Skyboot AI Markdown Changelog

The latest Skyboot AI Markdown version is 2.7.5, released on 2026-10-24. Every release is listed below, newest first, with its new features, improvements and fixes.

2.7.5 – 2026-10-24 #

• Fixed: imported FAQ sections opened in the block editor as a Custom HTML block instead of the Rank Math, Yoast or AIOSEO FAQ block. The FAQ markup now matches exactly what each SEO plugin saves, checked with the block editor’s own validation.
• Fixed: Yoast FAQ blocks now include the question and answer fields Yoast displays, not only its schema fields.
• Fixed: AIOSEO FAQs are created the way AIOSEO builds them: one FAQ block per question, with the answer as paragraph blocks.
• Fixed: Details-format FAQs keep each answer paragraph as its own paragraph block.
• Fixed: “Auto-Recover Gutenberg Blocks” turned a block it could not validate into plain HTML and kept only its first part. It now rebuilds the block from its own settings, as the editor’s “Attempt recovery” does, so an FAQ block stays an FAQ block.
• Improved: FAQ answers with “–” or other special characters no longer risk breaking the block comment.

2.7.4 – 2026-10-24 #

• Fixed: FAQ sections imported from a ZIP or .md file, applied by Review & Fix, or added with “Fix this issue” lost their Rank Math, Yoast or AIOSEO FAQ block data when saved, because WordPress removed the backslashes inside the block settings. “Fix this issue” also no longer damages FAQ blocks that were already on the post.
• Fixed: importing one .md file with front matter created two posts. Combined files are now split only at each post’s front matter or page separator, never inside code blocks or at a normal “—” line.
• Fixed: the “=====” separator of combined exports was not recognised on import.
• Fixed: ZIP exports lost a post when two posts had the same file name (for example two drafts with the same title, or a page and a post with the same slug). The second file now gets the post ID added to its name.
• Fixed: exported Markdown dropped every closing curly quote and the space after it, so quoted text was cut off.
• Improved: ZIP imports use the file name as the title when a file has no title, and AIOSEO FAQ answers keep separate paragraphs.

2.7.3 – 2026-10-24 #

• Improved: Markdown import reads the slug from the url front matter field (the field our own export writes) when there is no slug field, so exported files import back to the same URL without manual slug edits.

2.7.2 – 2026-10-24 #

• Fixed: llms.txt left out custom post types such as BetterDocs documentation, WooCommerce products and portfolio or theme post types. It was built before those plugins registered their content.
• New: Content types setting (AI Visibility > Settings). Choose which post types get Markdown versions, llms.txt entries, IndexNow pings and readiness scans. New post types are included automatically.
• New: Exclude IP addresses from AI analytics, including an “Exclude my IP” button on the AI Analytics tab. Visitor IPs are still never stored.
• Improved: no Markdown alternate link on 404 and search result pages.
• Fixed: Markdown import no longer takes a “# heading” inside a code block as the post title, and the front matter title is used when the text has no leading H1.

2.7.1 – 2026-10-22 #

• Fixed: the AI Markdown admin menu could appear twice with many repeated submenus after activating or updating the plugin, saving settings, or exporting content.

2.7.0 – 2026-10-22 #

• New: Permissions tab (administrators only). Choose which roles can use each feature: View AI Analytics, View AI Visibility, Review & fix content, Use the AI API key, Run site tools, Export, Import and Manage plugin settings.
• New: per-person access. Allow or deny any feature for a specific user, overriding their role.
• The admin menu, tabs and buttons follow each person’s permissions, and every action is also checked on the server. Fixing still requires permission to edit the post.
• Administrators always have every permission; permissions and the AI API key stay administrator-only. WordPress roles are never modified.
• Defaults: editors can view analytics and AI Visibility, fix content and use the AI API key; authors can fix their own posts.

2.6.1 – 2026-10-22 #

• One-issue fixes are now strictly based on the page: the prompt states the page title and SEO focus keyword, requires using only the page, and lets the AI answer NOT_ENOUGH_INFO instead of guessing.
• The plugin refuses to write a part when the page is too thin to support it (for example an FAQ needs 150+ words of page content) or when there are no headings to rewrite.
• New grounding check before applying any one-issue fix, including text pasted from ChatGPT or Claude: numbers that are not on the page and text that does not match the page are flagged, and the author must confirm.

2.6.0 – 2026-10-22 #

• New: fix one issue at a time. In Review & Fix, “Fix this issue” lets AI write only the missing part (meta description, answer-first intro, FAQ block, question headings, key-takeaways list or product summary) and inserts just that. Facts, sources and specs return suggestions only.
• New: re-scan a single post, and remove pages from (or add them back to) the readiness scan.
• New: “Write with AI” for the llms.txt summary (API key or ChatGPT/Claude/Gemini/Copilot).
• New: every list in AI Analytics and AI Visibility has search, sortable columns and pagination; “All scored content” lists every scored item with type, status, score, issue count, Edit and View links.
• New: Fake crawlers table (which crawler names were faked, how often, last attempt).
• New: Last seen column on page tables; feed, embed and AMP requests are counted under their page.
• Fix: builder templates and widgets (Elementor templates and floating buttons, Divi, Bricks, Oxygen, Beaver Builder, Kadence, Astra hooks, reusable blocks) are no longer scored, submitted to IndexNow or listed in llms.txt.
• Fix: AI meta descriptions are saved to every active SEO plugin, including All in One SEO, The SEO Framework, Slim SEO and SmartCrawl; without an SEO plugin the excerpt is used and counted.

2.5.4 – 2026-10-21 #

• SEO fix: Markdown responses for the blog (posts page), category/tag/product archives and author pages sent a canonical pointing to the first post in the list. Each listing now declares its own canonical URL, including page numbers.
• Fix: the posts page (e.g. /blog.md) returned the front page content; it now lists the posts. Its HTML page now links to /blog.md instead of the homepage Markdown.
• Fix: paginated listings (page 2, 3…) no longer share the page 1 cache.
• llms.txt summary: Markdown symbols are removed so the blockquote stays valid; without a summary the site tagline is used. New checklist step reminds you to describe the site.
• No duplicate meta description / JSON-LD when The SEO Framework, Slim SEO or SmartCrawl is active.
• llms.txt is cached by CDNs for 1 day instead of 7, so new pages appear sooner.

2.5.3 – 2026-10-21 #

• Security: password-protected posts no longer expose their content through .md URLs, ?format=md, Accept: text/markdown, archive excerpts, llms-full.txt, the meta description or JSON-LD.
• Security: export and cache folders now block direct web access (.htaccess + web.config, no directory listing). Abandoned exports are deleted after one hour; export session IDs are strictly validated.
• Security: the AI & Markdown Checker only requests URLs on this site (no server-side requests to other hosts or internal addresses) and does not follow redirects.
• Security: Markdown import validates post type, status and author; ZIP import skips files over 5 MB and stops after 2,000 files.
• Hardening: AI visitor counting only accepts existing pages and has a daily ceiling; spoofed crawler traffic to random URLs no longer creates new rows.
• llms-full.txt now skips noindex and checkout/cart/account pages, like llms.txt.

2.5.2 – 2026-10-21 #

• Fixed: buttons no longer show an underline on hover (checklist actions, item counts, fix modal).

2.5.1 – 2026-10-21 #

• AI fixing setup: choose a provider, enter the API key, click “Fetch models” and pick a model from the live list your account can use (Anthropic, OpenAI, Google Gemini, OpenRouter). Text-generation models only; OpenRouter shows prices per 1M tokens and has a filter box. “Other” lets you type any model ID. The fetched list is remembered for a week.

2.5.0 – 2026-10-21 #

• New: “Review & fix” for citation readiness. Click a page (Fix these first, the AI Ready column, analytics tables or the editor box) to see every issue, why it matters and how to fix it. Click an issue count under “Most common gaps” to see every page with that issue.
• Fix without an API key: one click opens ChatGPT, Claude, Gemini, Copilot or Perplexity with a ready prompt and the page as Markdown (prompt also copied to the clipboard). Paste the answer back and click “Update post”.
• Fix with an API key (optional): Anthropic (Claude), OpenAI, Google Gemini or OpenRouter. The improved page is generated in WordPress, shown for review, then applied.
• Updates go through the Markdown importer: Gutenberg blocks, FAQ blocks in the configured format (Rank Math, Yoast, AIOSEO, details), SEO meta description, product short description. A revision is always kept. Page-builder pages (Elementor, Divi, WPBakery, Bricks) are never overwritten.
• The AI is told never to invent facts, numbers, sources, specs or reviews. Missing figures and sources are returned as notes for the author (“AI suggests, you add”).
• Icons inside buttons now always match the button text colour.

2.4.1 – 2026-10-21 #

• Fixed: raw shortcodes (e.g. , [contact-form-7 …]) no longer appear in Markdown page bodies, front matter descriptions, archive lists, llms.txt or meta descriptions, whether or not the shortcode’s plugin is active. Normal bracketed text like [Step 1] is kept.
• Breadcrumb shortcodes from Rank Math, Yoast, AIOSEO, SEOPress and Slim SEO are removed from Markdown pages (page chrome, not content).
• Markdown front matter description now uses the SEO plugin’s meta description first.

2.4.0 – 2026-10-21 #

• Real data only: AI crawler hits are verified against the official IP lists of OpenAI, Google, Bing, Perplexity, Apple and DuckDuckGo (plus reverse DNS). Fake crawlers that copy a user agent are filtered out and shown separately.
• Requests from localhost, private networks, the server itself and the plugin’s own access test are never counted.
• Visitors from AI are counted once per visitor, page and day, only from real browsers on this site.
• Real visitor IP detection behind Cloudflare (trusted only for genuine Cloudflare edge IPs).
• SEO plugin compatibility layer: noindex, meta description and canonical are read from Yoast SEO, Rank Math, All in One SEO, SEOPress, The SEO Framework, Slim SEO and SmartCrawl, including post-type defaults.
• E-commerce: WooCommerce, Easy Digital Downloads, SureCart and FluentCart. Product Markdown pages get price, sale price, availability, SKU, GTIN, brand, rating and specifications (front matter + table). Category pages list products with price and stock. llms.txt gets product facts and product category sections. Cart, checkout, account and receipt pages are excluded everywhere.
• Product-specific citation readiness checks (summary, price, availability, identifiers, specs, description, reviews, FAQ) and a thin category page report.
• IndexNow also submits category pages when they are edited.
• Fixed: llms.txt was rebuilt during save_post (twice per save) before store plugins saved prices; it is now rebuilt once per request with complete data.
• Redesigned AI Analytics and AI Visibility screens: period comparison, verified status per crawler, “Get cited” checklist with one-click actions, full-width layout.
• Fixed: admin tab bar no longer scrolls horizontally.

2.3.0 – 2026-10-21 #

• New: AI Analytics dashboard. Counts AI crawler visits per page (ChatGPT, Claude, Perplexity, Google, Bing and more), split into live answer fetches, AI search crawls, search index crawls and training crawls, plus human visits referred by ChatGPT, Perplexity, Gemini, Copilot, Claude and other assistants (cache-safe beacon). Aggregated per day, no IP addresses stored, automatic retention cleanup.
• New: IndexNow. Auto-submits URLs on publish, update and trash; serves the key file; “Submit all URLs” tool and submission log.
• New: AI Crawler Access Test. Requests the homepage as each answer-engine crawler and checks the live robots.txt to catch firewall/CDN/robots blocks.
• New: AI training crawler policy. Optionally block training-only crawlers (GPTBot, ClaudeBot, Google-Extended, CCBot…) in robots.txt without affecting AI search and citations.
• New: Citation Readiness score. Scores each post on answer-first intro, question headings, FAQ, lists/tables, facts, sources, freshness, depth, meta description and indexability. “AI Ready” list column, editor checklist, bulk scan and site-wide gap report.
• Fixed llms.txt truncation: removed the 48,000-character hard cut that dropped most pages and the entire Posts section on larger sites.
• llms.txt now skips noindex (Rank Math / Yoast) and password-protected content, and builder template post types.
• Cleaner link descriptions: uses the SEO meta description first and strips leftover shortcodes such as .

2.2.3 – 2026-10-10 #

• Fixed FAQ Multi-Line & Multi-Paragraph Answers: Ensured FAQ answers with multiple lines or paragraphs are completely captured and formatted without truncating lines.
• Fixed Complete FAQ List Ingestion: Fixed parser boundary loop to ensure all FAQs in a section (e.g. 7 of 7) are imported without skipping items.
• Added FAQ Section Description Support: Added automatic capture and rendering of introductory description paragraphs placed directly under FAQ headings (e.g. ## FAQ).
• Enhanced 100% Roundtrip Fidelity: Synchronized HTML-to-Markdown export with Gutenberg FAQ blocks for exact 1-to-1 roundtrip parity.

2.2.2 – 2026-10-07 #

• Fixed LiteSpeed Cache Poisoning: Enforced early, universal non-cacheable directives (X-LiteSpeed-Cache-Control: no-cache, Cache-Control: no-store, no-cache, DONOTCACHEPAGE, LSCACHE_NO_CACHE, and LiteSpeed API litespeed_control_set_nocache) across all Markdown response paths prior to output.
• Fixed Search Engine Cloaking Risk: Stopped overriding canonical URLs based on User-Agent alone (ClaudeBot, PerplexityBot, OAI-SearchBot now receive normal HTML when requesting HTML like standard browsers and Googlebot). Content negotiation on canonical URLs is strictly driven by Accept: text/markdown.
• Added LiteSpeed Vary Separation: Registered litespeed_vary filter and X-LiteSpeed-Vary: value=md response header so Markdown and HTML responses are never stored under identical LiteSpeed cache keys.
• Added Canonical URL Content Negotiation Setting: Added a dedicated setting allowing site administrators to enable or disable Markdown negotiation on canonical URLs while keeping dedicated .md endpoints active.
• Added Original Request URI Preservation: Preserved and restored original REQUEST_URI during .md processing, preventing server caches and loggers from mapping dedicated .md responses to canonical HTML paths.
• Added Automatic Upgrade Cache Purge: Automatically calls do_action('litespeed_purge_all') and purges all internal Markdown caches upon plugin update to clear any previously poisoned cache entries.
• Fixed Dedicated .md URL Resolution: Resolved lifecycle hook race condition where early no-cache checks cached a negative markdown request decision prior to .md suffix detection, restoring direct .md URL access across all permalink structures.
• Updated .htaccess & Server Rules: Added dedicated .md bypass directives and dual mod_rewrite/LiteSpeed module blocks.

2.2.1 – 2026-10-04 #

• Markdown Title & H1 Import Prioritization: Markdown import now prioritizes the first # H1 heading in the content body as the post title, stripping it from the content body to avoid duplication.
• Preserved Rank Math Dynamic Variables: Default Rank Math dynamic template variables (e.g. %title% %sep% %sitename%) remain untouched and intact upon importing Markdown, only applying %currentyear% and letting Rank Math render %title% dynamically from the imported post title.
• Single Post Editor Import Flow: Single post editor Markdown import now parses and populates the editor blocks, title, and SEO meta in-memory without directly auto-publishing or updating the database via wp_update_post(), allowing users to review imported content before clicking “Update” or “Save Draft”.
• Added Dynamic Year Conversion: Automatically detects the current year (e.g. 2026) in imported SEO title and meta description fields and converts it to %currentyear%.
• Added Own Domain Affiliate Link Slug Support: Added custom affiliate link prefix/slug setting under “Open External Links in New Tab” so site-specific redirect/affiliate paths (e.g., domain.com/go/, domain.com/visit/, domain.com/recommend/) open in a new tab (target="_blank").
• Fixed Escaped Markdown Character Unescaping: Automatically unescapes backslashed currency amounts, numbers, and punctuation (e.g., $95.8 -> $95.8, December 25. -> December 25.) during import.
• Enhanced SEO Description Import: Added full support for importing and mapping description and meta_description from YAML frontmatter directly into Rank Math, Yoast, AIOSEO, and SEOPress.

2.2.0 – 2026-09-28 #

• Added Two-Tier Markdown Cache: Tier 1 Object Cache (Memcached/Redis) with Tier 2 File Cache fallback (wp-content/cache/skyboot-md/) delivering sub-300ms TTFB on cache hits.
• Added Stampede Lock Protection: Ensures only 1 worker process generates Markdown during crawler bursts while concurrent AI crawlers wait and receive cached content.
• Added Zero-Latency llms-full.txt: Pre-warmed in background via WP-Cron, completely preventing CPU spikes and runtime generation lag during crawler requests.
• Added 304 Not Modified & ETag Support: Enables HTTP 304 conditional responses for AI crawlers with fresh copies, saving server bandwidth and execution time.
• Added Cloudflare Edge Cache for Dedicated .md URLs: Applies public edge caching on direct .md URLs while keeping strict private content negotiation on identical URLs.
• Added Automatic Cache Purging & Background Warming: Post updates trigger targeted cache invalidation, LiteSpeed litespeed_purged_all purge, and scheduled background warming.
• Added Admin Cache Purge: One-click “Purge Markdown Cache” button in plugin settings.
• Cleaned User-Agents: Removed non-standard -Extended robots.txt tokens from .htaccess and Nginx rules.

2.1.2 – 2026-09-15 #

• Fixed Cache Contamination: Replaced all shared/public edge caching on Markdown responses with strict Cache-Control: no-store, no-cache, must-revalidate, max-age=0 and X-LiteSpeed-Cache-Control: no-cache.
• Fixed Page Cache Exclusion: Added automatic DONOTCACHEPAGE and LSCACHE_NO_CACHE constants across all Markdown requests to prevent LiteSpeed/Cloudflare edge cache key collisions for human visitors.
• Fixed Rank Math SEO FAQ block format as default setting.

2.1.1 – 2026-09-02 #

• Fixed AI Crawler Indexing: Updated X-Robots-Tag from noindex, nofollow to noindex, follow on Markdown endpoints so AI search engines and crawler bots can ingest content and traverse links.
• Fixed Discovery Endpoints: Removed noindex, nofollow restrictions from /llms.txt, /llms-full.txt, and /.well-known/api-catalog.
• Fixed Edge & CDN Caching: Added RFC-compliant Cache-Control (s-maxage=604800, stale-while-revalidate=60) and CDN-Cache-Control headers for ultra-fast Cloudflare and LiteSpeed cache hits.
• Fixed .md URL Delivery: Removed the aggressive 302 redirect on .md URLs so scrapers and bots requesting Markdown are served 200 OK directly.
• Enhanced Discovery Headers: Added dual rel="canonical" and rel="alternate"; type="text/html" HTTP Link headers to Markdown responses.
• Added Link Import Attributes: Added target=”_blank” and rel=”nofollow” options with customizable domain whitelist for imported markdown links.
• Added Engineering Standard: Introduced AI_CITATION_GUIDELINES.md to prevent future crawler regression.

2.1.0 – 2026-08-18 #

• Added 100% native Gutenberg block converter supporting Headings, Paragraphs, Lists, Tables, Quotes, Code, Separators, Images, and Links with zero recovery errors.
• Added SEO FAQ block compatibility for Rank Math SEO, Yoast SEO, All in One SEO (AIOSEO), and WordPress Details blocks.
• Added post import reload settings and safety confirmation modal popup with “Reload Now” button.
• Added automatic Gutenberg block recovery handler (auto_recover_blocks).
• Fixed placeholder text corruption bug during Markdown import.
• Polished modal CSS layout and button flex alignment.

2.0.0 – 2026-07-10 #

• Added bulk Markdown export engine with ZIP archive and single-file options.
• Added /llms.txt, /llms-full.txt and /.well-known/api-catalog virtual endpoints.
• Added Content Negotiation for Accept: text/markdown, .md, and ?format=md.
• Added single post metabox tools and copy-to-clipboard functionality.

1.1.1 – 2026-06-01 #

• Initial Accept-header Markdown negotiation release.

FAQ #