Editorial Policy

GitHub Copilot changes faster than most software documented on the open web. Plan names, pricing, model availability, agent terminology and slash commands have all changed materially in recent releases. Any site covering it will be wrong eventually; the question is whether it tells you when, and how quickly it corrects itself.

This page states our commitments so you can hold us to them.

How AI is used

AI tools assist with research and drafting on this site. That is stated plainly rather than hidden, because a site about AI-assisted development that concealed its own use of AI would be hard to take seriously.

What that does and does not mean:

• AI output is never assumed accurate. A model's recollection of a product's pricing or feature set is not evidence. Every factual claim is checked against a primary source before it is published.
• Model training data is treated as stale by default.Where a model and current documentation disagree, the documentation wins, and the model's version is discarded rather than hedged.
• Nothing is published on a model's authority. If a claim cannot be traced to documentation or to something we actually ran, it is cut or explicitly marked as unverified.

Sources

We prefer primary technical sources over secondary commentary. For GitHub Copilot that means, in order of preference:

1. GitHub's product documentation and its underlying source data
2. The GitHub Changelog and official engineering blog posts
3. First-party IDE vendor documentation for editor-specific behaviour
4. Our own reproducible testing

SEO blogs, aggregator articles and other people's tutorials are not used as sources of fact. Every lesson lists the sources actually used at the bottom of the page, each with the date it was consulted — not a decorative bibliography of things we did not read.

Verification and review intervals

Every lesson carries three separate dates, because collapsing them would be misleading:

• Published — when the lesson first went live.
• Updated — when the content last changed substantively. Typo fixes do not move this date.
• Last technically verified — when a human last re-checked the page's factual claims against primary documentation. This is the date that matters most, and it is the one to weigh if you are about to make a purchasing decision.

Pages that state volatile facts are flagged version-sensitive and given a short review interval. A build-time script (npm run content:stale) compares each page's verification date against its interval and reports anything overdue, so staleness surfaces as a failing check rather than as something a reader discovers.

Pages under the most aggressive review

Everything else in GitHub Copilot Fundamentals is reviewed on a longer cycle, because conceptual explanations age more slowly than prices do.

Code and testing

Code examples are tested where practical, and labelled honestly where they are not:

• An example we ran is described as such, and we say what environment it ran in.
• An example we did not run is called an example implementation. It is written to be correct and idiomatic, but we do not claim to have executed it.
• We never write "we tested this and it works" about code we did not run. Perceived authority is not worth the trade.

Copilot's own output is inherently non-deterministic. We therefore do not publish transcripts presented as "what Copilot said", because the same prompt will not reproduce them for you. Where a lesson describes what Copilot tends to do, it says so in those terms.

What we will not publish

• Fabricated screenshots. No mocked-up product UI presented as a genuine capture. Where a screenshot would help but we do not have a current one, the lesson is written to be useful without it.
• Invented commands, flags or shortcuts. Every documented shortcut and command on this site is traceable to GitHub's reference documentation. Where an entry is a natural-language prompt rather than real syntax, it is explicitly labelled as a prompt.
• Invented prices, quotas or credit allowances. All commercial figures come from GitHub's published data and live in one central file, so they cannot drift out of sync between pages.
• Unsupported statistics. No user counts, enrolment numbers, completion rates, page views or productivity percentages that we cannot source.
• Fake testimonials, quotes, benchmarks or test results.
• Awards or credentials we do not hold, and no implication that this site is an official GitHub certification provider.

Corrections

When we get something wrong, the fix goes into the page itself rather than into a note at the bottom. Specifically:

• Factual errors are corrected as soon as they are confirmed against a primary source.
• A meaningful correction moves the Updated date, and re-verification moves the Last technically verifieddate.
• Where a correction changes the practical advice — not just the wording — the lesson says what changed, so a returning reader is not silently working from an out-of-date mental model.
• Where GitHub renames something, we keep the former name visible rather than quietly swapping it, because readers searching the old term still need to find the page.

Independence and incentives

This site is not sponsored by GitHub or Microsoft, and nothing on it is written to encourage a particular purchase. Where a cheaper plan is the right answer — or where the free tier is genuinely sufficient — the lessons say so.

See the disclaimer for the full trademark and affiliation statement.