diff --git a/CHANGELOG.md b/CHANGELOG.md
index 5dd40fe..4e8883c 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,7 @@
 - [#50](https://github.com/dblock/ruby-gem-release/pull/50): Add optional release spec helpers, local publication adapters, and a complete example for testing project-specific release customizations in isolated repositories - [@dblock](https://github.com/dblock).
 - [#51](https://github.com/dblock/ruby-gem-release/pull/51): Carry changelog subsections into the next release and remove multiple contribution placeholders - [@dblock](https://github.com/dblock).
 - [#52](https://github.com/dblock/ruby-gem-release/pull/52): Support configured contribution placeholders and complete category templates in Keep a Changelog releases - [@dblock](https://github.com/dblock).
+- [#53](https://github.com/dblock/ruby-gem-release/pull/53): Add Markdown section replacement and text substitution for project-specific release documentation - [@dblock](https://github.com/dblock).
 - Your contribution here.
 
 ## 0.5.1 (2026/11/10)
diff --git a/docs/markdown.md b/docs/markdown.md
new file mode 100644
index 0000000..f7dd4b7
--- /dev/null
+++ b/docs/markdown.md
@@ -0,0 +1,30 @@
+# Updating Markdown Sections
+
+Use `RubyGemRelease::Markdown::Section` to replace a section's body while keeping project-specific wording and version links in your hooks:
+
+```ruby
+require 'ruby-gem-release/markdown/section'
+
+stable = RubyGemRelease::Markdown::Section.new(
+  path: 'README.md', heading: '## Stable Release'
+)
+
+workflow.before(:validate) { |_ctx| stable.validate! }
+workflow.after(:prepare_release) do |ctx|
+  stable.replace!("You're reading the documentation for version #{ctx.release.version}.")
+end
+```
+
+`heading:` is the complete, exact heading line, with one to six leading `#` characters. There must be exactly one matching top-level Markdown heading; headings inside code, quotes, and HTML examples do not count. Missing or duplicate headings raise before writing.
+
+`replace!` replaces everything after that heading through the next equal-or-shallower heading, including nested subsections. It preserves surrounding source, normalizes the replacement's line endings to the file, and accepts repeated replacement. Mixed or unsupported line endings are rejected.
+
+Use `gsub!(pattern, replacement)` to replace matching text without replacing the whole body:
+
+```ruby
+stable.gsub!(/version \d+\.\d+\.\d+/, 'version 0.2.0')
+```
+
+Like `String#gsub!`, the pattern can be a string or regular expression, and replacement strings support capture references. All matches within the body are replaced, including matches in nested subsections and code examples; the heading and surrounding sections remain unchanged. Existing formatting and the final newline are preserved, and inserted line endings follow the file's style. With no matches, it returns `nil` without writing. Missing or duplicate section headings still raise.
+
+The [lifecycle-hooks example](../examples/lifecycle_hooks/release_notes.rb) uses this helper for its stable and development README text. Keep assertions about that wording and the project's release behavior in the project's [custom workflow specs](testing.md).
diff --git a/docs/workflow/hooks.md b/docs/workflow/hooks.md
index ecc8dc0..ea383cf 100644
--- a/docs/workflow/hooks.md
+++ b/docs/workflow/hooks.md
@@ -14,6 +14,8 @@ For each stage, before hooks run in registration order, then the built-in operat
 
 Use the built-in [changelog contribution template options](../changelog/keep_a_changelog.md#contribution-templates) for placeholder removal and category restoration rather than implementing those operations as hooks.
 
+For README updates, use the [Markdown section helper](../markdown.md) to validate and replace one heading's body while keeping project-specific wording in the hook.
+
 ## Context
 
 Hooks receive an immutable execution context with the configured `ctx.workflow` and current `ctx.release`. `ctx.stage`, `ctx.hook`, and `ctx.hook_index` identify the stage, timing, and one-based position. Before capture, `ctx.release` is `nil`; semantic hooks see `ctx.analysis` and the selected version after validation. Next-version hooks see `ctx.release.next_version` after development preparation.
diff --git a/examples/lifecycle_hooks/release_notes.rb b/examples/lifecycle_hooks/release_notes.rb
index d61d622..d96e0cb 100644
--- a/examples/lifecycle_hooks/release_notes.rb
+++ b/examples/lifecycle_hooks/release_notes.rb
@@ -1,14 +1,14 @@
 # frozen_string_literal: true
 
-class ProjectReleaseNotes
-  STABLE_PARAGRAPH = /^## Stable Release\n\n[^\n]+$/
+require 'ruby-gem-release/markdown/section'
 
+class ProjectReleaseNotes
   def initialize(readme: 'README.md')
-    @readme = readme
+    @readme_section = RubyGemRelease::Markdown::Section.new(path: readme, heading: '## Stable Release')
   end
 
   def validate!(_ctx)
-    raise ArgumentError, 'expected exactly one stable release paragraph in README.md' unless File.read(@readme).scan(STABLE_PARAGRAPH).one?
+    @readme_section.validate!
   end
 
   def update_stable_readme!(ctx)
@@ -24,6 +24,6 @@ def update_development_readme!(ctx)
   private
 
   def write_readme(paragraph)
-    File.write(@readme, File.read(@readme).sub(STABLE_PARAGRAPH, "## Stable Release\n\n#{paragraph}"))
+    @readme_section.replace!(paragraph)
   end
 end
diff --git a/lib/ruby-gem-release/changelog/base.rb b/lib/ruby-gem-release/changelog/base.rb
index c5196f4..dff92a4 100644
--- a/lib/ruby-gem-release/changelog/base.rb
+++ b/lib/ruby-gem-release/changelog/base.rb
@@ -1,11 +1,13 @@
 # frozen_string_literal: true
 
 require 'date'
-require 'commonmarker'
+require_relative '../markdown/source'
 
 module RubyGemRelease
   module Changelog
     class Base
+      include Markdown::Source
+
       def initialize(path: 'CHANGELOG.md')
         @path = path
       end
@@ -44,38 +46,6 @@ def unreleased_notes(content, separator)
         content[line_end(content, position, separator)...section_end]
       end
 
-      def heading(content, pattern)
-        matches = heading_positions(content, pattern)
-        raise ArgumentError, "expected exactly one matching heading in #{@path}" unless matches.one?
-
-        matches.first
-      end
-
-      def heading_positions(content, pattern, max_level: 6)
-        lines = content.lines
-        offsets = line_offsets(lines)
-        matches = []
-        document(content).each do |node|
-          next unless node.type == :heading && node.header_level <= max_level
-
-          index = node.source_position[:start_line] - 1
-          matches << offsets[index] if pattern.match?(lines[index].chomp)
-        end
-        matches
-      end
-
-      def heading_prefix(content, position)
-        content[position, 6][/\A#+/]
-      end
-
-      def section_end_position(content, position, level)
-        heading_positions(content, /\A\#{1,6} /, max_level: level).find { |offset| offset > position } || content.length
-      end
-
-      def line_end(content, position, separator)
-        content.index(separator, position) || raise(ArgumentError, "heading must end with a newline in #{@path}")
-      end
-
       def check_release_notes(content, position, separator, version)
         section_end = section_end_position(content, position, heading_prefix(content, position).length)
         check_notes(content[line_end(content, position, separator)...section_end], version)
@@ -103,19 +73,6 @@ def editable_lines(content)
         end
       end
 
-      def document(content)
-        Commonmarker.parse(content, options: { extension: { shortcodes: false } })
-      end
-
-      def line_offsets(lines)
-        offset = 0
-        lines.map do |line|
-          position = offset
-          offset += line.length
-          position
-        end
-      end
-
       def excluded_lines(content)
         excluded = Set.new
         document(content).walk do |node|
@@ -127,12 +84,6 @@ def excluded_lines(content)
         excluded
       end
 
-      def newline(content)
-        raise ArgumentError, "mixed or unsupported line endings in #{@path}" if (content.match?(/(?<!\r)\n/) && content.include?("\r\n")) || content.match?(/\r(?!\n)/)
-
-        content.include?("\r\n") ? "\r\n" : "\n"
-      end
-
       def validate_version(version)
         raise ArgumentError, "invalid version: #{version.inspect}" unless version.is_a?(String) && /\A(0|[1-9]\d*)\.(0|[1-9]\d*)\.(0|[1-9]\d*)\z/.match?(version)
       end
diff --git a/lib/ruby-gem-release/markdown/section.rb b/lib/ruby-gem-release/markdown/section.rb
new file mode 100644
index 0000000..83d95df
--- /dev/null
+++ b/lib/ruby-gem-release/markdown/section.rb
@@ -0,0 +1,73 @@
+# frozen_string_literal: true
+
+require_relative 'source'
+
+module RubyGemRelease
+  module Markdown
+    class Section
+      include Source
+
+      def initialize(path:, heading:)
+        raise ArgumentError, 'heading must be a complete Markdown heading line' unless heading.is_a?(String) && /\A\#{1,6}[ \t]+\S[^\r\n]*\z/.match?(heading)
+
+        @path = path
+        @pattern = /\A#{Regexp.escape(heading)}\z/
+      end
+
+      def validate!
+        read_section
+        self
+      end
+
+      def replace!(body)
+        raise ArgumentError, 'section body must be a String' unless body.is_a?(String)
+
+        content, start, finish, separator = read_section
+        replacement = replacement_body(body, content, finish, separator)
+        File.write(@path, content[0...start] + replacement + content[finish..])
+      end
+
+      def gsub!(pattern, replacement)
+        content, start, finish, separator = read_section
+        body = content[start...finish]
+        return unless body.gsub!(pattern, replacement)
+
+        body = substitution_body(body, separator, following: finish < content.length)
+        File.write(@path, content[0...start] + body + content[finish..])
+      end
+
+      private
+
+      def read_section
+        content = File.read(@path)
+        separator = newline(content)
+        position = heading(content, @pattern)
+        finish = section_end_position(content, position, heading_prefix(content, position).length)
+        start = content.index(separator, position) || content.length
+        [content, start, finish, separator]
+      end
+
+      def replacement_body(body, content, finish, separator)
+        newline(body)
+        text = body.gsub(/\r?\n/, separator).sub(/\A(?:\r?\n)+/, '').sub(/(?:\r?\n)+\z/, '')
+        return separator * 2 if text.empty?
+
+        "#{separator}#{separator}#{text}#{trailing_newlines(content, finish, separator)}"
+      end
+
+      def substitution_body(body, separator, following:)
+        text = body.gsub(/\r?\n/, separator)
+        newline(text)
+        text.prepend(separator) unless text.empty? || text.start_with?(separator)
+        text.concat(separator) if following && !text.end_with?(separator)
+        text
+      end
+
+      def trailing_newlines(content, finish, separator)
+        return separator * 2 if finish < content.length
+
+        content.end_with?(separator) ? separator : ''
+      end
+    end
+  end
+end
diff --git a/lib/ruby-gem-release/markdown/source.rb b/lib/ruby-gem-release/markdown/source.rb
new file mode 100644
index 0000000..47d6efa
--- /dev/null
+++ b/lib/ruby-gem-release/markdown/source.rb
@@ -0,0 +1,62 @@
+# frozen_string_literal: true
+
+require 'commonmarker'
+
+module RubyGemRelease
+  module Markdown
+    module Source
+      private
+
+      def heading(content, pattern)
+        matches = heading_positions(content, pattern)
+        raise ArgumentError, "expected exactly one matching heading in #{@path}" unless matches.one?
+
+        matches.first
+      end
+
+      def heading_positions(content, pattern, max_level: 6)
+        lines = content.lines
+        offsets = line_offsets(lines)
+        matches = []
+        document(content).each do |node|
+          next unless node.type == :heading && node.header_level <= max_level
+
+          index = node.source_position[:start_line] - 1
+          matches << offsets[index] if pattern.match?(lines[index].chomp)
+        end
+        matches
+      end
+
+      def heading_prefix(content, position)
+        content[position, 6][/\A#+/]
+      end
+
+      def section_end_position(content, position, level)
+        heading_positions(content, /\A/, max_level: level).find { |offset| offset > position } || content.length
+      end
+
+      def line_end(content, position, separator)
+        content.index(separator, position) || raise(ArgumentError, "heading must end with a newline in #{@path}")
+      end
+
+      def document(content)
+        Commonmarker.parse(content, options: { extension: { shortcodes: false } })
+      end
+
+      def line_offsets(lines)
+        offset = 0
+        lines.map do |line|
+          position = offset
+          offset += line.length
+          position
+        end
+      end
+
+      def newline(content)
+        raise ArgumentError, "mixed or unsupported line endings in #{@path}" if (content.match?(/(?<!\r)\n/) && content.include?("\r\n")) || content.match?(/\r(?!\n)/)
+
+        content.include?("\r\n") ? "\r\n" : "\n"
+      end
+    end
+  end
+end
diff --git a/spec/ruby_gem_release/markdown/section_spec.rb b/spec/ruby_gem_release/markdown/section_spec.rb
new file mode 100644
index 0000000..1a8ef1d
--- /dev/null
+++ b/spec/ruby_gem_release/markdown/section_spec.rb
@@ -0,0 +1,145 @@
+# frozen_string_literal: true
+
+require 'spec_helper'
+require 'tmpdir'
+require 'ruby-gem-release/markdown/section'
+
+RSpec.describe RubyGemRelease::Markdown::Section do
+  subject(:section) { described_class.new(path: 'README.md', heading: '## Stable Release') }
+
+  around do |example|
+    Dir.mktmpdir do |directory|
+      Dir.chdir(directory) { example.run }
+    end
+  end
+
+  ["\n", "\r\n"].each do |separator|
+    it "substitutes all matching body text while preserving formatting and surrounding sections with #{separator.inspect} endings" do
+      before = "# Project\n\nOld.*\n\n## Stable Release"
+      body = "\n\n  Old.*\n\n### Details\n\n`Old.*`\n\n"
+      after = "## Installation\n\nOld.*\n"
+      File.binwrite('README.md', "#{before}#{body}#{after}".gsub("\n", separator))
+
+      2.times { section.gsub!('Old.*', "New.\nMore.") }
+
+      expected_body = "\n\n  New.\nMore.\n\n### Details\n\n`New.\nMore.`\n\n"
+      expect(File.binread('README.md')).to eq("#{before}#{expected_body}#{after}".gsub("\n", separator))
+    end
+  end
+
+  it 'supports regular expressions and replacement references without adding a final newline' do
+    File.write('README.md', "## Stable Release\n\nVersion 1.2.0; version 2.3.0.")
+
+    section.gsub!(/version (\d+)\.(\d+)\.\d+/i, 'Release \1.\2.9')
+
+    expect(File.read('README.md')).to eq("## Stable Release\n\nRelease 1.2.9; Release 2.3.9.")
+  end
+
+  it 'preserves heading boundaries when matching boundary newlines or inserting into an empty body' do
+    File.write('README.md', "## Stable Release\n\nOld.\n\n## Other\n\nKeep.\n")
+    section.gsub!(/\A\s+|\s+\z/, '')
+    expect(File.read('README.md')).to eq("## Stable Release\nOld.\n## Other\n\nKeep.\n")
+
+    File.write('README.md', '## Stable Release')
+    section.gsub!(/\A/, 'New.')
+    expect(File.read('README.md')).to eq("## Stable Release\nNew.")
+
+    File.write('README.md', "## Stable Release\n## Other\n")
+    section.gsub!(/\A/, 'New.')
+    expect(File.read('README.md')).to eq("## Stable Release\nNew.\n## Other\n")
+  end
+
+  it 'does not write when no body text matches, even if the heading or another section matches' do
+    content = "## Stable Release\n\nKeep.\n\n## Other\n\nStable Release\n"
+    File.write('README.md', content)
+    allow(File).to receive(:write).and_call_original
+
+    expect(section.gsub!('Stable Release', 'New.')).to be_nil
+
+    expect(File).not_to have_received(:write)
+    expect(File.read('README.md')).to eq(content)
+  end
+
+  it 'rejects substitutions that introduce unsupported line endings without writing' do
+    content = "## Stable Release\n\nOld.\n"
+    File.write('README.md', content)
+
+    expect { section.gsub!('Old.', "New.\rBroken.") }.to raise_error(ArgumentError, /line endings/)
+
+    expect(File.read('README.md')).to eq(content)
+  end
+
+  ["\n", "\r\n"].each do |separator|
+    it "replaces the full body, preserves surrounding source, and accepts repetition with #{separator.inspect} endings" do
+      before = "# Project\n\nAn introduction.\n\n## Stable Release"
+      after = "## Installation\n\nKeep *this* formatting.\n"
+      original = "#{before}\n\nOld paragraph.\n\n### Details\n\nOld details.\n\n#{after}"
+      File.binwrite('README.md', original.gsub("\n", separator))
+      section.validate!
+      expect(File.binread('README.md')).to eq(original.gsub("\n", separator))
+
+      2.times { section.replace!("New paragraph.\n\n    Indented code.\n") }
+
+      expect(File.binread('README.md')).to eq("#{before}\n\nNew paragraph.\n\n    Indented code.\n\n#{after}".gsub("\n", separator))
+    end
+  end
+
+  it 'ignores headings inside code blocks, quotes, and HTML examples' do
+    examples = "```\n## Stable Release\n```\n\n> ## Stable Release\n\n<div>\n## Stable Release\n</div>\n\n"
+    File.write('README.md', "#{examples}## Stable Release\n\nOld.\n")
+
+    section.replace!('New.')
+
+    expect(File.read('README.md')).to eq("#{examples}## Stable Release\n\nNew.\n")
+  end
+
+  it 'rejects missing and duplicate real headings without writing' do
+    ["# Other\n\nOld.\n", "## Stable Release\n\nOld.\n\n## Stable Release\n\nOther.\n"].each do |content|
+      File.write('README.md', content)
+      expect { section.validate! }.to raise_error(ArgumentError, /exactly one matching heading/)
+      expect { section.replace!('New.') }.to raise_error(ArgumentError, /exactly one matching heading/)
+      expect { section.gsub!('Old.', 'New.') }.to raise_error(ArgumentError, /exactly one matching heading/)
+      expect(File.read('README.md')).to eq(content)
+    end
+  end
+
+  it 'matches the heading literally rather than as a regular expression' do
+    literal = described_class.new(path: 'README.md', heading: '## Stable [Release]')
+    File.write('README.md', "## Stable Release\n\nKeep.\n\n## Stable [Release]\n\nOld.\n")
+
+    literal.replace!('New.')
+
+    expect(File.read('README.md')).to eq("## Stable Release\n\nKeep.\n\n## Stable [Release]\n\nNew.\n")
+  end
+
+  it 'stops at a shallower heading and leaves it unchanged' do
+    ['# History', "#\tHistory", "History\n======="].each do |following|
+      File.write('README.md', "## Stable Release\n\nOld.\n\n#{following}\n\nKeep.\n")
+      section.replace!('New.')
+      expect(File.read('README.md')).to eq("## Stable Release\n\nNew.\n\n#{following}\n\nKeep.\n")
+    end
+  end
+
+  it 'handles the last section with and without a final newline, including an empty body' do
+    ['## Stable Release', "## Stable Release\n\nOld.", "## Stable Release\n\nOld.\n"].each do |content|
+      File.write('README.md', content)
+      section.replace!('New.')
+      ending = content.end_with?("\n") ? "\n" : ''
+      expect(File.read('README.md')).to eq("## Stable Release\n\nNew.#{ending}")
+      section.replace!('')
+      expect(File.read('README.md')).to eq("## Stable Release\n\n")
+    end
+  end
+
+  it 'rejects mixed and unsupported line endings before editing' do
+    ["## Stable Release\r\n\nOld.\n", "## Stable Release\rOld."].each do |content|
+      File.binwrite('README.md', content)
+      expect { section.replace!('New.') }.to raise_error(ArgumentError, /line endings/)
+      expect { section.gsub!('Old.', 'New.') }.to raise_error(ArgumentError, /line endings/)
+      expect(File.binread('README.md')).to eq(content)
+    end
+    File.write('README.md', "## Stable Release\n\nOld.\n")
+    expect { section.replace!("New.\nMixed.\r\n") }.to raise_error(ArgumentError, /line endings/)
+    expect(File.read('README.md')).to eq("## Stable Release\n\nOld.\n")
+  end
+end