The best AI certification is the one that matches your role and the cloud you already use. AWS, Microsoft and Google Cloud each offer a foundational AI credential, and all three have rebuilt their higher levels around generative AI and agents. Here are the current exams at each level, who they suit and how to choose.

Foundational: prove you understand AI

These suit managers, analysts, sales and support staff, and anyone starting out.

  • AWS Certified AI Practitioner (AIF-C01): 65 questions in 90 minutes, with a passing score of 700 out of 1,000. It is for people who use AI on AWS but do not necessarily build it, and coding is out of scope. See our AI Practitioner guide.
  • Microsoft Azure AI Fundamentals (AI-901): this replaced AI-900 on 30 June 2026. It now expects basic Python and gives 55 to 60% of the exam to building simple solutions in Microsoft Foundry.
  • Google Cloud Generative AI Leader: 50 to 60 multiple choice questions in 90 minutes, for any job role, with or without hands-on technical experience. Google does not use exam codes.

Associate: build and run AI solutions

These suit developers, ML engineers and data scientists with hands-on experience.

  • AWS Certified Machine Learning Engineer, Associate: the last English sitting of MLA-C01 was on 28 September 2026, and the updated MLA-C02 has run as a beta since 29 September, adding generative AI, agentic AI and foundation model content. AWS recommends a year of experience with Amazon SageMaker AI and Amazon Bedrock. Our MLA-C02 guide has the detail.
  • Microsoft Azure AI Apps and Agents Developer Associate (AI-103): this replaced AI-102 on 30 June 2026. It is for Python developers building generative AI apps and agents in Microsoft Foundry.
  • Microsoft Machine Learning Operations Engineer Associate (AI-300): this replaced the Azure Data Scientist Associate (DP-100), which retired on 1 June 2026. It covers machine learning and generative AI operations for people with a data science background.
  • Microsoft Azure AI Cloud Developer Associate (AI-200): the successor to AZ-204, for back-end developers building AI into Azure services.
  • Google Cloud has no AI-specific associate certification; its AI path runs from Generative AI Leader to the professional level.

Professional and expert: design AI at scale

These suit experienced engineers and architects.

  • AWS Certified Generative AI Developer, Professional (AIP-C01): 75 questions in 180 minutes, with a passing score of 750. AWS recommends two or more years building production applications plus a year of hands-on generative AI work. It is about putting foundation models into applications, not training them.
  • Google Cloud Professional Machine Learning Engineer: 50 to 60 questions in two hours, with three or more years of industry experience recommended. Google has updated it for the move from Vertex AI to Gemini Enterprise Agent Platform.
  • Google Cloud Professional Agentic Architect: a new certification, with the full exam due to launch in mid-November 2026. It combines a three-hour multiple choice exam with hands-on labs, and Google recommends at least a year building agentic solutions on Google Cloud.
  • Microsoft Agentic AI Business Solutions Architect Expert: you pass exam AB-100 and hold one of several listed associate certifications, which include AI-103 and AI-300.

Retired and changing: check before you book

AI certifications are changing fast. In 2026 alone, AWS retired its Machine Learning Specialty (the last exam date was 31 March 2026), and Microsoft retired AI-900, AI-102 and DP-100. MLA-C02 is still in beta, and Microsoft updates the English AB-100 exam on 16 November 2026. A course from last year may describe an exam that no longer exists.

Check the official pages for current details before you book: AWS Certification, Microsoft Learn credentials and Google Cloud certification. Fees vary by country, so confirm yours when you register.

How to choose your path

Start with the cloud your employer or clients use. One credential on the platform you use daily is worth more than three you never touch.

  • You lead, sell or manage AI work: Generative AI Leader or AWS AI Practitioner.
  • You are new to AI but technical: AI-901 or AWS AI Practitioner, then an associate exam.
  • You build apps and agents: AI-103 on Azure, AIP-C01 on AWS once experienced, or Agentic Architect on Google Cloud.
  • You train, deploy and monitor models: MLA-C02, AI-300 or the Google Professional Machine Learning Engineer.

Our guide to AI skills worth learning in 2026 helps if you are still choosing a direction.

Practise under real exam conditions

Every exam above is timed and proctored, and the higher levels run for two or three hours, so pace matters as much as knowledge. On CandidatesPrep you can sit timed mocks that mirror the real exam's timer, navigation and flagging, then see your results broken down by topic, so weak areas such as responsible AI or deployment show up before exam day. Trainers can schedule mocks for a cohort and track each candidate's readiness.

The bottom line

There is no single best AI certification, only the right next step for your role and cloud. Start foundational if you are new, move to an associate exam once you build or run AI systems, and check the provider's page first, because codes and content change quickly.

Get ready with a timed mock on the CandidatesPrep simulator, or book a demo to prepare a whole cohort.