| 开发者 | athenalion |
|---|---|
| 更新时间 | 2026年10月17日 09:00 |
| PHP版本: | 8.0 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPL-2.0+ |
| 版权网址: | 版权信息 |
yoursite.com/llms.txt) is generated immediately. Open it in a browser and look.shop2llm folder to /wp-content/plugins/ (or install "Shop2LLM" directly via Plugins → Add New)Yes. Shop2LLM exposes a full MCP (Model Context Protocol) server at /wp-json/shop2llm/v1/mcp with four JSON-RPC 2.0 tools: search_products, get_product, get_policies, and get_store_overview. ChatGPT connects via your site's AI connection URL and calls these tools like native tools. The MCP server works with WooCommerce product catalogs and — on sites without WooCommerce — with posts and pages.
Yes. Shop2LLM auto-generates /llms.txt (machine-readable content catalog), /llms-full.txt (complete content for agents), and /agents.md (structured Markdown dossier with your UCP score, MCP tools, and schema types) the moment you activate — no configuration, no cron, no external service. The catalog rebuilds automatically when you publish, update, or delete content. If another plugin writes a physical llms.txt file that shadows the endpoint, Shop2LLM detects the conflict and offers one-click remove or take-over.
An llms txt file (written "llms.txt") is a simple text file that tells AI assistants like ChatGPT, Claude and Perplexity what your site contains and where to read it. Shop2LLM creates and keeps it up to date for you automatically.
Three checks, under a minute each:
yoursite.com/llms.txt in a browser — you should see your content catalog (works immediately after activation)wp eval-file wp-content/plugins/shop2llm/tools/activation-doctor.php) — it prints your onboarding funnel and points at the bottleneck (most often Cloudflare's "AI Scrapers and Crawlers" block, or a firewall returning 403 to bot user agents).
Shop2LLM appends 51 explicit Allow: / invite directives to robots.txt — GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, Amazonbot, Bytespider, meta-externalagent and 41 more. It never blocks bots — these are invitations. You can additionally declare a site-wide AI usage policy (Training / Summarization / Attribution) as an opt-in comment banner at the top of robots.txt (Settings → Shop2LLM → AI Agents).
Yes — three ways. (1) Product JSON-LD schema on every product page: name, price, availability, GTIN/MPN/brand, ratings. (2) A Google-Merchant-format XML feed at /product-feed.xml (WooCommerce stores; shopping AI platforms index it directly). (3) The MCP search_products tool, which lets ChatGPT query your live catalog at conversation time so answers use current prices and stock. An optional cloud Product Sync (off by default) additionally lists your products in the Shop2LLM AI search index.
Four scores, all computed on your server. The site-wide AI Visibility Score™ has 10 factors / 100 points: robots.txt invites, llms.txt health, llms-full.txt reachability, JSON-LD schema, MCP endpoint, product feed (WooCommerce), AI search queries, AI crawler activity, store policies, multilingual readiness. The post-editor Content Readiness audit has 25 items in five groups: Discoverability, Relevance, Content Quality, Entity Authority, Activity. WooCommerce products get an 11-dimension Product AI Score, and the UCP Readiness Score (5 dimensions, 100 points) measures agentic-commerce readiness (Catalog API, MCP & Agentic, llms.txt, Schema, robots.txt). Every item shows a fix action.
Yes — 19 JSON-LD types are auto-detected and emitted with zero configuration: Product, Article, BlogPosting, FAQ, HowTo, Recipe, Event, Course, JobPosting, SoftwareApplication, Book, MusicRecording, VideoObject, Person, Dataset, Review, LocalBusiness, WebPage, BreadcrumbList. If Yoast, Rank Math, AIOSEO, or SEOPress is active, Shop2LLM detects it and disables overlapping schema modules so nothing is emitted twice.
Tracking tools measure how you rank in AI answers; Shop2LLM builds the server-side layer those answers draw from — llms.txt, MCP server, JSON-LD schema, 51 crawler invites, and visibility scores, all running locally with no account. Because it coexists with your existing SEO plugin (auto-deduplicating overlapping modules), it can run alongside tracking tools rather than replacing them.
No. All core features (llms.txt, /agents.md, MCP tools, JSON-LD schema, crawler tracking, visibility scores) run entirely on your WordPress server and start working on activation — no account, no cloud signup, no API key. The optional Hero Test Search uses our cloud LLM proxy with 5 free queries per day, also without an account. All AI connections to your site use Bearer token authentication; the key is auto-generated and can be rotated anytime.
No. Shop2LLM works on any WordPress site out of the box — your posts and pages are fully included in the AI catalog (llms.txt, BlogPosting schema, MCP search). If WooCommerce 8.0+ (free) is active, product search, price and stock data, Product schema, and the /product-feed.xml XML feed join the catalog automatically.
Content data (full product descriptions, customer data, orders) stays on your server. AI agents connect directly to your WordPress REST API for detailed product information. External transmission is opt-in and off by default:
add_filter('shop2llm_disable_product_sync', '__return_true');first_search_done: true, dashboard_visits: "2-5") with a SHA-256 hashed site identifier — no PII, no exact counts, no URLs, no search termsFree works with ChatGPT plus any MCP-compatible client — Claude Desktop, Claude Code, Gemini CLI, and more — via your connection URL, with 4 AI tools (content search, detail lookup, store policies, store overview) plus everything listed above. Pro adds cloud connections (Claude, Gemini, Copilot, Meta AI, Amazon Bedrock) so consumers can discover your products without any manual connector setup, plus the analytics dashboard and Checkout Kit. Upgrade at https://shop2llm.com/
Your site returns to normal. llms.txt, robots.txt AI rules, schema markup, and the product feed are removed. All settings are preserved if you reactivate later. No data is lost.
No. Shop2LLM is independently tested and A-rated on MakeWPFast.com — a third-party WordPress speed certification service. The benchmark report covers Home, Blog, and Product page contexts and rates overall plugin impact as "Low" (measured TTFB, memory delta, and DB query delta against a clean WordPress baseline). All AI features run on your server through the standard WordPress REST API; external calls are non-blocking with 1s timeouts — even if our cloud is unreachable, your site stays fast.
Yes. llms.txt and llms-full.txt are served with Cache-Control: no-cache headers, so caching plugins (LiteSpeed Cache, WP Rocket, W3 Total Cache) and edge CDNs like Cloudflare never serve a stale copy after your catalog changes. With default settings no configuration is needed; if you run an aggressive "cache everything" edge rule, add an exclusion for llms.txt.
All of them. The 51 named robots.txt invites include Google-Extended (which powers Gemini groundings), ClaudeBot and Claude-SearchBot (Anthropic), so Gemini and Claude can discover and read your catalog. The MCP endpoint is the native protocol of Claude Desktop and Claude Code, and also connects from Gemini CLI and any other MCP-compatible client. The llms.txt and llms-full.txt catalogs are plain markdown every AI engine can read.
Yes. Shop2LLM auto-detects SEO plugins (Yoast, Rank Math, AIOSEO, SEOPress) and automatically disables conflicting modules (meta tags, schema, Open Graph). You keep the AI-specific features (MCP tools, llms.txt, crawler invites, visibility scores) and the other plugin keeps handling classic SEO.
/65 instead of /100) with "Unlock full score →" upgrade linksRedirects and Sitemap sub-menus into the main Shop2LLM → SEO tab as sub-tabs. The WP admin sidebar now shows a single Shop2LLM entry instead of three. Old URLs (?page=shop2llm-seo-redirects, ?page=shop2llm-seo-sitemap) auto-redirect to the new sub-tab locations.nav-tab sub-navigation inside SEO tab (SEO Settings / Redirects / Sitemap).initialize returns dynamic protocolVersion (echoes client-requested version, defaults to 2025-06-18) instead of hardcoded 2025-11-25notifications/* requests now return HTTP 202 with empty body per MCP 2025-06-18 spec (previously returned -32601 Method not found)tools/call results now wrapped in content[] + structuredContent structure per specmcp.version synchronized to 2025-06-18heroRemaining JavaScript error on activation (props WP.org Plugin Tests)wp eval-file tools/activation-doctor.php) for onboarding funnel, connectivity checks, and Site Hash for fast support