內容簡介
Royal AI Firewall 外掛讓使用者能夠掌控進入其 WordPress 網站的 AI 機器人，提供即時監控和管理功能，讓使用者可以選擇允許或封鎖特定的 AI 機器人，保護網站內容不被未經授權使用。
【主要功能】
• 實時儀表板顯示過去 24 小時的 AI 機器人訪問情況
• 每個機器人可選擇允許、封鎖或僅記錄
• 一鍵封鎖所有 AI 機器人的緊急按鈕
• 首次設置向導自動檢測 CDN 設定
• 兼容 GuardPress 和其他安全外掛
• 定期更新的機器人指紋目錄
外掛標籤
開發者團隊
原文外掛簡介
Every AI company is scraping WordPress sites to train models, feed AI search engines, and answer questions using your content — usually without your knowledge and rarely with a link back. Royal AI Firewall gives you agency at the WordPress layer: see every AI agent hitting your site, and decide who gets through with one-click per-bot policies.
Whether you want to allow AI search engines but block training crawlers, log everything for a month before deciding anything, or block every AI bot with one click — this plugin gives you the visibility and enforcement to make that choice.
This plugin gives you:
A live dashboard of which AI agents have visited your site in the last 24 hours
A per-bot dropdown to allow, block, or log-only any of 60+ recognized AI bots
A master “Block all AI bots” panic button on every dashboard load
A first-run setup wizard that detects your CDN (Cloudflare, Bunny CDN, Fastly, KeyCDN, Sucuri, StackPath, or Akamai) and, when it finds Cloudflare, tells you exactly which CF settings to dial down so this plugin can take over the AI-bot layer
Compatibility detection for GuardPress and other popular security plugins
A bundled bot fingerprint catalog that refreshes on every plugin update, with an optional opt-in to fetch fresher catalogs daily from fingerprints.royalplugins.com (see External Services below)
Free, Self-Hosted, Fully Featured
Royal AI Firewall is fully featured in its free, GPL-licensed release. There is no Pro version — every feature ships in the wp.org plugin, and updates go through the standard WordPress plugin updater.
Your data stays on your server. The plugin never sends your site’s traffic, customer data, IP addresses, or credentials to any third party. See the External Services section below for the full list of endpoints the plugin can contact, when, and how to disable each one.
AI Bots Recognized (69 as of v1.0.6)
The bundled catalog covers the major AI bot families. Each entry includes the bot’s owner, intended purpose, default policy, and the blocking consequences (for example, “blocking GPTBot may remove your site from ChatGPT search results”).
Training crawlers: GPTBot, ClaudeBot, anthropic-ai, Bytespider, TikTokSpider, FacebookBot, Meta-ExternalAgent, GoogleOther, GoogleOther-AI, Google-Extended, Google-CloudVertexBot, MistralBot, KimiBot, cohere-ai, cohere-training-data-crawler, ai2bot, ai2bot-dolma, Amazonbot, PetalBot
Retrieval bots (on-demand): ChatGPT-User, OAI-AdsBot, ClaudeBot-User, Claude-Web, claude-code, Perplexity-User, Kimi-User, YandexAdditionalBot, Meta-ExternalFetcher, facebookexternalhit, APIs-Google
AI search engines: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Kimi-SearchBot, MistralAI-Index, YandexAdditional, meta-webindexer, Applebot-Extended, MicrosoftCopilotBot, DuckAssistBot, YouBot, PhindBot, iAsk, Komo, Liner, Brave Leo, Andi
Search engines (always-allow guarded): Googlebot, Googlebot-Image, Googlebot-Video, Googlebot-News, Google-InspectionTool, Bingbot, BingPreview, Applebot, DuckDuckBot
Other search engines: Baiduspider
Agent browsers (newer category): OperatorAgent, ChatGPT-Atlas, Claude-Computer-Use
Dataset scrapers: CCBot (Common Crawl), Diffbot, ImagesiftBot, Omgilibot, Timpibot
Other Google crawlers: Storebot-Google, Mediapartners-Google, AdsBot-Google, adidxbot
The Dashboard
Open the AI Firewall menu in your WordPress admin to see:
A hero metric — total AI bot hits in the last 24 hours and the number of distinct bots involved
A per-bot list with hit count, bandwidth used, and a one-click policy dropdown for each row
An MCP / Abilities API activity widget when an MCP server plugin (Royal MCP or any plugin implementing the WordPress Abilities API) is detected on your site
A Cloudflare visibility status card with an honest estimate of how many AI bots may have been filtered by Cloudflare at the edge before reaching WordPress
Click any bot row to expand a drill-down view: top URLs the bot hit, recent activity, and what blocking the bot would cost you
Per-Bot Policy Controls
Each recognized bot row has a dropdown with four options:
Use default policy — falls back to your global mode (Log only, Block training, or Block all)
Always allow — bot is allowed regardless of default mode
Log only — bot is allowed and recorded; never blocked
Block — bot receives a 403 response immediately, before WordPress runs any heavy work
Major search engines (Googlebot, Bingbot, Applebot, DuckDuckBot) are protected from accidental blocking. The per-bot dropdown is disabled for these bots, and the API rejects block requests for them unless you explicitly enable the “Search engine override” toggle in Settings — with a clear warning that blocking Googlebot removes your site from Google Search.
Cloudflare Compatibility
If your site is behind Cloudflare, the setup wizard’s Cloudflare screen tells you exactly which CF settings to turn off so this plugin can take over the AI-bot layer:
AI Audit → set to “Allow”
AI Labyrinth → OFF
Custom WAF rules blocking AI bots → DELETE (the per-bot controls in this plugin replace them)
Security Level → Medium or Low
And which CF settings to leave on (they don’t conflict):
DDoS protection
Managed WAF rules
SSL/TLS
Bot Fight Mode (basic tier)
Browser Integrity Check
Caching
The dashboard also detects Cloudflare on every admin page load (looking for cf-ray, cf-connecting-ip, or CDN-Loop: cloudflare headers) and shows a status card with the detection state. A persistent 24-hour state ensures the UI stays stable even when an occasional admin request doesn’t pass through CF.
Other CDN Compatibility
The setup wizard also recognizes 6 additional CDNs by their vendor-forwarded request headers: Bunny CDN, Fastly, KeyCDN, Sucuri, StackPath, and Akamai. Detection is header-sniff only — no outbound HTTP, no DNS lookups.
When any of these are detected, the wizard shows a compatibility note rather than a network-specific dial-down (Cloudflare remains the only CDN with a full walkthrough because its AI controls are the most common source of operator confusion). Royal AI Firewall sees AI bot traffic that reaches WordPress regardless of which CDN sits in front — if you have edge-side AI-bot rules configured on your CDN, consult its documentation, as the two layers can coexist.
Other Security Plugin Compatibility
The plugin auto-detects these plugins when they’re active and shows compatibility notes on the dashboard and Settings page:
Edge-firewall security plugins — their firewalls run before WordPress loads. AI bots they block at their layer won’t appear in this plugin’s dashboard, but the two layers don’t conflict.
WordPress-layer security plugins — coexist cleanly at the WordPress layer.
GuardPress (Royal Plugins) — first-party Royal Plugins integration.
Royal MCP (Royal Plugins) — when Royal MCP is detected, MCP tool invocations from connected AI agents appear in the MCP Activity widget on the dashboard.
WordPress Abilities API & MCP Server Integration
This plugin listens for the WordPress Abilities API hooks wp_before_execute_ability and wp_after_execute_ability (WP 6.9+) and logs every ability invocation regardless of which MCP server triggers it. If you have any MCP server plugin installed and an AI agent calls an ability, you’ll see it in the MCP Activity widget on the dashboard.
If Royal MCP 1.4.33 or later is installed, an additional first-party bridge captures every MCP tool call from that server with full tool name and result status.
Search Engine Guard
Major search engines are protected from accidental blocking by default. The dashboard dropdown is disabled for Googlebot, Bingbot, Applebot, and DuckDuckBot. The REST API endpoints reject block attempts on these bots with a 409 Conflict response unless the customer has explicitly enabled the “Search engine override” toggle in Settings. The override toggle includes a clear warning that blocking Googlebot removes the site from Google Search.
Telemetry and Data
Anonymous usage data is OFF by default. The plugin makes no outbound HTTP call for telemetry unless you explicitly opt in — via the setup wizard’s final step or the “Anonymous usage data” toggle in Settings.
If you opt in, once per week the plugin POSTs a small JSON payload to telemetry.royalplugins.com. The payload contains:
Plugin version
WordPress version
PHP version
An opaque one-way hash of your site URL — used only to count unique installs. The URL itself cannot be recovered from the hash (SHA-256 with a per-install random salt that never leaves your site).
Whether you have completed the setup wizard
How many per-bot policy overrides you have configured
A bucketed range for AI bot hits in the last 24 hours (e.g. “11–50”)
The following are NEVER sent, regardless of toggle state:
Your site URL or domain (only its salted hash)
Customer email addresses
Invocation log contents
Specific IP addresses
Specific bot identities
User-Agent strings of visitors
Data retention: raw payloads are retained for 90 days. Aggregated statistics are retained indefinitely. Per-site fingerprints are purged after 12 months of no reports. You can revoke consent at any time in Settings — the plugin unschedules the weekly cron immediately.
The other outbound HTTP call the plugin can make (also opt-in) is a single daily GET request to fingerprints.royalplugins.com for a fresher bot catalog when you enable “Keep catalog updated between releases.” That request body is empty and includes only the plugin version in the User-Agent header. See the “External Services” section below.
Log retention defaults to 7 days. The retention window is filterable via raif_log_retention_days for developers who need a different value.
How Activation Works
On activation the plugin:
Creates five custom database tables: raif_invocation_log, raif_daily_rollup, raif_bot_policy, raif_bot_access_baseline, raif_page_type_daily
Seeds safe default options (Log only mode; telemetry off; uninstall data-delete off; live catalog updates off)
Schedules WP-Cron events for hourly rollup and daily log prune, both running entirely inside your WordPress install with no network calls
Loads the bundled bot fingerprint catalog from the plugin zip
Redirects the activating admin to the 4-step setup wizard
The wizard is skippable from any step
No outbound HTTP calls are made until the customer explicitly opts in to live catalog updates on the wizard’s final screen or via Settings → Bot fingerprint database. The plugin is fully functional without ever making a network call — the bundled catalog refreshes from the plugin zip on every plugin update.
On deactivation the plugin unschedules all WP-Cron events. Data is preserved by default so a re-activation continues where you left off. To remove all data on uninstall, check the “Delete all logs, tables, and options when the plugin is uninstalled” toggle in Settings → Data before deactivating.
External Services
The plugin can contact five categories of endpoint. Two are opt-in and off by default. Three run automatically to support features the plugin advertises (bot signature verification, bot-access alerting, and Content Signals). Every endpoint is disable-able. No customer data, IP address, credentials, or site traffic is ever transmitted to any third party.
You can globally block every outbound request by setting define( 'WP_HTTP_BLOCK_EXTERNAL', true ); in wp-config.php. The plugin degrades cleanly when blocked.
1. Royal AI Firewall Fingerprint Catalog (opt-in, off by default)
Endpoint: https://fingerprints.royalplugins.com/v1/index.json
When it runs: Only when the customer enables the “Keep catalog updated between releases” toggle. Off by default.
Frequency: Once per day via WordPress cron (raif_fingerprint_update).
Data sent: None. Request body is empty. User-Agent header carries the plugin version (e.g. royal-ai-firewall/X.Y.Z), plus a standard If-None-Match cache validator. No site URL, no IP address, no customer data.
Data received: A JSON catalog of recognized AI bot fingerprints, approximately 55 KB.
How to disable: Untick the toggle in Settings → Bot fingerprint database, or filter raif_fingerprint_endpoint to an empty string.
Privacy Policy: royalplugins.com/privacy/
Terms of Service: royalplugins.com/terms/
2. Anonymous Usage Telemetry (opt-in, off by default)
Endpoint: https://telemetry.royalplugins.com/v1/raif/report
When it runs: Only when the customer enables the telemetry toggle in Settings → Data.
Frequency: Once per week via WordPress cron (raif_telemetry_report).
Data sent: Plugin version, WordPress version, PHP version, a locally-salted install fingerprint hash (not derivable back to a URL), coarse-bucketed feature-usage counters. No site URL, no IP address, no bot hit content, no customer data.
How to disable: Untick the telemetry toggle in Settings → Data.
Privacy Policy: royalplugins.com/privacy/
3. Web Bot Auth Key Directories (runs automatically)
Endpoints:
https://openai.com/.well-known/http-message-signatures-directory (OpenAI)
https://www.anthropic.com/.well-known/http-message-signatures-directory (Anthropic)
https://www.perplexity.ai/.well-known/http-message-signatures-directory (Perplexity)
https://developers.google.com/.well-known/http-message-signatures-directory (Google)
https://www.facebook.com/.well-known/http-message-signatures-directory (Meta)
When it runs: Weekly via WordPress cron (raif_web_bot_auth_refresh), scheduled at activation.
Data sent: None. Request body is empty. Default WordPress wp_remote_get User-Agent. No site URL, no customer data.
Data received: Public JWK signing keys the operators publish so RAIF can verify cryptographically-signed bot requests (RFC 9421 Web Bot Auth). Keys are cached locally.
Purpose: Enables the “trust verified operator” policy tier so bots that sign their requests with published keys can be distinguished from user-agent spoofers.
How to disable: Filter raif_web_bot_auth_directories to an empty array, or block outbound HTTP globally with WP_HTTP_BLOCK_EXTERNAL.
4. Bot Access Self-Probe (runs automatically)
Endpoint: Your own site’s home URL (home_url()).
When it runs: Twice daily via WordPress cron (raif_bot_access_check), scheduled at activation.
Data sent: Two GET requests. One with a browser User-Agent, one with a Googlebot User-Agent. No body, no site data beyond the requests themselves.
Data received: HTTP status code and a truncated response body used to detect whether upstream CDNs/hosts are silently blocking real search-engine bots.
Purpose: Powers the “Bot Access” alerting that warns when your site becomes unreachable to major search bots. Zero external calls; the probe loops back to your own origin.
How to disable: Untick “Enable bot access monitoring” in Settings, or filter raif_bot_access_probe_enabled to false.
5. Content Signals robots.txt read (runs automatically)
Endpoint: Your own site’s /robots.txt (via home_url( '/robots.txt' )).
When it runs: At most once every six hours during a live bot request that requires signal evaluation. Loops back to your own origin.
Data sent: One GET request. No body, no site data.
Data received: Your robots.txt content, parsed for Content Signals declarations (ai_train, ai_input, search).
Purpose: Lets the classifier know which AI purposes you’ve opted out of so it can flag violators.
How to disable: Delete the Content Signals lines from your robots.txt, or filter raif_content_signals_source to return an empty array.
DNS lookups. For search-engine bots that publish a verification method (Googlebot, Bingbot, Applebot, DuckDuckBot, YandexBot), the plugin performs reverse-DNS and forward-confirm lookups against the bot’s IP. These are DNS queries, not HTTP requests; they hit whatever resolver your PHP install is configured to use. Cached for 24 hours per IP.
Dashboard rendering, bot classification, policy decisions, and logging all run entirely inside your WordPress install. The endpoints above are the complete list; no license check, license activation, or analytics beacon is ever contacted.
