TWT AEO Ultimate

llms.txt and `llms-full.txt`, Markdown content negotiation for AI agents, Content-Signal headers, agent-discovery endpoints (Agent Skills Index, MCP and WebMCP, OAuth/OIDC discovery, RFC 9727 API catalog, RFC 8288 Link headers), and per-bot blocking and rate limits for 20+ AI crawlers such as GPTBot, ClaudeBot, PerplexityBot and Google-Extended. You decide which AI systems may read, cite or train on your content. Where creating `/.well-known/` files needs server access the plugin doesn't have, you get copy-and-paste instructions.

Changelog

2.25.0

• New: RAG Engine (switch it on under Modules; it replaces the Chunk View screen). One screen for how AI retrieval sees your site: Chunk View, Documents, Search Placement and AI Citations.
• New: Documents. Upload PDFs, Word files and text: spec sheets, manuals, FAQs. They are read on your own server and the uploaded file is deleted once read; only the searchable passages are kept. Large files upload in pieces and resume after a dropped connection.
• New: tables in PDFs are rebuilt as real tables. Standard/option charts read “Standard”, “Option” or “N/A” under the right model, and spec tables keep each value beside its label. Copy pastes a table straight into the editor.
• New: document labels. Each document is marked as a law, code, safety data sheet, contract, manual, spec sheet or other, with its place, edition and what to cite it as. Laws, codes and safety data are copied word for word with their citation. An optional switch lets AI read the first few pages to label them; you can correct any label.
• New: law check. For a law or code, an online search (Gemini or Perplexity) looks for changes after your copy’s date and links to the official source. Only the citation is sent, never your document’s text.
• New: Search Placement lists the Google and Bing searches you rank 4th to 15th for that your documents answer and the ranking page does not. Searches for a business to hire (“near me”, “services”, a town) are left out, with a note saying how many.
• New: AI Citations lists AI Visibility questions where an engine left you out, with the passages that answer them. It reads your existing runs, including paused or unfinished ones, and never runs new checks.
• Changed: the Search Console setting is now a dropdown of your Google account’s own properties. Fixed: domain properties (“sc-domain:”) could not be saved, and an address that did not exactly match a property failed with “User does not have sufficient permission”.
• Changed: AI Visibility stops asking an engine for the rest of a run once it rejects its API key.

2.24.0

• Changed: Perplexity now runs on its Agent API. Perplexity retires the Sonar API on September 27, 2026; earlier versions of this plugin lose Perplexity answers in AI Visibility and Perplexity lookups after that date. Perplexity errors now say what went wrong in plain words: a rejected key, no credit, rate limiting, or an API Perplexity has retired.
• New: choose the model for each AI provider under Settings → AI Models, or type in any model ID. Adds GPT-6 (Luna, Sol, Astra), Gemini 3.5–3.8 Flash, Claude Sonnet 5 and Opus 5. Sites that already had a key keep the model they were using; new setups start on Claude Haiku 4.5, GPT-6 Luna and Gemini 3.6 Flash.
• Fixed: newer models reject settings older ones needed (temperature on Claude 5, “minimal” reasoning on GPT-6, “minimal” thinking on Gemini 3.7/3.8). Requests now adapt to the model chosen.
• Fixed: Gemini 2.5 Flash is now limited to Google accounts that already used it, so new Gemini keys could fail. New setups use Gemini 3.6 Flash; web-grounded lookups on a free-tier key fall back to 2.5 where the account allows it.
• Fixed: IndexNow no longer submits URLs from local, development or staging copies of a site (or from .local, .test, localhost and private-network addresses). The log shows these as “Skipped” with the reason.
• Fixed: Chunk View labelled sibling sections as nested (“Features › Pricing”) on pages whose headings start at H2, which is most WordPress content. Each passage now shows the heading it actually sits under.

2.23.0

• New: Chunk View (off by default; switch it on under Modules). Shows a page the way a retrieval index holds it: fetched like a no-JavaScript AI crawler, split into passages, and each passage scored on whether it still makes sense on its own. Works on any URL, including competitor pages, with no API key.
• New: host governor. Heavy work measures what your host can handle and pauses instead of timing out on smaller hosting plans.
• Fixed: AI Visibility ignored allocation changes once a question list had been saved, and brand questions were never asked.
• Fixed: the AI Visibility question counts ignored “Per brand” and overcounted categories, posts and products.

2.22.0

• New: AI Visibility shows the exact passage where each engine named your company or brands, so you see how you were mentioned, not just that you were.
• Fixed: stored checks dropped owned-profile citation data, so brand totals undercounted. Re-run a check to refresh them.

2.21.0

• New: Bot View. See any page the way no-JavaScript AI crawlers (GPTBot, ClaudeBot, PerplexityBot) see it: what schema is visible, what is locked behind JavaScript, and which hidden script data could become schema.
• New: every Bot View audit fetches the page twice, as GPTBot and as a normal browser, exposing firewalls that turn AI crawlers away.
• New: a weekly Bot View scan of your homepage and recently updated content, with every finding linked to its fix.
• Fixed: AI Crawler Watch no longer counts Bot View’s own audits as crawler visits.

2.20.1

• New: duplicate meta descriptions are flagged on the Not Indexed screen, with an inline AI fix.
• Fixed: headings that were not questions could be published as FAQ schema.
• Fixed: saved custom schema of an unusual type could be invisible and impossible to delete.
• Fixed: accented letters, dashes and emoji in stored schema could be saved as literal codes.
• Fixed: Easy Digital Downloads checkout and account pages were reported as not indexed.

2.20.0

• New: content provenance. Articles can declare how they were made (human, AI-assisted or AI-generated) using schema.org’s digitalSourceType.
• New: author archives publish ProfilePage schema.
• New: multilingual schema for WPML and Polylang: each translation declares its language and links to the others.
• New: a contextual authority statement, drafted with AI, that explains an author’s credentials in plain words.

2.19.0

• Critical: Google Merchant Center sync moved to Google’s Merchant API before the old Content API shut down.
• New: author certifications are machine-recognizable credentials, with a category, the awarding body and its Wikipedia or Wikidata link.
• New: AI Visibility turns every question where you were left out into a brief of what to publish.
• New: AI Visibility checks its citations against real referral visits, and tracks each brand over time.
• Fixed: pushing a promotion to Merchant Center caused a fatal error.

2.18.0

• New: brand citation tracking. Each brand you track is asked about by name, and a brand can be a whole website, not only social profiles.
• New: the press-release features work on WordPress 7.0 without an API key of your own.
• Improved: Search Console top queries no longer fill up with rank-tracker noise.
• Fixed: /okf/ and /llms.txt could fail on hosts without PHP’s mbstring extension.

2.17.0

• New: AI Visibility, the citation engine. Asks AI engines the questions your customers ask and records whether you were cited, named or left out.
• New: Grok (xAI) and Le Chat (Mistral) join ChatGPT, Gemini, Claude and Perplexity.
• New: an Open Knowledge Format (OKF) bundle at /okf/.
• New: runs continue in the background without the tab open, and awkward questions can be polished with AI.

2.16.0

• New: the AEO Score, a 0 to 100 citability grade that shows where every point is lost.
• New: one-click “Auto-Fill Missing Schema” on the Dashboard, and “Generate with AI” for schema descriptions.
• New: Review and HowTo forms in the schema editor.
• Fixed: Search Console domain properties could not be connected.

2.15.0

• Fixed: every scanning screen stopped at the 200 most recent items. Scans now cover the whole site.
• New: catalogue-wide search and pagers on the big tables, and the FAQ detector scans products.
• Changed: replacing WooCommerce’s own Product schema now needs the store owner’s consent.
• Improved: one consistent image fallback for social cards, and products with gallery images are no longer reported as missing an image.

2.14.0

• New: the full commerce schema suite from AEO Ultimate for WooCommerce: Product and ProductGroup graphs, Smart Collections and Promotions.
• New: the Setup Wizard’s Autopilot now does real work, running every setup step that needs no AI in the background.
• Fixed: “Generate Schema for All” and “Scan All Pages” stopped after 200 items.
• Compatibility: verified with WordPress 7.1 and WooCommerce 11.

2.13.0

• New: bulk actions on every screen that had a per-row button, so a whole site can be brought to full coverage in one pass. AI-powered ones show the billed call count and ask first.
• New: a switch for every output that could collide with another plugin, on the Schema Conflict Detector page.
• Fixed: the text served to AI crawlers (llms.txt, llms-full.txt, Markdown pages and Agent Skills documents) was HTML-encoded, so characters such as > and & reached crawlers as > and &, breaking the llms.txt format.
• Fixed: the WordPress 7.0 AI Client connection.

2.12.1

• Fixed: hosts without PHP’s libsodium extension could hit a critical error when saving keys.
• Fixed: the Not Indexed page could fail on scan results from older versions.
• Improved: WooCommerce cart, checkout and account pages are no longer reported as not indexed, and fixable issues get a Fix button.

2.12.0

• Fixed: the Author Entity (E-E-A-T) page could fail to load on some sites.

2.11.0

• New: a unified Knowledge Graph. Every schema block on a page is woven into one connected @graph, with duplicate Organization output consolidated.
• New: the Knowledge Graph page, to define entities, relationships and per-page about and mentions, grounded in Wikipedia and Wikidata.

2.10.0

• New: Smart 404 Rescue (off by default). Heals broken internal links, ignores hacker probes, and recovers pages Google still ranks you for with suggested 301 redirects, without a bloated 404 log.
• New: Access & Roles. Let an Editor or an SEO Manager role use the tools without being a full administrator.

2.9.1

• New: Product schema includes Google’s merchant-listing fields: price valid-until dates and sale periods.
• New: an AI resolver maps each product to its exact Google Product Category.
• New: Merchant Center sync flags products whose category disagrees with the feed.

2.9.0

• First release on WordPress.org.
• Removed: the built-in content generator (now the separate TWT Content Generator plugin) and dormant bulk image generation code.
• Improved: every outside service the plugin can contact is listed with its terms and privacy links.

The complete, detailed changelog for every release ships in the changelog.txt file bundled with the plugin.