diff --git a/packages/svelte/src/compiler/phases/1-parse/state/element.js b/packages/svelte/src/compiler/phases/1-parse/state/element.js
index 9082b76c4972..f59f2566e387 100644
--- a/packages/svelte/src/compiler/phases/1-parse/state/element.js
+++ b/packages/svelte/src/compiler/phases/1-parse/state/element.js
@@ -14,7 +14,8 @@ import { get_attribute_expression, is_expression_attribute } from '../../../util
 import { closing_tag_omitted } from '../../../../html-tree-validation.js';
 import { list } from '../../../utils/string.js';
 
-const regex_invalid_unquoted_attribute_value = /^(\/>|[\s"'=<>`])/;
+const regex_invalid_unquoted_attribute_value_or_self_closing_tag = /^(\/>|[\s"'=<>`])/;
+const regex_invalid_unquoted_attribute_value = /^[\s"'=<>`]/;
 const regex_closing_textarea_tag = /^<\/textarea(\s[^>]*)?>/i;
 const regex_closing_comment = /-->/;
 const regex_whitespace_or_slash_or_closing_tag = /(\s|\/|>)/;
@@ -634,9 +635,15 @@ function read_attribute_value(parser) {
 	try {
 		value = read_sequence(
 			parser,
-			() => {
+			/** @param {number} chunk_length */
+			(chunk_length) => {
 				// handle common case of quote marks existing outside of regex for performance reasons
-				if (quote_mark) return parser.match(quote_mark);
+				if (quote_mark) return !!parser.match(quote_mark);
+
+				// if chunk is not empty, handle possibility of self-closing tag
+				if (chunk_length > 0) {
+					return !!parser.match_regex(regex_invalid_unquoted_attribute_value_or_self_closing_tag);
+				}
 				return !!parser.match_regex(regex_invalid_unquoted_attribute_value);
 			},
 			'in attribute value'
@@ -670,7 +677,7 @@ function read_attribute_value(parser) {
 
 /**
  * @param {Parser} parser
- * @param {() => boolean} done
+ * @param {(value: number) => boolean} done
  * @param {string} location
  * @returns {any[]}
  */
@@ -700,7 +707,7 @@ function read_sequence(parser, done, location) {
 	while (parser.index < parser.template.length) {
 		const index = parser.index;
 
-		if (done()) {
+		if (done(current_chunk.raw.length)) {
 			flush(parser.index);
 			return chunks;
 		} else if (parser.eat('{')) {
diff --git a/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/input.svelte b/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/input.svelte
index 4bab0df72f3e..959042384e69 100644
--- a/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/input.svelte
+++ b/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/input.svelte
@@ -1 +1,2 @@
-<div class=foo></div>
\ No newline at end of file
+<div class=foo></div>
+<a href=/>home</a>
\ No newline at end of file
diff --git a/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/output.json b/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/output.json
index 5df4d66ab668..554da4827775 100644
--- a/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/output.json
+++ b/packages/svelte/tests/parser-legacy/samples/attribute-unquoted/output.json
@@ -2,12 +2,12 @@
 	"html": {
 		"type": "Fragment",
 		"start": 0,
-		"end": 21,
+		"end": 40,
 		"children": [
 			{
 				"type": "Element",
 				"start": 0,
-				"end": 21,
+				"end": 40,
 				"name": "div",
 				"attributes": [
 					{
@@ -27,7 +27,46 @@
 					}
 				],
 				"children": []
+			},
+			{
+				"type": "Text",
+				"start": 21,
+				"end": 22,
+				"raw": "\n",
+				"data": "\n"
+			},
+			{
+				"type": "Element",
+				"start": 22,
+				"end": 40,
+				"name": "a",
+				"attributes": [
+					{
+						"type": "Attribute",
+						"start": 25,
+						"end": 31,
+						"name": "href",
+						"value": [
+							{
+								"start": 30,
+								"end": 31,
+								"type": "Text",
+								"raw": "/",
+								"data": "/"
+							}
+						]
+					}
+				],
+				"children": [
+					{
+						"type": "Text",
+						"start": 32,
+						"end": 36,
+						"raw": "home",
+						"data": "home"
+					}
+				]
 			}
 		]
 	}
-}
+}
\ No newline at end of file