| 开发者 | ksanyok |
|---|---|
| 更新时间 | 2026年8月16日 18:29 |
| PHP版本: | 7.4 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
[placeholder] text, facts from your content only./wp-content/plugins/ directory.Yes. The plugin connects to the RankBotAI service API to generate SEO suggestions and optimizations. A free tier with monthly free credits is available, and no credit card is required to start.
The plugin is free and GPL-licensed. The RankBotAI service provides a free credit tier and offers paid plans for additional usage and higher limits.
Yes! RankBotAI is specifically designed to complement Rank Math and Yoast SEO. It reads existing SEO data from these plugins and writes optimized values back into their own fields — so your SEO data stays exactly where it was, and nothing is lost if you deactivate RankBotAI.
Yes. RankBotAI works standalone and stores SEO data in its own meta fields. However, for the best SEO results, we recommend pairing it with Rank Math or Yoast SEO.
Only the items you select. Bulk Optimization has filters such as "only items without a focus keyword" and "only unprocessed items", so you can target the gaps and leave your hand-written work untouched. And every change is snapshotted before it is applied.
Yes — that is the point of the Backups & Restore page. You can roll back a single post or product from the History log, roll back one content type, or roll back the entire site to its pre-RankBot state.
Yes. Products, product categories and product attributes are supported, including a dedicated "RankBot AI" tab inside the Product Data box. Bulk optimization of a full catalog is the most common use case.
Yes. Any public taxonomy term — categories, tags, product categories, custom taxonomies — can be optimized individually or in bulk, with the same backup and restore guarantees.
The same language as the source content. The AI detects the language of each post or product and writes in it. WPML and Polylang sites are detected per post.
Yes. Global Instructions (Settings) let you set site-wide rules that are injected into every generation — for example "never promise free shipping", "always mention that packaging may vary", or "describe only the product in the title".
Most single-page optimizations complete in 5–15 seconds. Bulk optimization runs in the background at roughly 700 items/hour on a healthy connection, so 100+ pages typically finish within 15–30 minutes.
No. All AI processing happens on the RankBotAI service and in background WP-Cron jobs in wp-admin. On the frontend the plugin only outputs meta tags and JSON-LD structured data.
It sends only the data needed for the action you trigger. This can include your site URL, environment info (WordPress/plugin version), and the post/product/term content you request to optimize.
RankBotAI uses its own RankBot AI engine — an advanced multi-model system with automatic failover. You don't need to choose or configure any models. The system automatically selects the best available model for each request.
Only if you explicitly enable content optimization in settings. By default, RankBotAI only updates SEO meta fields (title, description, focus keyword). Content changes require opt-in.
llms.txt is a plain-text map of your site aimed at AI crawlers (ChatGPT, Perplexity, AI Overviews). It helps them understand and cite your pages correctly. RankBotAI generates and manages it for you.
%i table-identifier placeholders in $wpdb->prepare(), which were introduced in WordPress 6.2 — on older versions those queries silently returned nothing.str_contains() / str_starts_with(), which crashed the site with "Call to undefined function". Polyfills added.rankbot_ai_force_schema_output filter is available to restore the old behaviour.</script> could break out of the structured-data block and inject markup into the page head.state value generated by this site, so an activation link crafted by someone else can no longer silently re-bind the site to another RankBotAI account.wp_head, i.e. after headers were already sent, so it was never actually stored and visitors were re-randomised on every pageview. The variant is now resolved on template_redirect.<meta name="description"> next to Rank Math / Yoast. With an SEO plugin active, the variant is applied through that plugin's own description filter, so the test actually takes effect.fonts.googleapis.com. No admin IP addresses are shared with Google any more, and wp-admin loads one render-blocking request less. The UI falls back to the system font stack that the rest of the plugin already used.wp_json_encode(), which sanity-checks UTF-8 — content with broken encoding no longer produces an empty request body.wp_rand().payments / user_subscriptions tables predated the gateway and billing-period columns could not pay at all — every top-up failed with "Unknown column 'billing_period'". The missing columns are now migrated automatically.gemini-2.0-flash-001 model — the provider had started answering 404 "no endpoints" — with gemini-2.5-flash.seo_title / seo_description / keyword result aliases)..mo / .l10n.php from shadowing fresh strings.jobs and site_operations payloads (cron) — heavy AI request/response data is pruned after 30 days, preventing database bloat.jobs table if usage_logs is empty, fixing counter display for sites with historical data.usage_logs billing model yet.usage_logs) to avoid mismatches./api/v1/history stats no longer inflate "processed" by counting internal lifecycle events from site_operations.period_7_ops, period_7_spent, period_30_ops, period_30_spent) for accurate plugin period cards.GET /api/v1/site/custom-prompt and POST /api/v1/site/custom-prompt (Bearer-auth, per-site).GET /api/v1/runtime-config — exposes the active AI pool mode (paid_first / free_with_paid_fallback / free_only) and the matching dispatch profile (parallel jobs, polling cadence, stuck-job timeouts). The bulk runner can read this on the fly to scale itself up/down without a plugin update.CLIENT'S MANDATORY GLOBAL INSTRUCTIONS block at the very top of the system prompt, with explicit "must override defaults" phrasing.Database::getInstance() typo in OpenAIService was silently falling back to default mode — pool mode and active models are now actually read from DB.Cache-Control header on /api/v1/runtime-config switched to no-store so mode changes propagate to WP within seconds (CDN was caching the old JSON).