docs: document AI usage

README.md

@@ -2,6 +2,10 @@
 
 Repository for team IT2810-H26-T23 in IT2810-H26.
 
+## Use of AI tools
+
+See [AI usage documentation](docs/ai-usage.md).
+
 ## Working in this repository
 
 `main` is protected — you cannot push to it directly. Work on a branch and
@@ -18,7 +22,7 @@ Before a pull request can be merged:
    request, and approvals from staff do not count — you review each other's
    code.
 2. **Every review comment must be marked resolved.** Reply to each one, fix it
-   or explain why not, then click *Resolve conversation*.
+   or explain why not, then click _Resolve conversation_.
 3. **Pushing new commits dismisses the approval**, so if you change something
    after being approved, ask for another review. This is deliberate: what was
-   approved is what gets merged.
+   approved is what gets merged.

docs/ai-usage.md

@@ -0,0 +1,34 @@
+# Use of AI tools
+
+## How we used AI
+
+We used **ChatGPT, Claude, and Microsoft Copilot** as supplementary tools to
+speed up our workflow and help solve smaller problems. They assisted with
+translating documentation and README text into English, refining wording, and
+providing code examples. **OpenAI Codex** also helped edit this section and run
+project checks.
+
+For documentation, we used AI to turn Norwegian text into English and suggest
+clearer wording while keeping the intended meaning. For programming, code examples
+served as references for smaller tasks and were assessed against our existing code
+and project requirements.
+
+## Review and quality assurance
+
+AI suggestions were starting points; the group made the final implementation
+and design decisions. We critically reviewed suggestions to understand how they
+worked and adapted them where necessary to fit the project. AI was never used
+to generate entire modules or features without human review.
+
+During this documentation update, `npm run lint` and
+`npm run build` both passed, checking ESLint rules, TypeScript types, and the
+production build. These checks supplement human review; they do not verify all
+application behaviour.
+
+## Experiences
+
+AI was useful for reducing time spent on translation and wording, and for getting
+a starting point when working on smaller coding problems. Our main takeaway was
+that suggestions still required understanding and critical review: clear wording
+or plausible code alone was not enough to accept an answer. The group retained
+responsibility for the final text, code, and design decisions.