diff --git a/materialize-elasticsearch/driver.go b/materialize-elasticsearch/driver.go
index bd984337b9..7f590a67c5 100644
--- a/materialize-elasticsearch/driver.go
+++ b/materialize-elasticsearch/driver.go
@@ -305,19 +305,24 @@ func normalizeIndexName(index string, truncateLimit int) string {
 	afterPrefix := false
 	var b strings.Builder
 	for _, r := range index {
-		// Replace disallowed characters with underscore. Of the various problematic characters in
-		// an Elasticsearch index name, a "." is the only one that Flow allows in a collection name.
-		if r == '.' {
-			r = '_'
+		// Strip disallowed prefixes.
+		if !afterPrefix && slices.Contains([]rune{'_', '-', '+', '.'}, r) {
+			continue
 		}
 
-		// Strip disallowed prefixes that may be present in a Flow collection name. A '.' is also a
-		// bad prefix, but those were already replaced with '_' above.
-		if !afterPrefix && (r == '-' || r == '_') {
-			continue
+		// Replace disallowed characters with underscore as long as we aren't still at the beginning
+		// of the string. Most of the characters in this list are named in the docs, but ES will
+		// also truncate index names that contain a '#' character to drop everything including &
+		// after that '#' character, so we'll normalize those too.
+		if slices.Contains([]rune{'*', '<', '"', ' ', '\\', '/', ',', '|', '>', '?', ':', '#'}, r) {
+			if !afterPrefix {
+				continue
+			}
+			r = '_'
 		}
 		afterPrefix = true
 
+		// Index names must be lowercase.
 		char := strings.ToLower(string(r))
 		if b.Len()+len(char) > truncateLimit {
 			// Truncate extremely long names. These must be less than 255 bytes.
diff --git a/materialize-elasticsearch/driver_test.go b/materialize-elasticsearch/driver_test.go
index 2c75ed98c0..51f815a4d7 100644
--- a/materialize-elasticsearch/driver_test.go
+++ b/materialize-elasticsearch/driver_test.go
@@ -118,7 +118,13 @@ func TestNormalizeIndexName(t *testing.T) {
 			name:       "dots",
 			byteLength: maxByteLength,
 			input:      ".some..collection",
-			want:       "some__collection",
+			want:       "some..collection",
+		},
+		{
+			name:       "normalized characters",
+			byteLength: maxByteLength,
+			input:      `??weird*<"\/,|>?:#index`,
+			want:       "weird___________index",
 		},
 		{
 			name:       "capitalized",