diff --git a/website/docs/api/annotation.jade b/website/docs/api/annotation.jade
index 93511899b..adc6b28f7 100644
--- a/website/docs/api/annotation.jade
+++ b/website/docs/api/annotation.jade
@@ -79,3 +79,33 @@ p
 +h(2, "named-entities") Named Entity Recognition
 
 include _annotation/_named-entities
+
++h(2, "json-input") JSON input format for training
+
+p
+    |  spaCy takes training data in the following format:
+
++code("Example structure").
+    doc: {
+        id: string,
+        paragraphs: [{
+            raw: string,
+            sents: [int],
+            tokens: [{
+                start: int,
+                tag: string,
+                head: int,
+                dep: string
+            }],
+            ner: [{
+                start: int,
+                end: int,
+                label: string
+            }],
+            brackets: [{
+                start: int,
+                end: int,
+                label: string
+            }]
+        }]
+    }
diff --git a/website/docs/usage/cli.jade b/website/docs/usage/cli.jade
index ca33e4e40..74d6554b0 100644
--- a/website/docs/usage/cli.jade
+++ b/website/docs/usage/cli.jade
@@ -143,7 +143,8 @@ p
     +tag experimental
 
 p
-    |  Train a model. Expects data in spaCy's JSON format.
+    |  Train a model. Expects data in spaCy's
+    |  #[+a("/docs/api/annotation#json-input") JSON format].
 
 +code(false, "bash").
     python -m spacy train [lang] [output_dir] [train_data] [dev_data] [--n_iter] [--parser_L1] [--no_tagger] [--no_parser] [--no_ner]