Use GitHub's AI to Draft Pull Request Descriptions

What This Does

Under deadline pressure, PR descriptions are the first thing that gets skipped, which leaves a reviewer staring at a diff with no context. GitHub's AI generated pull request summary reads the diff and drafts a prose overview plus a bulleted list of what changed, giving reviewers a starting point instead of an empty text box.

Before You Start

• A repository where you're opening or reviewing pull requests
• A Pro subscription active on your account or organization ($10/month): github.com/features/copilot. Pull request summaries are not part of every Copilot tier, so if the option isn't there, check with whoever manages your org's Copilot licenses
• A pull request with a real diff behind it (this feature reads your actual code changes)

Steps

1. Find the AI feature

Open a new pull request you're creating, or an existing one where you're editing the description. Look for a Copilot icon or "Summary" control near the description text box. The same option can also appear as something you trigger from a comment on the PR's main timeline.

2. Tell it what you need

Start with a blank description field before running the summary. Copilot builds its draft from the diff itself and doesn't read or merge anything already typed into the description, so text you wrote first just gets left behind unused.

3. Review and use the result

Read the generated summary against the actual diff, especially if your change touched more than a few hundred lines. Very large diffs get partially excluded from the summary, so a big pull request might come back with a description that covers most, not all, of what changed. Add a sentence for the missing piece and for any context a diff alone can't carry, like why you chose this approach over an alternative.

Real Example

Scenario: You just finished an incremental dbt model change touching the merge logic and two downstream tests, and you're about to open the PR five minutes before a meeting.

What you type/do: Open the new PR, leave the description blank, and trigger the Copilot summary control.

What you get: A prose paragraph describing the merge logic change plus a bulleted list naming the model file and the two test files touched, ready for you to add one line about why the merge key changed before requesting review.

Tips

• Never let sample data or test fixtures in your diff contain real production values, customer names, or credentials. A summary of a diff exposes whatever's in it just as plainly as the diff itself does, and the underlying leak risk exists with or without the AI summary feature.
• Add the "why" yourself even after a good summary. Copilot can describe what changed from the code; it can't know why you rejected the simpler approach or what incident prompted the fix.
• Run the summary again after pushing follow-up commits from review feedback. It regenerates from the current diff, not the one from your first commit.

Tool interfaces change. If a button has moved, look for similar AI/magic/smart options in the same menu area.