diff --git a/crates/engine/src/tests.rs b/crates/engine/src/tests.rs
index 09b6dbf..f1f2348 100644
--- a/crates/engine/src/tests.rs
+++ b/crates/engine/src/tests.rs
@@ -125,6 +125,27 @@ fn reset_shortcuts_remove_direct_formatting_but_keep_styles() {
     assert_eq!(text(&s), "Hello world—");
 }
 
+#[test]
+fn non_english_runs_are_not_proofed() {
+    // #100: runs tagged with a proofing language that has no dictionary are skipped
+    // by both proofing paths; English still is proofed.
+    let mut s = s();
+    run(&mut s, "text.insert", json!({"text": "ąžuolas labas"}));
+    let before = run(&mut s, "review.issues", json!({}));
+    assert!(!before.as_array().map(|a| a.is_empty()).unwrap_or(true), "default language flags the words: {before}");
+
+    run(&mut s, "select.all", json!({}));
+    run(&mut s, "review.language", json!({"lang": "lt-LT"}));
+    let after = run(&mut s, "review.issues", json!({}));
+    assert_eq!(after, json!([]), "Lithuanian runs are skipped: {after}");
+
+    // English typos are still caught.
+    let mut e = Session::new(wordcraft_doc::Document::new());
+    run(&mut e, "text.insert", json!({"text": "Hello helo world"}));
+    let en = run(&mut e, "review.issues", json!({}));
+    assert_eq!(en.as_array().map(|a| a.len()).unwrap_or(0), 1, "English typo still flagged: {en}");
+}
+
 #[test]
 fn typing_enter_undo() {
     let mut s = s();
diff --git a/crates/proof/src/lib.rs b/crates/proof/src/lib.rs
index 71a0e4c..34693d7 100644
--- a/crates/proof/src/lib.rs
+++ b/crates/proof/src/lib.rs
@@ -186,6 +186,19 @@ fn known_core(w: &str) -> bool {
     false
 }
 
+/// Is there a proofing dictionary for this BCP 47 language tag? `None` (and empty tags) mean the
+/// document default, which is en-US. Runs in a language without one are skipped (#100) instead
+/// of flagging every word against another language's dictionary; see [`ProofLang::from_tag`].
+pub fn has_dictionary(lang: Option<&str>) -> bool {
+    ProofLang::from_tag(lang) != ProofLang::Other
+}
+
+/// Is `word` spelled correctly in the run's proofing language? A language without a dictionary
+/// can't be checked, so every word counts as correct (no red underlines for it).
+pub fn is_correct_for(word: &str, lang: Option<&str>) -> bool {
+    is_correct_in(word, ProofLang::from_tag(lang))
+}
+
 /// Is `word` spelled correctly? Numbers, single letters, ALL-CAPS acronyms, URLs and words with
 /// digits are accepted.
 pub fn is_correct(word: &str) -> bool {
@@ -688,6 +701,35 @@ mod tests {
         assert!(is_correct("zxqwordcrafty"));
     }
 
+    /// #100: runs are proofed when their language has a dictionary (the default language too)
+    /// and skipped otherwise.
+    #[test]
+    fn dictionary_lang_tags() {
+        for l in [None, Some(""), Some("  "), Some("en"), Some("en-US"), Some("en-GB"), Some("EN-us"), Some("en_US")] {
+            assert!(has_dictionary(l), "{l:?}");
+        }
+        for l in [Some("lt-LT"), Some("fr-FR"), Some("de"), Some("uk"), Some("es"), Some("-US"), Some("e")] {
+            assert!(!has_dictionary(l), "{l:?}");
+        }
+        // The gate follows the dictionaries, not English: every proofed language is proofed.
+        for d in ProofLang::PROOFED.map(ProofLang::code) {
+            assert!(has_dictionary(Some(d)), "{d}");
+            assert!(has_dictionary(Some(&format!("{}-XX", d.to_ascii_uppercase()))), "{d}");
+        }
+    }
+
+    /// #100: runs in a language without a dictionary are never flagged; English still is.
+    #[test]
+    fn non_english_words_are_not_flagged() {
+        for w in ["ąžuolas", "élève", "über", "labas", "bonjour"] {
+            assert!(is_correct_for(w, Some("lt-LT")), "{w}");
+            assert!(is_correct_for(w, Some("fr-FR")), "{w}");
+        }
+        assert!(!is_correct_for("helo", Some("en-US")), "English typos still flagged");
+        assert!(!is_correct_for("helo", None), "default language still proofed");
+        assert!(is_correct_for("hello", Some("en-US")));
+    }
+
     /// File › Options › Proofing (#485): each "Ignore" choice turned off makes those words count.
     #[test]
     fn ignore_options_decide_what_is_checked() {
@@ -757,6 +799,8 @@ mod tests {
             let _ = check_spelling(&s);
             let _ = check_grammar(&s);
             let _ = suggest(&s, 3);
+            let _ = has_dictionary(Some(&s));
+            let _ = is_correct_for(&s, Some(&s));
             let _ = hyphen::hyphen_points(&s, &hyphen::Limits::default());
             for lang in [ProofLang::En, ProofLang::Pl, ProofLang::Other] {
                 let _ = check_spelling_in(&s, lang);
diff --git a/docs/gaps.md b/docs/gaps.md
index 578532f..a0d7636 100644
--- a/docs/gaps.md
+++ b/docs/gaps.md
@@ -69,6 +69,7 @@ The same gaps grouped the way the parity documents are, for agents working in on
 
 | Date | Change | Summary |
 |---|---|---|
+| 2026-10-11 | trivial | #8 narrowed: runs in a proofing language without a dictionary are skipped instead of flagged against the English dictionary (#100); dictionaries still English-only |
 | 2026-10-11 | trivial | #16: DOCX open/save preserves independent complex-script formatting inheritance; RTL sections, tables and typography remain open |
 | 2026-10-11 | trivial | `.doc` pictures, shapes and text boxes read (#551); gap #14 narrowed to the writer and Word 6/95 files |
 | 2026-10-11 | trivial | Table column and row borders and the resize handle drag on the page (#217); UI/UX gap narrowed to the ruler's column markers |
diff --git a/docs/target-app-parity.md b/docs/target-app-parity.md
index 79c94d6..63265d5 100644
--- a/docs/target-app-parity.md
+++ b/docs/target-app-parity.md
@@ -251,6 +251,7 @@ can be used without committing Word output.
 
 | Date | Change | Summary |
 |---|---|---|
+| 2026-10-11 | trivial | Proofing: runs in a language without a dictionary are skipped instead of flagged against the English dictionary (#100); no percentage change |
 | 2026-10-11 | minor | Polish interface (13 UI languages) and Polish proofing with per-run languages: Proofing 35% → 40% (25–45 h → 20–40 h), Localization 25% → 27%; feature depth and readiness unchanged at this precision |
 | 2026-10-11 | trivial | Tables: column and row borders and the resize handle drag on the page (#217); Tables row and interaction-fidelity evidence updated, no percentage change |
 | 2026-10-11 | trivial | Review: Show revisions in balloons takes tracked deletions in the body out of the text into "Deleted: …" balloons (it drew like the comments-and-formatting mode before); Review row updated, no percentage change |