diff --git a/agent-tools/neo4jTools.js b/agent-tools/neo4jTools.js
index 1d5b46a..8b14369 100644
--- a/agent-tools/neo4jTools.js
+++ b/agent-tools/neo4jTools.js
@@ -1,4 +1,5 @@
 const neo4jDriver = require('./mentionsNeo4jDriver');
+const { publicAudioUrl } = require('../utils/audioFormat');
 
 // Neo4j query tools - each does exactly one thing
 const neo4jTools = {
@@ -34,7 +35,7 @@ const neo4jTools = {
                 return result.records.map(record => ({
                     episode: record.get('episode'),
                     creator: record.get('creator'),
-                    audioUrl: record.get('audioUrl'),
+                    audioUrl: publicAudioUrl(record.get('audioUrl')),
                     artworkUrl: record.get('artworkUrl'),
                     quote: record.get('quote'),
                     date: record.get('date'),
diff --git a/docs/details-apis/HIERARCHY_API.md b/docs/details-apis/HIERARCHY_API.md
index a2a91e6..71a1eda 100644
--- a/docs/details-apis/HIERARCHY_API.md
+++ b/docs/details-apis/HIERARCHY_API.md
@@ -50,7 +50,7 @@ curl -X GET "http://localhost:4132/api/get-hierarchy?paragraphId=https___lexfrid
     "paragraph": {
       "id": "https___lexfridman_com__p_6320_p715",
       "metadata": {
-        "audioUrl": "https://cascdr-chads-stay-winning.nyc3.cdn.digitaloceanspaces.com/745287/httpslexfridmancomp6320.mp3",
+        "audioUrl": "https://audio.pullthatupjamie.ai/745287/httpslexfridmancomp6320.mp3",
         "creator": "Lex Fridman Podcast",
         "end_time": 15686.3545,
         "episode": "#481 – Norman Ohler: Hitler, Nazis, Drugs, WW2, Blitzkrieg, LSD, MKUltra & CIA",
@@ -96,7 +96,7 @@ curl -X GET "http://localhost:4132/api/get-hierarchy?paragraphId=https___lexfrid
     "episode": {
       "id": "episode_https___lexfridman_com__p_6320",
       "metadata": {
-        "audioUrl": "https://cascdr-chads-stay-winning.nyc3.cdn.digitaloceanspaces.com/745287/httpslexfridmancomp6320.mp3",
+        "audioUrl": "https://audio.pullthatupjamie.ai/745287/httpslexfridmancomp6320.mp3",
         "creator": "Lex Fridman",
         "description": "Norman Ohler is a historian and author of \"Blitzed: Drugs in the Third Reich\"...",
         "duration": 16281,
diff --git a/routes/jamieExploreRoutes.js b/routes/jamieExploreRoutes.js
index 28334df..564e292 100644
--- a/routes/jamieExploreRoutes.js
+++ b/routes/jamieExploreRoutes.js
@@ -1,4 +1,5 @@
 const express = require('express');
+const { publicAudioUrl } = require('../utils/audioFormat');
 const router = express.Router();
 const { OpenAI } = require('openai');
 const { findSimilarDiscussions, getEpisodeByGuid, getParagraphWithEpisodeData, getClipById, formatResults } = require('../agent-tools/pineconeTools.js');
@@ -373,7 +374,7 @@ router.post('/search-quotes-3d', serviceHmac({ optional: true }), createEntitlem
           if (ep.episodeImage) result.episodeImage = ep.episodeImage;
         }
         if (!result.audioUrl || result.audioUrl === 'URL unavailable') {
-          if (ep.audioUrl) result.audioUrl = ep.audioUrl;
+          if (ep.audioUrl) result.audioUrl = publicAudioUrl(ep.audioUrl);
         }
         if (!result.creator || result.creator === 'Creator not specified') {
           if (ep.creator) result.creator = ep.creator;
@@ -1383,7 +1384,7 @@ router.post('/fetch-research-id', async (req, res) => {
           if (ep.episodeImage) result.episodeImage = ep.episodeImage;
         }
         if (!result.audioUrl || result.audioUrl === 'URL unavailable') {
-          if (ep.audioUrl) result.audioUrl = ep.audioUrl;
+          if (ep.audioUrl) result.audioUrl = publicAudioUrl(ep.audioUrl);
         }
         if (!result.creator || result.creator === 'Creator not specified') {
           if (ep.creator) result.creator = ep.creator;
diff --git a/routes/researchSessions.js b/routes/researchSessions.js
index c8f5e1f..981aa30 100644
--- a/routes/researchSessions.js
+++ b/routes/researchSessions.js
@@ -1,4 +1,5 @@
 const express = require('express');
+const { publicAudioUrl } = require('../utils/audioFormat');
 const router = express.Router();
 const mongoose = require('mongoose');
 
@@ -1694,7 +1695,7 @@ router.post('/enrich', async (req, res) => {
           episode: rest.episode || 'Unknown episode',
           creator: rest.creator || 'Unknown creator',
           episodeImage: rest.episodeImage || null,
-          audioUrl: rest.audioUrl || null,
+          audioUrl: publicAudioUrl(rest.audioUrl) || null,
           date: rest.date || null,
           feedId: rest.additionalFields?.feedId || null,
           guid: rest.additionalFields?.guid || null,
diff --git a/services/tape/tapeShared.js b/services/tape/tapeShared.js
index ae7de4f..9f9cf8a 100644
--- a/services/tape/tapeShared.js
+++ b/services/tape/tapeShared.js
@@ -5,6 +5,7 @@
  */
 
 const crypto = require('crypto');
+const { publicAudioUrl } = require('../../utils/audioFormat');
 const { TapeHttpError } = require('./tapeErrors');
 
 /**
@@ -28,7 +29,7 @@ function candidateFromResult(r) {
     creator: r.creator || null,
     feedId: r.feedId != null ? String(r.feedId) : null, // for source gating (tapeTaste.feed_allow)
     episodeImage: r.episodeImage && r.episodeImage !== 'Image unavailable' ? r.episodeImage : null,
-    audioUrl: r.audioUrl && r.audioUrl !== 'URL unavailable' ? r.audioUrl : null,
+    audioUrl: publicAudioUrl(r.audioUrl && r.audioUrl !== 'URL unavailable' ? r.audioUrl : null),
     startTime: Number.isFinite(start) ? start : null,
     endTime: Number.isFinite(end) ? end : null,
     publishedDate: r.date && r.date !== 'Date not provided' ? r.date : null,
diff --git a/skills/pullthatupjamie/references/podcast-rra.md b/skills/pullthatupjamie/references/podcast-rra.md
index 4820940..93e9456 100644
--- a/skills/pullthatupjamie/references/podcast-rra.md
+++ b/skills/pullthatupjamie/references/podcast-rra.md
@@ -243,7 +243,7 @@ Each result contains:
 - `quote` — transcript text
 - `episode` — episode title
 - `creator` — podcast name
-- `audioUrl` — direct audio file link
+- `audioUrl` — direct audio file link. **Use this value verbatim** to play or download audio; do not construct audio URLs from `feedId`/`guid` or hardcode a CDN host.
 - `date` — publish date
 - `similarity.combined` — relevance score (0-1, aim for >0.84)
 - `timeContext.start_time` / `end_time` — timestamp in seconds
diff --git a/utils/entitlementMiddleware.js b/utils/entitlementMiddleware.js
index 50ff2dc..61bc60b 100644
--- a/utils/entitlementMiddleware.js
+++ b/utils/entitlementMiddleware.js
@@ -6,6 +6,7 @@
  */
 
 const { resolveIdentity, TIERS } = require('./identityResolver');
+const { publicAudioUrl } = require('./audioFormat');
 const { Entitlement } = require('../models/Entitlement');
 const { AgentInvoice } = require('../models/AgentInvoice');
 const JamieVectorMetadata = require('../models/JamieVectorMetadata');
@@ -438,7 +439,7 @@ async function generateDebugMockResponse(entitlementType, query) {
     description: null,
     episode: doc.episode || doc.metadataRaw?.episode || 'Mock Episode',
     creator: doc.creator || doc.metadataRaw?.creator || 'Mock Creator',
-    audioUrl: doc.audioUrl || doc.metadataRaw?.audioUrl || '',
+    audioUrl: publicAudioUrl(doc.audioUrl || doc.metadataRaw?.audioUrl) || '',
     episodeImage: doc.episodeImage || doc.metadataRaw?.episodeImage || '',
     date: doc.publishedDate || 'Date not provided',
     published: doc.publishedDate || null,
diff --git a/utils/workflowOutputFormatter.js b/utils/workflowOutputFormatter.js
index 1947ad5..dfce591 100644
--- a/utils/workflowOutputFormatter.js
+++ b/utils/workflowOutputFormatter.js
@@ -1,4 +1,5 @@
 const BASE_URL = process.env.FRONTEND_URL || 'https://www.pullthatupjamie.ai';
+const { publicAudioUrl } = require('./audioFormat');
 
 /**
  * Format workflow results based on the requested outputFormat.
@@ -48,7 +49,7 @@ function formatAsStructured(result) {
             : null,
           episode: item.episode || null,
           speaker: item.creator || null,
-          audioUrl: item.audioUrl || null,
+          audioUrl: publicAudioUrl(item.audioUrl) || null,
         },
         guid: item.guid || null,
         feedId: item.feedId || null,
diff --git a/utils/workflowSteps.js b/utils/workflowSteps.js
index 268bf36..c132043 100644
--- a/utils/workflowSteps.js
+++ b/utils/workflowSteps.js
@@ -1,4 +1,5 @@
 const { printLog } = require('../constants.js');
+const { publicAudioUrl } = require('./audioFormat');
 const JamieVectorMetadata = require('../models/JamieVectorMetadata');
 const WorkProductV2 = require('../models/WorkProductV2');
 const { findSimilarDiscussions } = require('../agent-tools/pineconeTools');
@@ -26,7 +27,7 @@ function buildMiniPlayer(result) {
       : null,
     episode: result.episode || null,
     speaker: result.creator || null,
-    audioUrl: result.audioUrl || null,
+    audioUrl: publicAudioUrl(result.audioUrl) || null,
   };
 }
 
@@ -132,7 +133,7 @@ async function stepSearchQuotes({ query, feedIds = [], guids = [], limit = 10, m
         quote: metadata.text || metadata.summary || metadata.headline || 'Quote unavailable',
         episode: metadata.episode || metadata.title || 'Unknown episode',
         creator: metadata.creator || 'Unknown',
-        audioUrl: metadata.audioUrl || null,
+        audioUrl: publicAudioUrl(metadata.audioUrl) || null,
         episodeImage: metadata.episodeImage || null,
         date: metadata.publishedDate || null,
         similarity: parseFloat(r.score.toFixed(4)),