SEO: fix FAQPage schema to read the question from saved <summary> markup

projects/packages/seo/changelog/fix-seo-faq-schema-summary

@@ -0,0 +1,4 @@
+Significance: patch
+Type: fixed
+
+FAQ schema: read the question from the saved `<summary>` markup so FAQPage JSON-LD emits from real editor-saved Details blocks.

projects/packages/seo/src/class-post-schema-node.php

@@ -128,15 +128,15 @@ private static function build_faq( WP_Post $post ) {
 			if ( 'core/details' !== ( $block['blockName'] ?? '' ) ) {
 				continue;
 			}
-			$question = trim( (string) ( $block['attrs']['summary'] ?? '' ) );
+			$question = self::question_from_details_block( $block );
 
 			// Render only the inner blocks for the answer. Rendering the whole
 			// core/details block would re-include the <summary> (the question).
 			$answer_html = '';
 			foreach ( $block['innerBlocks'] ?? array() as $inner_block ) {
 				$answer_html .= render_block( $inner_block );
 			}
-			$answer = trim( wp_strip_all_tags( $answer_html ) );
+			$answer = self::to_plain_text( $answer_html );
 			if ( '' === $question || '' === $answer ) {
 				continue;
 			}
@@ -159,4 +159,37 @@ private static function build_faq( WP_Post $post ) {
 			'mainEntity' => $items,
 		);
 	}
+
+	/**
+	 * Extract the question text from a `core/details` block's `<summary>`.
+	 *
+	 * The Details block declares `summary` as a `source: "rich-text"` attribute,
+	 * so the value is saved in the `<summary>…</summary>` markup, not in the
+	 * `<!-- wp:details … -->` comment. `parse_blocks()` does not resolve
+	 * source-based attributes (it only returns what's written into the comment),
+	 * so `$block['attrs']['summary']` is always empty for real, editor-saved
+	 * blocks. The summary text does survive in the block's inner HTML, so we read
+	 * it from there instead.
+	 *
+	 * @param array $block A parsed `core/details` block.
+	 * @return string The plain-text question, or '' when the block has no summary.
+	 */
+	private static function question_from_details_block( array $block ) {
+		$inner_html = (string) ( $block['innerHTML'] ?? '' );
+		if ( ! preg_match( '#<summary\b[^>]*>(.*?)</summary>#is', $inner_html, $matches ) ) {
+			return '';
+		}
+		return self::to_plain_text( $matches[1] );
+	}
+
+	/**
+	 * Reduce a fragment of post HTML to the plain text used for a schema value:
+	 * tags stripped, entities decoded, surrounding whitespace trimmed.
+	 *
+	 * @param string $html HTML fragment.
+	 * @return string
+	 */
+	private static function to_plain_text( $html ) {
+		return trim( html_entity_decode( wp_strip_all_tags( (string) $html ), ENT_QUOTES, 'UTF-8' ) );
+	}
 }

projects/packages/seo/tests/php/PostSchemaNodeTest.php

@@ -131,13 +131,18 @@ public function test_published_page_has_no_default_schema() {
 	}
 
 	/**
-	 * FAQPage answers are built from a `core/details` block's inner blocks only,
-	 * so the question (the `<summary>`) is not duplicated into the answer text.
+	 * The FAQ question is read from the saved `<summary>` markup, and the answer
+	 * from the inner blocks only — so the question is not duplicated into the
+	 * answer text. The fixture uses realistic, editor-saved markup: the summary
+	 * lives only in `<summary>` (a `source: "rich-text"` attribute), never in the
+	 * `<!-- wp:details -->` comment, which is what `parse_blocks()` actually
+	 * returns. Baking the summary into the comment instead would mask the bug
+	 * this builder has to handle (see JETPACK-1793).
 	 */
-	public function test_faq_answer_excludes_the_question() {
+	public function test_faq_question_from_summary_and_answer_excludes_it() {
 		\Jetpack_SEO_Posts::$schema_type = 'faq';
 
-		$content  = '<!-- wp:details {"summary":"What is SEO?"} -->';
+		$content  = '<!-- wp:details -->';
 		$content .= '<details class="wp-block-details"><summary>What is SEO?</summary>';
 		$content .= '<!-- wp:paragraph --><p>Search engine optimization.</p><!-- /wp:paragraph -->';
 		$content .= '</details><!-- /wp:details -->';
@@ -155,6 +160,46 @@ public function test_faq_answer_excludes_the_question() {
 		$this->assertStringNotContainsString( 'What is SEO?', $item['acceptedAnswer']['text'] );
 	}
 
+	/**
+	 * The summary is rich text, so it may carry inline formatting and HTML
+	 * entities. The question must be reduced to decoded plain text, and multiple
+	 * Details blocks each become their own Question entity in document order.
+	 */
+	public function test_faq_summary_is_decoded_plain_text_across_multiple_blocks() {
+		\Jetpack_SEO_Posts::$schema_type = 'faq';
+
+		$content  = '<!-- wp:details -->';
+		$content .= '<details class="wp-block-details"><summary>What is <strong>SEO</strong> &amp; AEO?</summary>';
+		$content .= '<!-- wp:paragraph --><p>Optimization for search and answer engines.</p><!-- /wp:paragraph -->';
+		$content .= '</details><!-- /wp:details -->';
+		$content .= '<!-- wp:details -->';
+		$content .= '<details class="wp-block-details"><summary>Is it free?</summary>';
+		$content .= '<!-- wp:paragraph --><p>Yes.</p><!-- /wp:paragraph -->';
+		$content .= '</details><!-- /wp:details -->';
+
+		$node = Post_Schema_Node::build( $this->make_post( array( 'post_content' => $content ) ) );
+
+		$this->assertIsArray( $node );
+		$this->assertCount( 2, $node['mainEntity'] );
+		$this->assertSame( 'What is SEO & AEO?', $node['mainEntity'][0]['name'] );
+		$this->assertSame( 'Is it free?', $node['mainEntity'][1]['name'] );
+	}
+
+	/**
+	 * A Details block whose `<summary>` is empty produces no question, so it is
+	 * skipped rather than emitting a Question with a blank name.
+	 */
+	public function test_faq_skips_details_block_with_empty_summary() {
+		\Jetpack_SEO_Posts::$schema_type = 'faq';
+
+		$content  = '<!-- wp:details -->';
+		$content .= '<details class="wp-block-details"><summary></summary>';
+		$content .= '<!-- wp:paragraph --><p>An answer with no question.</p><!-- /wp:paragraph -->';
+		$content .= '</details><!-- /wp:details -->';
+
+		$this->assertNull( Post_Schema_Node::build( $this->make_post( array( 'post_content' => $content ) ) ) );
+	}
+
 	/**
 	 * A "faq" override with no `core/details` blocks yields no node, rather than
 	 * an empty/invalid FAQPage.

projects/packages/seo/tests/php/SchemaBuilderTest.php

@@ -195,7 +195,7 @@ public function test_emits_graph_with_article_for_published_post() {
 	public function test_emits_graph_with_faqpage_for_faq_override() {
 		\Jetpack_SEO_Posts::$schema_type = 'faq';
 
-		$content  = '<!-- wp:details {"summary":"What is SEO?"} -->';
+		$content  = '<!-- wp:details -->';
 		$content .= '<details class="wp-block-details"><summary>What is SEO?</summary>';
 		$content .= '<!-- wp:paragraph --><p>Search engine optimization.</p><!-- /wp:paragraph -->';
 		$content .= '</details><!-- /wp:details -->';
@@ -235,7 +235,7 @@ public function test_faqpage_has_no_publisher_but_graph_has_organization() {
 		$this->set_site_name( 'Acme Co' );
 		\Jetpack_SEO_Posts::$schema_type = 'faq';
 
-		$content  = '<!-- wp:details {"summary":"What is SEO?"} -->';
+		$content  = '<!-- wp:details -->';
 		$content .= '<details class="wp-block-details"><summary>What is SEO?</summary>';
 		$content .= '<!-- wp:paragraph --><p>Search engine optimization.</p><!-- /wp:paragraph -->';
 		$content .= '</details><!-- /wp:details -->';