diff --git a/README.md b/README.md
index 7e6d8c87..d41c7a58 100644
--- a/README.md
+++ b/README.md
@@ -50,46 +50,37 @@
 
 </div>
 
-## Download and install
+## Install
 
 1. [Download **Blurt.dmg**](https://github.com/AssemblyAI/blurt/releases/latest/download/Blurt.dmg).
 2. Open the disk image and drag **Blurt.app** into `Applications`.
 3. Launch Blurt and follow setup: Microphone, Accessibility, and your [AssemblyAI API key](https://www.assemblyai.com/dashboard/api-keys).
-4. Start dictating with **right command** by default. Tap to toggle, or hold for push-to-talk.
+4. Dictate with **right command** by default. Tap to toggle, or hold for push-to-talk.
+
+Blurt needs macOS 15 or later on Apple Silicon, plus an AssemblyAI API key
+(free tier available).
 
 ## Why Blurt
 
 - **Works anywhere you can type.** Blurt pastes the transcript into the focused Mac app.
-- **No model downloads.** Audio goes to AssemblyAI's Sync STT API and comes back as text.
-- **No extra cleanup step.** The polishing instruction rides with the transcription request.
+- **Polished in one step.** Audio goes to AssemblyAI's Sync STT API and comes back as clean text — no model downloads, no separate cleanup pass.
 - **Small native app.** The setup window, overlay, hotkey, and paste flow are built for macOS.
 - **Actual synth cues.** Start and stop can be cued by real Yamaha DX7 or Roland Juno-106 sounds, or turned off.
 
-## Requirements
-
-| What           | Requirement                             |
-| -------------- | --------------------------------------- |
-| macOS          | macOS 15 Sequoia or later               |
-| Mac            | Apple Silicon                           |
-| Speech-to-text | AssemblyAI API key, free tier available |
-
 ## Privacy
 
 Blurt stores your API key in the macOS Keychain. Audio is captured only while
 you are dictating, then sent over HTTPS to AssemblyAI for transcription. Blurt
 stores no audio and no transcripts.
 
-Release builds send crash reports and a few handled errors to
-[Datadog](https://www.datadoghq.com). Those reports are diagnostic only: stack
-traces and error details, never your audio, transcripts, or API key. Blurt also
-sends an anonymous, random install identifier and app-launch events so we can
-measure unique installs and version adoption — neither is tied to your identity.
-
-Because transcription is processed by AssemblyAI, their policies apply to that
-audio:
+Release builds send crash reports, handled errors, and anonymous install and
+app-launch events to [Datadog](https://www.datadoghq.com) — diagnostics only,
+never your audio, transcripts, or API key, and nothing tied to your identity.
 
-- [AssemblyAI Privacy Policy](https://www.assemblyai.com/legal/privacy-policy)
-- [AssemblyAI Terms of Service](https://www.assemblyai.com/legal/terms-of-service)
+Because transcription is processed by AssemblyAI, their
+[Privacy Policy](https://www.assemblyai.com/legal/privacy-policy) and
+[Terms of Service](https://www.assemblyai.com/legal/terms-of-service) apply to
+that audio.
 
 ## Build from source
 
diff --git a/site/index.html b/site/index.html
index 7259a019..d6b9e9c2 100644
--- a/site/index.html
+++ b/site/index.html
@@ -32,19 +32,6 @@
     />
 
     <meta name="twitter:card" content="summary_large_image" />
-    <meta name="twitter:title" content="Blurt — native macOS dictation" />
-    <meta
-      name="twitter:description"
-      content="Blurt your first draft in any Mac app."
-    />
-    <meta
-      name="twitter:image"
-      content="https://blurtblurt.com/assets/og-card.png"
-    />
-    <meta
-      name="twitter:image:alt"
-      content="Blurt your first draft. Open-source macOS dictation."
-    />
 
     <link rel="icon" type="image/png" href="assets/blurt-b-icon.png" />
 
@@ -114,8 +101,9 @@
           <p class="eyebrow">Native macOS dictation</p>
           <h1>Blurt your first draft.</h1>
           <p class="lede">
-            A small, open-source Mac dictation app cued by real Yamaha DXY and
-            Roland Juno-106 synths.
+            A small, open-source Mac dictation app that turns speech into
+            polished text in the app you're already using — cued by real Yamaha
+            DX7 and Roland Juno-106 synths.
           </p>
 
           <div class="actions">
@@ -136,7 +124,7 @@ <h1>Blurt your first draft.</h1>
           <img
             class="shot"
             src="assets/blurt.png"
-            alt="Blurt's ready screen: tap or hold the hotkey, speak, and cleaned-up text is typed into the focused app"
+            alt="Blurt's ready screen: tap or hold the hotkey, speak, and transcribed text is pasted into the focused app"
             width="1096"
             height="628"
             fetchpriority="high"
diff --git a/site/sitemap.xml b/site/sitemap.xml
index 0f3d9314..6d6a166b 100644
--- a/site/sitemap.xml
+++ b/site/sitemap.xml
@@ -2,7 +2,7 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
   <url>
     <loc>https://blurtblurt.com/</loc>
-    <lastmod>2026-06-27</lastmod>
+    <lastmod>2026-07-01</lastmod>
     <changefreq>monthly</changefreq>
     <priority>1.0</priority>
   </url>