diff --git a/src/confluence.rs b/src/confluence.rs
index e0b100e..a44eae2 100644
--- a/src/confluence.rs
+++ b/src/confluence.rs
@@ -15,8 +15,8 @@ use tracing::{debug, warn};
 use url::Url;
 
 use crate::utils::{
-    HeaderHints, decode_html_attribute, ensure_dir, get_file_name_from_url_or_headers, resolve_url,
-    to_markdown_asset_path,
+    HeaderHints, URI_COMPONENT, decode_html_attribute, ensure_dir,
+    get_file_name_from_url_or_headers, resolve_url, to_markdown_asset_path,
 };
 
 // ── Public types ───────────────────────────────────────────────────
@@ -444,18 +444,19 @@ pub async fn download_images_and_rewrite_html(
     }
 
     static REPLACE_RE: Lazy<Regex> =
-        Lazy::new(|| Regex::new(r#"(?is)(<img\b[^>]*\bsrc=)(["'])(.*?)(["'])"#).unwrap());
+        Lazy::new(|| Regex::new(r#"(?is)(<img\b[^>]*\bsrc=)(?:"([^"]*)"|'([^']*)')"#).unwrap());
 
     let result = REPLACE_RE.replace_all(html, |caps: &regex::Captures<'_>| {
         let prefix = &caps[1];
-        let quote_open = &caps[2];
-        let src = &caps[3];
-        let quote_close = &caps[4];
-        if quote_open != quote_close {
+        let (quote, src) = if let Some(m) = caps.get(2) {
+            ('"', m.as_str())
+        } else if let Some(m) = caps.get(3) {
+            ('\'', m.as_str())
+        } else {
             return caps[0].to_owned();
-        }
+        };
         match src_to_local.get(src) {
-            Some(local) => format!("{prefix}{quote_open}{local}{quote_close}"),
+            Some(local) => format!("{prefix}{quote}{local}{quote}"),
             None => caps[0].to_owned(),
         }
     });
@@ -464,7 +465,7 @@ pub async fn download_images_and_rewrite_html(
 }
 
 fn is_local_markdown_asset(src: &str, markdown_image_prefix: &str) -> bool {
-    let encoded_prefix = utf8_percent_encode(markdown_image_prefix, PATH_SEGMENT).to_string();
+    let encoded_prefix = utf8_percent_encode(markdown_image_prefix, &URI_COMPONENT).to_string();
     src.starts_with(&format!("{markdown_image_prefix}/"))
         || src.starts_with(&format!("{encoded_prefix}%2F"))
 }
@@ -708,6 +709,53 @@ mod tests {
         assert_eq!(id, "777888");
     }
 
+    #[tokio::test]
+    async fn rewrite_html_rewrites_image_src_when_url_contains_apostrophe() {
+        let server = MockServer::start().await;
+        // Confluence embeds the raw page title (with apostrophe) in the URL.
+        let img_path = "/download/attachments/123/My%20team's%20page/image.png";
+        Mock::given(method("GET"))
+            .and(path(img_path))
+            .respond_with(ResponseTemplate::new(200).set_body_bytes(b"\x89PNG\r\n\x1a\n"))
+            .mount(&server)
+            .await;
+
+        let html = format!(r#"<img src="{}{}" />"#, server.uri(), img_path);
+
+        let tmp_dir = std::env::temp_dir().join(format!(
+            "confluence2md_test_{}",
+            std::time::SystemTime::now()
+                .duration_since(std::time::UNIX_EPOCH)
+                .unwrap()
+                .subsec_nanos()
+        ));
+        let assets_dir = tmp_dir.join("assets");
+        let client = Client::new();
+        let mut used = HashSet::new();
+        let result = download_images_and_rewrite_html(
+            &client,
+            &html,
+            DownloadImagesOptions {
+                base_url: &server.uri(),
+                personal_access_token: "token",
+                assets_abs_dir: &assets_dir,
+                markdown_image_prefix: "assets",
+                used_names: &mut used,
+            },
+        )
+        .await
+        .unwrap();
+
+        assert!(
+            !result.contains(&server.uri()),
+            "src should be rewritten to local path, got: {result}"
+        );
+        assert!(
+            result.contains("assets"),
+            "src should point into assets dir, got: {result}"
+        );
+    }
+
     #[tokio::test]
     async fn fetch_confluence_page_preserves_content_json_response() {
         let server = MockServer::start().await;
@@ -729,4 +777,37 @@ mod tests {
         assert_eq!(page.export_html, "<p>export</p>");
         assert_eq!(page.storage_html.as_deref(), Some("<p>storage</p>"));
     }
+
+    // Regression test for: is_local_markdown_asset fails when the markdown_image_prefix
+    // contains an apostrophe because the old code used PATH_SEGMENT encoding (which
+    // encodes `'` → `%27`) while to_markdown_asset_path uses URI_COMPONENT encoding
+    // (which keeps `'` as a literal). The mismatch caused local draw.io / image assets
+    // to be treated as remote URLs and re-downloaded.
+    #[test]
+    fn is_local_markdown_asset_recognizes_apostrophe_in_prefix() {
+        let prefix = "confluence2md's_test_assets";
+        let src = to_markdown_asset_path(prefix, "single.drawio.png");
+        // With the old PATH_SEGMENT encoding, encoded_prefix would contain %27 instead
+        // of the literal apostrophe used by to_markdown_asset_path, so starts_with
+        // would return false and this assertion would fail.
+        assert!(
+            is_local_markdown_asset(&src, prefix),
+            "local asset not recognized (apostrophe in prefix): {src}"
+        );
+    }
+
+    #[test]
+    fn is_local_markdown_asset_recognizes_plain_prefix() {
+        let prefix = "my_page_assets";
+        let src = to_markdown_asset_path(prefix, "image.png");
+        assert!(is_local_markdown_asset(&src, prefix));
+    }
+
+    #[test]
+    fn is_local_markdown_asset_rejects_remote_url() {
+        assert!(!is_local_markdown_asset(
+            "https://example.com/image.png",
+            "my_page_assets"
+        ));
+    }
 }
diff --git a/src/utils.rs b/src/utils.rs
index 4b307c9..7da377a 100644
--- a/src/utils.rs
+++ b/src/utils.rs
@@ -313,7 +313,7 @@ pub fn make_assets_info(page_id: &str, title: &str, output_path: &Path) -> Asset
 
 // `encodeURIComponent` percent-encodes everything that is not in the unreserved
 // set `A-Z a-z 0-9 - _ . ! ~ * ' ( )`. Construct the inverse `AsciiSet`.
-const URI_COMPONENT: AsciiSet = NON_ALPHANUMERIC
+pub const URI_COMPONENT: AsciiSet = NON_ALPHANUMERIC
     .remove(b'-')
     .remove(b'_')
     .remove(b'.')