diff --git a/packages/rich-text/CHANGELOG.md b/packages/rich-text/CHANGELOG.md
index 36c9d8850351a0..13bca1b68bc7ce 100644
--- a/packages/rich-text/CHANGELOG.md
+++ b/packages/rich-text/CHANGELOG.md
@@ -2,6 +2,10 @@
## Unreleased
+### Bug Fixes
+
+- `toHTMLString`: Close non-void inline objects and preserve their children ([#85895](https://github.com/WordPress/gutenberg/pull/85895)).
+
## 8.2.0 (2026-11-18)
## 8.1.0 (2026-11-04)
diff --git a/packages/rich-text/src/test/to-html-string-objects.jsdom.test.ts b/packages/rich-text/src/test/to-html-string-objects.jsdom.test.ts
new file mode 100644
index 00000000000000..ccd3743a4f536e
--- /dev/null
+++ b/packages/rich-text/src/test/to-html-string-objects.jsdom.test.ts
@@ -0,0 +1,135 @@
+import { afterEach, describe, expect, it } from 'vitest';
+import '../store';
+import { create } from '../create';
+import { applyFormat } from '../apply-format';
+import { insertObject } from '../insert-object';
+import { registerFormatType } from '../register-format-type';
+import { unregisterFormatType } from '../unregister-format-type';
+import { toHTMLString } from '../to-html-string';
+import { OBJECT_REPLACEMENT_CHARACTER } from '../special-characters';
+
+const FORMAT_NAME = 'test/serialization-object';
+const CLASS_NAME = 'test-serialization-object';
+
+function registerObjectFormat( tagName: string, contentEditable?: boolean ) {
+ registerFormatType( FORMAT_NAME, {
+ title: 'Serialization object',
+ tagName,
+ className: CLASS_NAME,
+ object: true,
+ contentEditable,
+ } );
+}
+
+describe( 'toHTMLString object elements', () => {
+ afterEach( () => {
+ unregisterFormatType( FORMAT_NAME );
+ } );
+
+ it.each( [ 'span', 'a', 'custom-object', 'SPAN' ] )(
+ 'closes a non-void %s object before following text',
+ ( tagName ) => {
+ registerObjectFormat( tagName );
+ const value = insertObject(
+ create( { text: 'before after' } ),
+ { type: FORMAT_NAME },
+ 7,
+ 7
+ );
+ const html = toHTMLString( { value } );
+
+ expect( html ).toBe(
+ `before <${ tagName } class="${ CLASS_NAME }">${ tagName }>after`
+ );
+
+ const element = document.createElement( 'div' );
+ element.innerHTML = html;
+ expect( element.querySelector( tagName )?.textContent ).toBe( '' );
+ expect( element.lastChild?.textContent ).toBe( 'after' );
+ }
+ );
+
+ it( 'round-trips an empty object with registered attributes', () => {
+ registerFormatType( FORMAT_NAME, {
+ title: 'Inline token',
+ tagName: 'span',
+ className: CLASS_NAME,
+ object: true,
+ attributes: { id: 'data-id' },
+ } );
+ const value = insertObject(
+ create( { text: 'before after' } ),
+ { type: FORMAT_NAME, attributes: { id: 'sample' } },
+ 7,
+ 7
+ );
+ const html = toHTMLString( { value } );
+
+ expect( html ).toBe(
+ `before after`
+ );
+ const parsed = create( { html } );
+
+ expect( parsed.text ).toBe( value.text );
+ expect( parsed.replacements[ 7 ] ).toMatchObject( {
+ type: FORMAT_NAME,
+ attributes: { id: 'sample' },
+ } );
+ } );
+
+ it( 'preserves the children of a non-editable object and the following text', () => {
+ registerObjectFormat( 'span', false );
+ const html = `before Tokenafter`;
+ const value = create( { html } );
+
+ expect( value.text ).toBe(
+ `before ${ OBJECT_REPLACEMENT_CHARACTER }after`
+ );
+ const serialized = toHTMLString( { value } );
+
+ expect( serialized ).toBe( html );
+ expect( create( { html: serialized } ).text ).toBe( value.text );
+ } );
+
+ it.each( [
+ 'img',
+ 'br',
+ 'input',
+ 'IMG',
+ 'param',
+ 'keygen',
+ 'frame',
+ 'basefont',
+ 'bgsound',
+ ] )( 'keeps a void %s object without a closing tag', ( tagName ) => {
+ registerObjectFormat( tagName );
+ const value = insertObject(
+ create( { text: 'before after' } ),
+ { type: FORMAT_NAME },
+ 7,
+ 7
+ );
+
+ expect( toHTMLString( { value } ) ).toBe(
+ `before <${ tagName } class="${ CLASS_NAME }">after`
+ );
+ } );
+
+ it( 'preserves text in a non-object format registered with a void tag name', () => {
+ registerFormatType( FORMAT_NAME, {
+ title: 'Image wrapper',
+ tagName: 'img',
+ className: CLASS_NAME,
+ } );
+ const value = applyFormat(
+ create( { text: 'selected' } ),
+ { type: FORMAT_NAME },
+ 0,
+ 8
+ );
+
+ expect( toHTMLString( { value } ) ).toBe(
+ `
selected`
+ );
+ } );
+} );
diff --git a/packages/rich-text/src/to-html-string.ts b/packages/rich-text/src/to-html-string.ts
index 937d6a3315ae0d..929027415ddccb 100644
--- a/packages/rich-text/src/to-html-string.ts
+++ b/packages/rich-text/src/to-html-string.ts
@@ -6,6 +6,28 @@ import {
import { toTree } from './to-tree';
import type { HTMLTreeNode, RichTextValue } from './types';
+const VOID_ELEMENTS = new Set( [
+ 'area',
+ 'base',
+ 'br',
+ 'col',
+ 'embed',
+ 'hr',
+ 'img',
+ 'input',
+ 'link',
+ 'meta',
+ 'source',
+ 'track',
+ 'wbr',
+ // Browsers also keep these obsolete elements void when parsing HTML.
+ 'basefont',
+ 'bgsound',
+ 'frame',
+ 'keygen',
+ 'param',
+] );
+
/**
* Create an HTML string from a Rich Text value.
*
@@ -146,7 +168,7 @@ function createElementHTML( {
) }"`;
}
- if ( object ) {
+ if ( object && type && VOID_ELEMENTS.has( type.toLowerCase() ) ) {
return `<${ type }${ attributeString }>`;
}