confirm the professional exam code, use the five official domains to prioritize work, build end-to-end GenAI architectures, and test every design for security, responsibility, performance, cost, and failure recovery. The next action is straightforward: open the current official AIP-C01 guide, compare its target candidate and domain tasks with your experience, then create a study plan that closes the largest evidence-based gaps before you schedule.