diff --git a/crates/ork-server/src/state.rs b/crates/ork-server/src/state.rs
index 4bcc1cf..34cbb58 100644
--- a/crates/ork-server/src/state.rs
+++ b/crates/ork-server/src/state.rs
@@ -1977,23 +1977,81 @@ impl OrkState {
             return Ok("kind: notfound\n".to_string());
         }
 
-        // Fuzzy: first try an exact headline-title match (a bare target that
-        // happens to name a heading), then fall back to the slug id. Phase 2
-        // will extend this with <<target>>/#+NAME/plain-text search returning a
-        // byte offset within the owning section.
+        // Fuzzy resolution follows the Emacs `org-link-search` ladder:
+        //   1. a `<<target>>` dedicated target (or `<<<radio>>>`)
+        //   2. an element named with `#+NAME: target`
+        //   3. a headline whose title matches
+        //   4. (compat) a section whose slug id matches
+        //   5. the target text appearing literally in a section's content
+        // Each branch returns the owning section plus a byte offset relative to
+        // the start of that section's raw content (headline end), so a caller
+        // can scroll within `/body`. Headline/slug matches carry offset 0.
         let want = normalize_search(target);
+
+        // 1. Dedicated target: `<<target>>` or radio `<<<target>>>`.
+        for section in d.doc.all_sections() {
+            let (_content_start, raw) = self.section_content(&d.content, section);
+            if let Some(off) = find_org_target(raw, &want) {
+                return Ok(fuzzy_result(doc_name, &self.section_id(section), off));
+            }
+        }
+
+        // 2. Named element: `#+NAME: target`.
+        for section in d.doc.all_sections() {
+            let (content_start, _raw) = self.section_content(&d.content, section);
+            if let Some(elem) = section.content.iter().find(|e| {
+                e.affiliated.iter().any(|a| {
+                    a.key.eq_ignore_ascii_case("NAME")
+                        && matches!(&a.value,
+                            org_ast::AffiliatedValue::Text(v) if normalize_search(v) == want)
+                })
+            }) {
+                let off = elem.span.start.offset.saturating_sub(content_start);
+                return Ok(fuzzy_result(doc_name, &self.section_id(section), off));
+            }
+        }
+
+        // 3. Headline title match.
         if let Some(section) = d.doc.all_sections()
             .find(|s| normalize_search(&s.headline.title_text()) == want)
         {
             return Ok(fuzzy_result(doc_name, &self.section_id(section), 0));
         }
+
+        // 4. Compat: a section whose slug id matches the bare target.
         if let Some(section) = self.find_section(&d.doc, target) {
             return Ok(fuzzy_result(doc_name, &self.section_id(section), 0));
         }
 
+        // 5. Plain text: the target appears literally in a section's content.
+        for section in d.doc.all_sections() {
+            let (_content_start, raw) = self.section_content(&d.content, section);
+            if let Some(off) = find_plain_text(raw, target) {
+                return Ok(fuzzy_result(doc_name, &self.section_id(section), off));
+            }
+        }
+
         Ok("kind: notfound\n".to_string())
     }
 
+    /// Return a section's raw content region: the slice from the end of its
+    /// headline to the start of its first child (or the section end), together
+    /// with that region's starting byte offset in the document. The offset lets
+    /// callers translate an absolute document position into a section-relative
+    /// one.
+    fn section_content<'a>(&self, content: &'a str, section: &Section) -> (usize, &'a str) {
+        let start = section.headline.span.end.offset;
+        let end = section
+            .children
+            .first()
+            .map(|c| c.headline.span.start.offset)
+            .unwrap_or(section.span.end.offset);
+        if start >= end || end > content.len() {
+            return (start, "");
+        }
+        (start, &content[start..end])
+    }
+
     /// Map a `file:` link's file component to a served document name, if any.
     /// Matches on the file stem (e.g. `notes.org`, `./notes.org`, `notes`
     /// all map to the document `notes`).
@@ -2046,6 +2104,52 @@ fn normalize_search(s: &str) -> String {
     s.split_whitespace().collect::<Vec<_>>().join(" ")
 }
 
+/// Find a `<<target>>` dedicated target (or `<<<target>>>` radio target) in a
+/// section's raw content. Matching is whitespace-insensitive on the target
+/// name (`want` must already be normalized). Returns the byte offset of the
+/// first `<` relative to the start of `content`.
+fn find_org_target(content: &str, want: &str) -> Option<usize> {
+    let bytes = content.as_bytes();
+    let mut i = 0;
+    while let Some(rel) = content[i..].find("<<") {
+        let open = i + rel;
+        // Count the full run of leading '<' so `<<<radio>>>` works too.
+        let mut run = 0;
+        while open + run < bytes.len() && bytes[open + run] == b'<' {
+            run += 1;
+        }
+        let inner_start = open + run;
+        if let Some(close_rel) = content[inner_start..].find(">>") {
+            let inner = &content[inner_start..inner_start + close_rel];
+            // A target name is non-empty and contains no newline.
+            if !inner.is_empty() && !inner.contains('\n')
+                && normalize_search(inner) == want
+            {
+                return Some(open);
+            }
+            // Advance past this candidate's closing delimiter.
+            i = inner_start + close_rel + 2;
+        } else {
+            break;
+        }
+    }
+    None
+}
+
+/// Find the first literal, case-insensitive occurrence of `needle` in a
+/// section's raw `content`. Returns the byte offset relative to the start of
+/// `content`. Whitespace is not normalized — this is the lowest-priority
+/// fuzzy fallback.
+fn find_plain_text(content: &str, needle: &str) -> Option<usize> {
+    let needle = needle.trim();
+    if needle.is_empty() {
+        return None;
+    }
+    let hay = content.to_lowercase();
+    let want = needle.to_lowercase();
+    hay.find(&want)
+}
+
 /// Format a resolution result that names only the owning document (used for a
 /// bare `file:DOC` link with no in-document target).
 fn doc_result(doc: &str) -> String {
@@ -2438,6 +2542,73 @@ mod tests {
         assert!(r.contains("doc: a"), "{}", r);
     }
 
+    #[test]
+    fn test_resolve_fuzzy_dedicated_target() {
+        // `<<target>>` resolves to the owning section with a byte offset into
+        // the section's raw content.
+        let (_d, state) = state_with(
+            "#+TITLE: W\n* Alpha\nIntro line.\nHere is a <<my-anchor>> target.\n");
+        let r = state.resolve_target("w", "my-anchor").unwrap();
+        assert!(r.contains("kind: fuzzy"), "{}", r);
+        assert!(r.contains("section: alpha"), "{}", r);
+        // Offset is relative to section content; it points past the headline.
+        assert!(!r.contains("offset: 0\n"), "offset should be non-zero: {}", r);
+    }
+
+    #[test]
+    fn test_resolve_fuzzy_radio_target() {
+        // `<<<radio>>>` radio targets resolve like dedicated targets.
+        let (_d, state) = state_with(
+            "#+TITLE: W\n* Alpha\nA <<<radio-tgt>>> here.\n");
+        let r = state.resolve_target("w", "radio-tgt").unwrap();
+        assert!(r.contains("kind: fuzzy"), "{}", r);
+        assert!(r.contains("section: alpha"), "{}", r);
+    }
+
+    #[test]
+    fn test_resolve_fuzzy_named_element() {
+        // `#+NAME: foo` makes `[[foo]]` resolve to the named element's section.
+        let (_d, state) = state_with(
+            "#+TITLE: W\n* Alpha\nText.\n#+NAME: fig-one\n| a | b |\n");
+        let r = state.resolve_target("w", "fig-one").unwrap();
+        assert!(r.contains("kind: fuzzy"), "{}", r);
+        assert!(r.contains("section: alpha"), "{}", r);
+    }
+
+    #[test]
+    fn test_resolve_fuzzy_plain_text() {
+        // A bare target that is neither a target, name, nor headline resolves
+        // to the first section whose content contains it (case-insensitive).
+        let (_d, state) = state_with(
+            "#+TITLE: W\n* Alpha\nThe quick brown FOX jumps.\n");
+        let r = state.resolve_target("w", "brown fox").unwrap();
+        assert!(r.contains("kind: fuzzy"), "{}", r);
+        assert!(r.contains("section: alpha"), "{}", r);
+    }
+
+    #[test]
+    fn test_resolve_fuzzy_target_beats_headline() {
+        // When a `<<name>>` target and a same-named heading both exist, the
+        // dedicated target wins (Emacs precedence).
+        let (_d, state) = state_with(
+            "#+TITLE: W\n* Intro\nSee <<spot>> here.\n* Spot\n");
+        let r = state.resolve_target("w", "spot").unwrap();
+        assert!(r.contains("section: intro"), "target in Intro wins: {}", r);
+    }
+
+    #[test]
+    fn test_resolve_fuzzy_cross_document() {
+        // Fuzzy targets fall back across the org path like other link kinds.
+        let (_d, state) = state_with_docs(&[
+            ("a", "#+TITLE: A\n* Start\n"),
+            ("b", "#+TITLE: B\n* Beta\nContains a <<faraway-target>>.\n"),
+        ]);
+        let r = state.resolve_target("a", "faraway-target").unwrap();
+        assert!(r.contains("kind: fuzzy"), "{}", r);
+        assert!(r.contains("doc: b"), "{}", r);
+        assert!(r.contains("section: beta"), "{}", r);
+    }
+
     #[test]
     fn test_relevel_section() {
         let (_d, state) = state_with("#+TITLE: W\n* Parent\n** Child\n* Other\n");