AWS AIP-C01: AWS Certified Generative AI Developer - Professional

Last checked October 22, 2026 against the official AWS exam page.

1. Question 1Foundation Model Integration, Data Management, and Compliance

   A RAG application built on an Amazon Bedrock knowledge base returns outdated policy answers after the team edited documents in an Amazon S3 bucket. Which action makes the updated content searchable?

   1. AIncrease the number of results the retriever returns from 5 to 25 and raise the minimum relevance score
   2. BSwitch to a foundation model with a more recent training cutoff date and a larger maximum context window
   3. CDelete and recreate the foundation model access grant in the Amazon Bedrock console and then re-run the setup
   4. DStart an ingestion job for the knowledge base data source to re-sync the updated S3 documents
   Show answer and explanationHide answer

   Correct answer: D

   Knowledge base content is indexed when an ingestion (sync) job runs, so edits in S3 are not reflected in the vector index until the data source is re-synced. Changing the model or raising the retrieval count does not fix stale indexed content, and model access grants are unrelated to indexing.

2. Question 2Implementation and Integration

   A Bedrock agent must look up order status from an existing REST API. Which component connects the agent to that API?

   1. AKnowledge base data source that is queried with the user's question, returning ranked document passages from an S3 prefix
   2. BGuardrail with a denied topic filter that blocks outbound calls to external REST systems
   3. CPrompt template with the full OpenAPI schema pasted into the system prompt so the model can format HTTP requests as plain text
   4. DAction group whose OpenAPI schema defines the operations, executed by a Lambda function or API integration
   Show answer and explanationHide answer

   Correct answer: D

   An action group describes the callable operations with an OpenAPI schema, and the agent invokes a Lambda function or API integration to perform them. A knowledge base is for retrieval over documents, and prompt text cannot execute an API call.

3. Question 3AI Safety, Security, and Governance

   A company wants to stop customer personal data from appearing in prompts sent to a foundation model and in the responses returned to users. Which Amazon Bedrock control addresses this?

   1. AGuardrail sensitive information filters applied to the model invocation
   2. BProvisioned Throughput with reserved model units for capacity
   3. CModel invocation logging to CloudWatch Logs and S3 with request and response bodies
   4. DBatch inference jobs that read records from an S3 input prefix
   Show answer and explanationHide answer

   Correct answer: A

   Guardrail sensitive information filters detect PII in prompts and responses and can block or mask it at invocation time. Invocation logging records traffic but does not filter it, while throughput and batch settings do not inspect content.

4. Question 4Operational Efficiency and Optimization for GenAI Applications

   A support chatbot sends every request to the largest model, although most questions are simple FAQs. Which optimization reduces cost while keeping quality for complex questions?

   1. ARoute simple FAQs to a smaller model and reserve the largest model for complex requests
   2. BRaise the maximum output tokens for every request so answers are not cut short
   3. CRemove prompt caching because it adds complexity and operational overhead
   4. DUse the largest model for every request to keep answers consistent across users
   Show answer and explanationHide answer

   Correct answer: A

   Model routing sends easy requests to a cheaper model and keeps the expensive model for hard cases, which lowers average cost. Larger outputs and removing caching increase cost, and consistency does not require the largest model everywhere.

5. Question 5Testing, Validation, and Troubleshooting

   A team changes a prompt in its RAG application and needs to detect quality regressions before release. Which practice is most appropriate?

   1. ARely on the model's self-reported confidence score attached to each answer, which the pipeline logs for each build
   2. BAsk one developer to try a single question on a local branch and confirm the answer looks fine to them
   3. CShip the change to production and wait for user complaints to surface through support tickets
   4. DRun an automated evaluation on a golden question set and compare faithfulness and relevance with the previous baseline
   Show answer and explanationHide answer

   Correct answer: D

   A golden dataset with automated metrics gives repeatable comparisons against a baseline, which makes regressions visible before release. Single spot checks, waiting for complaints, and self-reported confidence are not reliable signals.