The Google Cloud credential for the people who decide what an organisation does with generative AI. Google states plainly that the exam tests strategic leadership and influence, which makes it the closest any cloud vendor comes to a product certification.
Practice questions (60)Free. Sign in to answer, so your result is kept and reported by module.
At a glance
Cost
$99 Google lists the exam at $99 plus tax where it applies. There is no separate registration fee, and Google charges the full price for a retake. A waiting period applies between attempts and lengthens with each failure.
Format
50 to 60 multiple choice questions in 90 minutes. Candidates sit it online with a proctor or at a test centre, and Google offers the exam in English, Japanese, Spanish and Portuguese.
Pass mark
Google does not publish one. A Google Cloud certification reports a pass or a fail with no score and no breakdown by section, so there is no figure to aim at beyond passing.
Prerequisites
None. Google states the exam is for anyone in any job role, with or without hands on technical experience. Google recommends familiarity with generative AI concepts and asks for no prior certification.
Renewal
Valid for three years. A candidate recertifies by taking the exam again, and Google opens a renewal window some months before expiry.
Based on
The published Generative AI Leader exam guide, which sets out four sections with approximate weightings and lists the considerations under each
What it covers
What it covers. Google Cloud sets and awards this certification, and Product Digest supplies the preparation for it. Google describes the candidate as somebody who can hold a conversation with technical and non technical teams, spot use cases across business functions and influence what gets funded. No question on the exam asks the candidate to build anything.
The heaviest section is Google's own offerings at 35 per cent, and it is the section that ages fastest. Google renamed the platform this exam is built on, so Vertex AI is now Gemini Enterprise Agent Platform and Vertex AI Search is now Agent Search. Older study material still carries the earlier names, and a candidate revising from it will lose marks.
A fifth of the exam is techniques for improving output, covering grounding, retrieval, prompting and sampling parameters. This is the most transferable material on the exam, and it is worth reading even for somebody who never sits it.
The smallest section, business strategy at 15 per cent, carries the Secure AI Framework and responsible AI. Product people find this the easiest reading and skip it, which is a poor trade on a section where the questions are answerable from a page of notes.
The syllabus4 domains · free
Google Cloud’s own domains. Each opens its own page listing the concepts beneath it, and each concept has a page of its own.
A retailer wants a customer service assistant that answers from its own returns policy, checks an order in its warehouse system, and issues a refund where the policy allows it. Which layer of the gen AI landscape does this describe?
AModels, because a foundation model produces every response.
BAgents, because the system plans steps, calls tools and acts rather than only answering.
CInfrastructure, because it requires compute to serve requests.
DApplications, because customers interact with it directly.