diff --git a/README.md b/README.md
index 86d1a24b..75c2db10 100644
--- a/README.md
+++ b/README.md
@@ -9,6 +9,7 @@ Categorize changes, highlight contributors, and maintain consistent release docu
 - [Overview](#overview)
 - [Motivation](#motivation)
 - [Quick Start](#quick-start)
+- [How Release Notes Are Recognized](#how-release-notes-are-recognized)
 - [Requirements](#requirements)
 - [Configuration](#configuration)
 - [Example Workflow](#example-workflow)
@@ -76,6 +77,26 @@ https://github.com/org/repo/compare/v1.1.0...v1.2.0
 3. Extract release note text and contributors.
 4. Output a Markdown section ready for publishing.
 
+## How Release Notes Are Recognized
+
+The action looks for a section whose heading matches the first regex `[Rr]elease [Nn]otes:` (configurable via `release-notes-title`) in the **issue body**, then in each **linked PR body**. Every bullet line (`-`, `*`, or `+`) immediately below that heading becomes one release note entry.
+
+**Example — issue or PR body:**
+
+```markdown
+Some intro text
+
+Release Notes:
+- Add user MFA enrollment flow
+- Improve cache invalidation logic
+
+Other commentary...
+```
+
+Only the first matching section per body is used; content outside it is ignored. If no section is found in any linked PR body and [CodeRabbit Integration](docs/features/coderabbit_integration.md) is enabled, the AI-generated summary is used as a fallback.
+
+For full details — including the configurable title regex, `skip-release-notes-labels`, and edge cases — see [Release Notes Extraction](docs/features/release_notes_extraction.md).
+
 ## Requirements
 
 To run this action successfully, make sure your environment meets the following requirements:
diff --git a/docs/features/custom_chapters.md b/docs/features/custom_chapters.md
index a6827563..953c2a66 100644
--- a/docs/features/custom_chapters.md
+++ b/docs/features/custom_chapters.md
@@ -321,9 +321,12 @@ Feature #2 has no `scope:security` label of its own, but its aggregated set incl
 Children within each rendered node are sorted **ascending by issue number**.
 
 ### Validation
-- Entries missing `title` or `label`/`labels` are skipped with a warning.
 - Non-dict entries are skipped with a warning.
-- Empty labels after normalization cause the entry to be skipped with a warning.
+- A missing `title` key causes the entry to be skipped with a warning.
+- A `title` that is not a string or is blank/whitespace-only is skipped with a warning.
+- A missing `label`/`labels` key causes the entry to be skipped with a warning.
+- An invalid `label`/`labels` type (not a string or list) is treated as empty and causes the entry to be skipped with a warning.
+- Labels that are empty after normalization (e.g. only blank tokens) cause the entry to be skipped with a warning.
 
 ## Related Features
 - [Duplicity Handling](./duplicity_handling.md) – governs multi-chapter visibility.