← All changes | includes/seo/class-seo-analyzer-fixer.php +150 -12 2.4.0 → 2.12.0 View file →

		@@ -72,8 +72,19 @@
72	72	'image_alt_text' => [
73	73	'label' => __('Fill in missing image alt text', 'thinkrank'),
74	74	'warning' => __('Runs in batches over your media library. With AI alt text enabled this uses your AI provider key.', 'thinkrank'),
75	75	],
	76	+ 'ai_crawler_access' => [
	77	+ 'label' => __('Allow AI answer engines to crawl this site', 'thinkrank'),
	78	+ // Only the answer engines are unblocked. Saying so matters:
	79	+ // people block the training crawlers on purpose, and a fix
	80	+ // that quietly reversed that decision would be a betrayal.
	81	+ 'warning' => __('Sets the answer-engine crawlers (ChatGPT, Claude, Perplexity, Google-Extended) to Allow. Crawlers that only collect training data keep whatever setting you gave them.', 'thinkrank'),
	82	+ ],
	83	+ 'llms_txt' => [
	84	+ 'label' => __('Publish llms.txt', 'thinkrank'),
	85	+ 'warning' => __('Publishes /llms.txt from your saved LLMs.txt settings. It needs those fields filled in first.', 'thinkrank'),
	86	+ ],
76	87	];
77	88
78	89	/**
79	90	* Filter the checks the analyzer can fix automatically.
		@@ -118,8 +129,12 @@
118	129	case 'schema':
119	130	return $this->fix_schema();
120	131	case 'image_alt_text':
121	132	return $this->fix_image_alt_text();
	133	+ case 'ai_crawler_access':
	134	+ return $this->fix_ai_crawler_access();
	135	+ case 'llms_txt':
	136	+ return $this->fix_llms_txt();
122	137	}
123	138
124	139	/**
125	140	* Let a third party handle a fix it registered.
		@@ -285,25 +300,148 @@
285	300	],
286	301	];
287	302	}
288	303
	304	+ $message = $remaining > 0
	305	+ ? sprintf(
	306	+ /* translators: 1: images updated in this batch, 2: images still to process. */
	307	+ __('Added alt text to %1$d images. %2$d still to go — run the fix again to continue.', 'thinkrank'),
	308	+ $updated,
	309	+ $remaining
	310	+ )
	311	+ : sprintf(
	312	+ /* translators: %d: number of images updated. */
	313	+ __('Added alt text to %d images. Your media library is done.', 'thinkrank'),
	314	+ $updated
	315	+ );
	316	+
	317	+ // The write is done, but a cache ThinkRank cannot reach still decides
	318	+ // what the visitor sees. Saying so beats a green toast over a page
	319	+ // that has not changed (#763).
	320	+ $warnings = (array) ($result['cache']['warnings'] ?? []);
	321	+ if ([] !== $warnings) {
	322	+ $message .= ' ' . implode(' ', $warnings);
	323	+ }
	324	+
289	325	return [
290	326	'fixed' => true,
291		- 'message' => $remaining > 0
292		- ? sprintf(
293		- /* translators: 1: images updated in this batch, 2: images still to process. */
294		- __('Added alt text to %1$d images. %2$d still to go — run the fix again to continue.', 'thinkrank'),
295		- $updated,
296		- $remaining
297		- )
298		- : sprintf(
299		- /* translators: %d: number of images updated. */
300		- __('Added alt text to %d images. Your media library is done.', 'thinkrank'),
301		- $updated
302		- ),
	327	+ 'message' => $message,
303	328	'data' => [
304	329	'updated' => $updated,
305	330	'remaining' => $remaining,
306	331	],
	332	+ ];
	333	+ }
	334	+
	335	+ /**
	336	+ * Set every AI answer-engine crawler to `allow`.
	337	+ *
	338	+ * Deliberately narrow: only the crawlers that decide whether the site can
	339	+ * be CITED are touched. GPTBot, ClaudeBot, CCBot and the other
	340	+ * training-only agents keep whatever the user chose for them — blocking
	341	+ * those is an editorial position, not a misconfiguration.
	342	+ *
	343	+ * @since 2.5.0
	344	+ * @return array
	345	+ * @throws \Exception When the rules cannot be saved, or when something
	346	+ * other than the rule map is still blocking them.
	347	+ */
	348	+ private function fix_ai_crawler_access(): array {
	349	+ $manager = new Site_Identity_Manager();
	350	+ $settings = $manager->get_settings('site');
	351	+ $settings = is_array($settings) ? $settings : [];
	352	+
	353	+ $rules = isset($settings['ai_crawler_rules']) && is_array($settings['ai_crawler_rules'])
	354	+ ? $settings['ai_crawler_rules']
	355	+ : [];
	356	+
	357	+ foreach (SEO_Analyzer::GEO_ANSWER_AGENTS as $slug) {
	358	+ if (AI_Crawlers::exists($slug)) {
	359	+ $rules[$slug] = 'allow';
	360	+ }
	361	+ }
	362	+
	363	+ $settings['ai_crawler_rules'] = $rules;
	364	+
	365	+ if (!$manager->save_settings('site', null, $settings)) {
	366	+ throw new \Exception(esc_html__('The AI crawler rules could not be saved. Check the PHP error log for the ThinkRank line naming the cause.', 'thinkrank'));
	367	+ }
	368	+
	369	+ // The directives are composed into the served body at render time, but
	370	+ // a physical robots.txt in the web root is a copy the web server hands
	371	+ // out directly — saving the rules does not touch it. Re-sync from a
	372	+ // fresh manager (the one above holds the pre-save settings) or the fix
	373	+ // changes only the file nobody is reading.
	374	+ (new Site_Identity_Manager())->sync_robots_txt_file();
	375	+
	376	+ // The rules are only one of the things that can disallow a crawler: a
	377	+ // hand-written robots.txt body, a physical robots.txt in the web root,
	378	+ // and the site-wide search block all outrank them. Ask the served file
	379	+ // again rather than reporting a success it contradicts.
	380	+ $still_blocked = (new SEO_Analyzer())->blocked_answer_agents();
	381	+
	382	+ if (!empty($still_blocked)) {
	383	+ throw new \Exception(
	384	+ sprintf(
	385	+ /* translators: %s: comma-separated crawler names. */
	386	+ esc_html__('The crawler rules were saved, but your robots.txt still blocks %s. That block comes from your own robots.txt content, a robots.txt file in your site root, or the site-wide search-engine setting — edit it under Essential SEO → Crawling and AI Indexing.', 'thinkrank'),
	387	+ esc_html(implode(', ', $still_blocked))
	388	+ )
	389	+ );
	390	+ }
	391	+
	392	+ return [
	393	+ 'fixed' => true,
	394	+ 'message' => __('AI answer engines can now crawl and cite your site.', 'thinkrank'),
	395	+ 'data' => [],
	396	+ ];
	397	+ }
	398	+
	399	+ /**
	400	+ * Generate and publish /llms.txt from the saved LLMs.txt settings.
	401	+ *
	402	+ * The document is built from settings the user already wrote, so this is a
	403	+ * publish, not authorship. Incomplete settings produce an invalid document
	404	+ * and are refused rather than written — a half-empty llms.txt is worse
	405	+ * than none, because an assistant will read it and believe it.
	406	+ *
	407	+ * @since 2.5.0
	408	+ * @return array
	409	+ * @throws \Exception When the module is unavailable, the settings are
	410	+ * incomplete, or the write fails.
	411	+ */
	412	+ private function fix_llms_txt(): array {
	413	+ if (!class_exists('ThinkRank\\SEO\\LLMs_Txt_Manager')) {
	414	+ throw new \Exception(esc_html__('The llms.txt module is unavailable.', 'thinkrank'));
	415	+ }
	416	+
	417	+ $manager = new LLMs_Txt_Manager();
	418	+ $generated = $manager->generate_llms_txt([], []);
	419	+
	420	+ $validation = isset($generated['validation']) && is_array($generated['validation'])
	421	+ ? $generated['validation']
	422	+ : [];
	423	+
	424	+ if (empty($validation['valid'])) {
	425	+ throw new \Exception(esc_html__('llms.txt could not be published because its settings are incomplete. Fill in Website Description, Key Features and Target Audience under Essential SEO → Crawling and AI Indexing → LLMs.txt, then try again.', 'thinkrank'));
	426	+ }
	427	+
	428	+ $content = (string) ($generated['content'] ?? '');
	429	+ $write = $manager->write_llms_txt_to_file($content);
	430	+
	431	+ if (empty($write['success'])) {
	432	+ $reason = (string) ($write['message'] ?? '');
	433	+
	434	+ throw new \Exception(
	435	+ '' !== $reason
	436	+ ? esc_html($reason)
	437	+ : esc_html__('llms.txt could not be written. Check that your site root is writable.', 'thinkrank')
	438	+ );
	439	+ }
	440	+
	441	+ return [
	442	+ 'fixed' => true,
	443	+ 'message' => __('Your llms.txt is now published at /llms.txt.', 'thinkrank'),
	444	+ 'data' => [],
307	445	];
308	446	}
309	447	}