GH-300 study guide: domains, format and a 4-week plan

GH-300 is the GitHub Copilot certification exam. It checks that you can use Copilot responsibly and effectively across the IDE, the CLI and github.com, understand how it handles data, write good prompts, and configure its privacy and organisation settings. It suits developers, DevOps engineers and technical leads who use Copilot day to day. The proctored exam lasts 100 minutes, is delivered through Pearson VUE online or at a test centre, and needs a score of 700 to pass.

GH-300 at a glance

Details as of September 2026 — confirm on the official exam page before booking.

The skills outline was last updated on 7 August 2026. GitHub certifications are valid for two years, and the exam page notes that GitHub is moving to Microsoft's recertification process. If you fail, you can retake the exam 24 hours after the first attempt; later retakes have longer waits under Microsoft's retake policy.

Who this exam is for

The official audience profile describes candidates with expertise in using Copilot to improve productivity, quality and security, covering responsible AI, prompt engineering, Copilot features across plans and privacy safeguards. It also expects familiarity with GitHub fundamentals and experience in at least one programming language.

There is no prerequisite certification. In practice, the exam rewards people who have used Copilot in several surfaces, not only inline suggestions in one editor. If you have never touched organisation policies, content exclusions or the Copilot CLI, plan hands-on time for them. Administrators and team leads get an advantage from skill area 2's policy tasks; individual developers usually need to study that part deliberately.

What the exam covers

Microsoft publishes the weights as ranges. The study guide's summary lists the features skill area twice under slightly different names, apparently a drafting slip; the body of the guide has six skill areas.

Use GitHub Copilot responsibly (15–20%)

This area tests whether you understand the risks and limits of generative AI tools, such as inaccurate or insecure output, bias and over-reliance, and how to mitigate them. The central idea is that you remain responsible for code Copilot suggests: validate it, review it and test it like any other code. Expect scenario questions where the right answer is a human check rather than a setting.

Use GitHub Copilot features (25–30%)

The largest area, and the broadest. It covers enabling and triggering Copilot in the IDE (inline suggestions, chat, agent mode), the GitHub Copilot CLI (installing it, its commands, interactive sessions, generating scripts), and newer capabilities: agent mode, Copilot Edits, MCP, agent sessions and sub-agents, code review, Copilot Spaces, pull request summaries, and custom instructions and prompt files. It ends with organisation administration: policy management, code review policies, audit log events and managing subscriptions through the REST API. Learn which capability fits which job, and where each is configured.

Understand GitHub Copilot data and architecture (10–15%)

Here you explain what happens between a keystroke and a suggestion: how the prompt is built from your context, how data flows to the service and back, and what filtering and post-processing happen on the way. You also need the code suggestion lifecycle and the limitations of the underlying models. GitHub's documentation and Trust Center are the right sources, since marketing summaries are too vague for these questions.

Apply prompt engineering and context crafting (10–15%)

This area covers prompt structure, how Copilot decides which context to use (open files, the current selection, chat history), zero-shot and few-shot prompting, and good practice such as being specific, breaking tasks down and giving examples. Questions often show a weak prompt and ask what would improve the result.

Improve developer productivity with GitHub Copilot (10–15%)

Expect use-case questions: generating, refactoring and documenting code, learning an unfamiliar codebase, producing sample data and modernising legacy code. The testing half covers writing unit and integration tests, finding edge cases and assertions, and asking Copilot for security and performance improvements.

Configure privacy, content exclusions, and safeguards (10–15%)

This area covers content exclusions (what they do, where they are set, and what they do not cover), editor settings, output ownership, the filter that blocks suggestions matching public code, and troubleshooting when suggestions or exclusions do not behave as expected.

A 4-week study plan

Time follows the weights, with the features area getting about a week and a half.

Week 1: Responsible use and prompt engineering. Read GitHub's responsible use documentation and write down the risks it names and the mitigation for each. Then practise prompting in your editor: rewrite vague prompts and compare the results. Take mixed papers for both areas.

Week 2: Features, IDE and CLI. Use Copilot inline, in chat and in agent mode on a real project. Install the Copilot CLI and work through its main commands. Add a custom instructions file and a prompt file to a repository and see how the answers change.

Week 3: Features, organisation settings, then data and architecture. If you can, use a test organisation to look at Copilot policies, audit log events and seat management. Then study data flow, prompt building and filtering. Take topic papers on each administration task.

Week 4: Productivity, privacy, then full mocks. Cover testing and refactoring use cases, and set up and test a content exclusion. Then sit a full-length mock in 100 minutes, review every explanation, fix weak topics and sit a second mock.

Where candidates lose marks

• Studying only the editor. A large share of the features area is about the CLI, github.com features and organisation policy. Developers who only use inline suggestions often miss these.
• Misunderstanding content exclusions. Know the level at which exclusions are configured and their documented limits. Questions test the edge cases, not the happy path.
• Confusing the public code filter with content exclusions. One filters suggestions that match public code; the other stops chosen files being used as context. They answer different concerns.
• Treating Copilot output as trustworthy. The responsible use area consistently rewards validation, review and testing. An answer that removes the human check is rarely right.
• Studying a retired feature. The guide still names GitHub Spark, which stopped accepting new users on 4 August 2026 and closed to existing users on 31 August 2026. Check features against GitHub's current documentation rather than older courses.
• Reading organisation questions too quickly. Note who is making the change and whether the scenario calls for an organisation-wide policy or an individual setting before choosing.

Practising for GH-300 on CertifyAIx

CertifyAIx has 592 original GH-300 questions written against the current study guide, weighted by skill area and explained option by option. Read [how CertifyAIx questions are written](/blog/how-certifyaix-practice-questions-are-written) for our method.

• 3 full-length mock exams of 60 questions in 100 minutes; Microsoft does not publish the real item count, so the mock matches the time limit
• 27 mixed domain papers and 30 topic papers of up to 25 questions each (60 minutes per 25)
• Explanations for every option, with pointers to GitHub's documentation
• Progress tracked by skill area and topic

Start with the [GH-300 practice questions](/certifications/github-copilot-gh-300). For a wider view of where GH-300 fits, see [how to choose an AI certification](/blog/how-to-choose-an-ai-certification).

Frequently asked questions

How difficult is GH-300?

Microsoft rates it intermediate. Regular Copilot users find the prompting and productivity areas familiar; the harder parts are usually data handling, content exclusions and organisation administration, which many developers never touch.

How long should I prepare for GH-300?

If you already use Copilot daily, two to four weeks is a realistic range. Allow more if you have used only one editor or have no access to organisation settings, because those tasks need hands-on time.

How many questions are on GH-300?

Microsoft does not publish a question count. The exam page gives a 100-minute limit and says the exam may include interactive components.

What is the retake policy?

According to the exam page, you can retake GH-300 24 hours after a failed first attempt. Waiting periods for later attempts are set by Microsoft's exam retake policy.

Are CertifyAIx questions real GH-300 exam questions?

No. Every CertifyAIx question is original and written against the public study guide. We never use content from a live exam.

CertifyAIx is independent and not affiliated with GitHub. GitHub Copilot is a trademark of its owner. All CertifyAIx practice questions are original.