AI FAQ Schema - AEO, GEO & llms.txt for AI Search
| 开发者 | erinkwon1 |
|---|---|
| 更新时间 | 2026年10月7日 04:12 |
| PHP版本: | 8.0 及以上 |
| WordPress版本: | 7.1 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
详情介绍:
Built for AEO and GEO — write FAQs that AI Overviews, ChatGPT, and Perplexity can easily quote. This is an AI FAQ generator and FAQ schema plugin made for the new way people search: answer engines, not just blue links.
🪄 Fix with AI: score your FAQ quality, then fix it in one click. Run the analyzer, click one button, and every answer is rewritten as a citation-ready 40-80 word answer capsule — direct Yes/No openers, hard data points, self-contained sentences that can be quoted on their own. Your SEO and GEO scores update in real time. 🌍 19-language Auto Generate. ✅ Tested with WordPress 7.1.
No API key required. FAQ parsing, manual input, SEO/GEO quality scoring, FAQPage schema, 7 display styles, llms.txt, and AI crawler control all work out of the box. The AI features (Auto Generate, Fix with AI) are optional - bring your own OpenAI key only if you want them.
📘 Guides and documentation: answerschema.com — plain-English guides on FAQ schema, llms.txt and AI crawler control for WordPress, plus our own test results.
AI Search Visibility - no API key needed:
- llms.txt - serves an auto-generated
llms.txtat your site root (llmstxt.org convention): a machine-readable index of your pages and latest posts, following the proposed llms.txt convention. AI companies have not confirmed that their crawlers use it, so treat it as a low-cost extra. Rebuilt automatically when content changes. You choose which post types go in, exclude single items, and add your own links; noindex and password-protected content is always left out. On multilingual sites (Polylang or WPML) each language gets its own file, such as /de/llms.txt, linked from the main one. - AI Crawler Control - a one-screen table of 14 AI crawlers (GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended, CCBot, Bytespider and more) with per-bot allow/block via your robots.txt. Everything is allowed by default - being read by AI is the point of AEO - and blocking any bot is one checkbox. Need more? Add your own robots.txt lines in the same screen.
- AI Crawler Visits - see which AI crawlers actually read your site: visits per bot over the last 30 days, how often each one fetched robots.txt and llms.txt, the last page it read, and the pages AI crawlers read most. Stored only in your own database, no external calls.
- Install and activate the plugin, then open any post or page in the block editor.
- Search for "FAQ Schema" in the block inserter and add the block.
- Use Auto Generate (AI reads your post), paste existing FAQs in any format, or type them manually.
- Publish. Valid FAQPage JSON-LD schema is injected automatically — no shortcodes, no code.
- 🪄 Fix with AI — scores your FAQ quality and rewrites your answers to pass. One click.
- ✅ AI writes the FAQs for you — one click reads your actual post content. No pasting, no templates.
- ✅ Real structured data — valid FAQPage JSON-LD, not just visual styling
- ✅ Built-in SEO quality check — 12-point analyzer tells you exactly what to improve
- ✅ GEO quality check — 8-point AI citation analyzer for AI Overview, ChatGPT, and Perplexity visibility
- ✅ 19 languages — FAQ output auto-matches your post language, or force any of 18 specific languages
- ✅ Works with any theme — no shortcodes, no theme dependency
- ✅ Zero build tools — drop it in and activate. No npm, no React, no headaches
- Auto Generate (AI) — One click. The plugin reads your post content directly from the editor and generates relevant FAQs — no copying, no pasting. Powered by OpenAI.
- Paste Existing FAQs — Already have FAQs? Paste them in almost any format (Q:/A:, Question:/Answer:, numbered lists, plain text, HTML, or markdown) and the smart parser auto-detects and structures them instantly.
- Manual Input — Prefer full control? Type your questions and answers directly in the block editor.
- Direct answer in the first sentence — Yes/No questions open with "Yes," or "No," plus a restated claim
- 40-80 words total, every sentence under 20 words
- At least one concrete data point per answer (numbers, dates, timeframes)
- Fully self-contained — extractable by AI engines without any page context
- Keeps your questions and your language untouched (works in all 19 supported languages)
- Basic checks — FAQ count, duplicate detection, question format
- Direct answer quality — first-sentence scoring, search intent alignment
- Keyword SEO — keyword presence in questions and answers, sentence length, uniqueness
- Answer-First structure — no filler phrases, direct answers in the AI citation sweet spot (8-60 words)
- Self-Contained — no relative references, no vague endings, each FAQ stands alone
- Specificity and E-E-A-T — concrete data points, no promotional superlatives
- Clarity — no ambiguous pronouns, structured multi-sentence answers
<head> — the standard format read by Bing and the AI assistants built on its index, and understood by every search and answer engine. You don't touch any code.
Multiple OpenAI Models
Choose the model that fits your needs and budget:
gpt-5.6-luna— Fastest, lowest cost. Great for most blogs.gpt-5.6-terra— Balanced quality and cost.gpt-5.6-sol— Flagship model. Best quality for competitive niches.
- Minimal — Clean text list with Q prefix
- Card — Elevated cards with hover effects
- Accordion — Click to expand, great for long answers
- Boxed — Colored header with distinct sections
- Timeline — Vertical timeline with numbered dots
- Chat Bubble — Messenger-style question/answer bubbles
- Numbered — Large index numbers for tutorial-style posts
{count} as a placeholder for the number of FAQs. The plugin always enforces JSON formatting automatically.
Target Keyword Optimization
Set a target keyword per block (or let AI detect it from your content). The quality analyzer then checks whether your FAQs are properly optimized around that keyword.
Who Is This For?
- Bloggers who want more Google visibility without touching code
- SEO professionals who need to scale FAQ structured data across many posts
- Content marketers building topical authority in competitive niches
- Agencies managing client sites and needing a reliable FAQ solution
- Small business sites looking to get their answers surfaced and cited in AI search (AI Overviews, ChatGPT)
安装:
- Upload the
mkjb-ai-seo-faq-schemafolder to/wp-content/plugins/ - Activate the plugin through the Plugins menu in WordPress
- Open any post or page in the block editor and search for "FAQ Schema" in the block inserter
- Add the block, paste or type your FAQs, and publish — schema is injected automatically
- Optional: to enable the AI features (Auto Generate, Fix with AI), add your OpenAI API key under Settings > AI FAQ Schema
屏幕截图:
常见问题:
Where can I read guides about FAQ schema and AI search?
Our site answerschema.com publishes free guides on FAQ schema, llms.txt, AI crawler control and what changed in 2026, including the tests we ran ourselves. You can also request a feature or report a bug in the support forum.
Do I need an OpenAI API key to use this plugin?
No — not for everything. The "I Have FAQ" (paste mode) and "Manual Input" modes work without any API key. You only need a key for the "Auto Generate" mode and the AI keyword detection feature. You can get an API key at platform.openai.com/api-keys.
Will this help me get found in AI search?
It gives AI engines — Google AI Overviews, ChatGPT, Perplexity, Claude — clean, structured, citation-ready Q&A content, plus valid FAQPage schema. No plugin can guarantee that an AI assistant will cite you. It won't directly change your Google keyword rankings; its job is to make your answers the ones AI assistants quote. Note: Google retired FAQ rich results from Search in 2026, so this plugin is built for AI citation (AEO/GEO) and page understanding, not a Google visual snippet.
What is llms.txt and why would I want it?
llms.txt (llmstxt.org convention) is a plain-text index of your site - your pages and latest posts with one-line summaries - served at yoursite.com/llms.txt. It is a proposed convention: AI companies have not confirmed that their crawlers use it, and Google has said it does not use it for Search, so treat it as a low-cost extra rather than a ranking factor. This plugin generates and serves it automatically, keeps it updated as you publish, and needs no API key. If you already have a physical llms.txt file, the plugin never overrides it.
Does llms.txt work on multilingual sites?
Yes, with Polylang or WPML. The main /llms.txt lists your default language and links to one file per language that has its own directory or domain, for example /de/llms.txt with the German pages and their German URLs. Languages that share the main URL (a ?lang= parameter) are listed in the main file under their own headings instead. Developers can supply languages from other plugins through the mkjb_llms_txt_languages filter.
Can I block GPTBot or other AI crawlers?
Yes. Settings - AI FAQ Schema lists 14 known AI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended, CCBot, Bytespider, meta-externalagent, Amazonbot and more) with a Block checkbox each. Checking one adds a Disallow rule for that bot to your site's robots.txt. Each row explains the trade-off - for example, blocking OAI-SearchBot removes you from ChatGPT Search results. Blocking Google-Extended does NOT affect your normal Google Search ranking.
Does blocking AI crawlers conflict with wanting AI citations?
They are different bots for different jobs. Blocking a TRAINING crawler (GPTBot, CCBot, Bytespider) stops your content being used to train future models, while SEARCH crawlers (OAI-SearchBot, PerplexityBot) are what get you cited in live AI answers. The table separates them so you can, say, block training bots but stay visible in AI search. By default everything is allowed.
How much does it cost to use the OpenAI API?
Very little for typical use. With gpt-5.6-luna, generating 5 FAQs from a blog post costs a fraction of a cent. Even a high-traffic content site generating hundreds of FAQs per month would spend just a few dollars. You only pay when you click "Generate" — never in the background.
Is the FAQ schema valid?
Yes. The plugin outputs standard schema.org FAQPage JSON-LD that validates on the Schema.org validator and is consumed by Bing and the AI assistants built on its index. Note: Google retired FAQ rich results from Search (and removed FAQPage from its Rich Results Test) in 2026, so the markup no longer produces a Google visual snippet — today its value is page understanding and AI-search citation (AEO/GEO), not a Google rich result.
Does this work with Yoast SEO or RankMath?
Yes. This plugin works alongside Yoast SEO, Rank Math, and other SEO plugins. It adds its own FAQPage JSON-LD only for posts that contain this plugin's FAQ block. If the same post also has a Yoast, Rank Math, AIOSEO, Spectra or Kadence FAQ block that prints FAQ schema, the block sidebar warns you, and the Schema Output switch turns off this block's schema while the visible FAQ stays on the page. If you're only using Yoast or Rank Math for general SEO (title tags, meta descriptions, sitemaps), there's no conflict at all.
Does this work with the Classic Editor?
No — this plugin requires the WordPress Block Editor (Gutenberg). It is a native Gutenberg block and does not work with the Classic Editor plugin. If your site uses the Classic Editor, you would need to switch to the block editor to use this plugin.
What happens if I deactivate or delete the plugin?
Your FAQ content is always preserved in your posts — it never disappears. On deactivation, schema injection and the visual display simply pause until you reactivate, and all settings are kept. On deletion, plugin settings (API key, model, prompt, style preferences) are removed from the database per WordPress best practices, but the FAQ text in your posts stays.
Does the AI always produce perfect FAQs?
Not always — AI output quality depends on your content and prompt. That's why the built-in FAQ Quality Analyzer exists. Run it before you publish and follow the recommendations to improve your score. You can also edit any generated FAQ manually.
Is my content sent to OpenAI every time I load the editor?
No. Content is only sent to OpenAI when you click "Generate FAQ" or "AI Detect." Nothing is sent automatically in the background.
What is the difference between the SEO score and the GEO score?
The SEO score (out of 100) checks 12 traditional ranking factors: question format, answer length, direct first sentence, keyword placement, sentence length, duplicate detection, and so on. It tells you how well-structured your FAQ is for search engines and featured-snippet-style extraction. The GEO score (out of 100) checks 8 AI-citation factors that matter when ChatGPT, Perplexity, Google AI Overview, or Claude decides whether to quote you. These include: filler-word detection, citation-window length (8-60 words is the sweet spot AI assistants prefer), self-contained answers (no "as mentioned above"), specificity and E-E-A-T signals, pronoun ambiguity, and structured multi-sentence formatting. Both scores display side by side, and clicking either one shows a detailed per-FAQ breakdown.
Will this slow down my site?
No. Schema generation runs once per post when you save, then the JSON-LD is cached as post metadata. On every page view, the plugin reads the cache directly instead of re-parsing your block content. The frontend overhead is a single small <script type="application/ld+json"> tag in the page head — the standard schema.org format. The CSS for display styles is also enqueued only on pages that actually contain a FAQ block.
Does this work with non-English content?
Yes. The schema markup is language-agnostic — Google reads FAQPage schema correctly in any language. The AI generation also works in any language OpenAI supports (60+ languages); if your post content is in Korean, Japanese, Spanish, French, German, etc., the generated FAQs will be in the same language. The "I Have FAQ" paste mode also accepts FAQs in any language.
更新日志:
1.12.0
- Added: multilingual llms.txt for Polylang and WPML. The main file lists the default language and links to a file per language (for example /de/llms.txt) with that language's pages and URLs; llms-full.txt follows the same rule. Before, only one language was listed and the others were missing.
- Added:
mkjb_llms_txt_languagesfilter for other multilingual plugins. - Fixed: titles and summaries in llms.txt showed HTML entities such as
–instead of the dash. They are now plain text, and square brackets in titles no longer break the Markdown link. - AI Crawler Visits now also counts fetches of per-language files such as /de/llms.txt.
- Added: AI Crawler Visits under Settings - AI FAQ Schema. Counts requests from known AI crawlers (GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, PerplexityBot and more) for the last 30 days, with robots.txt and llms.txt fetches, last seen time, last page and the most-read pages. Data stays in one option in your database; a Reset button clears it, and a checkbox turns counting off. Matching is by User-Agent, and requests answered by a page cache never reach WordPress, so treat the numbers as a lower bound.
- Added: optional llms-full.txt (off by default) with the text of the same pages llms.txt lists, including this plugin's FAQ questions and answers. Capped at 20,000 characters per page and about 2 MB in total.
- Added: Duplicate FAQ schema guard - the block sidebar warns when the same post also has a Yoast, Rank Math, AIOSEO, Spectra or Kadence FAQ block that prints FAQPage schema, and lists active FAQ plugins that can print it too.
- Added: Schema Output switch per block - turn off this block's FAQPage schema while the visible FAQ stays on the page.
- Added: llms.txt contents - choose which post types are listed, exclude items by ID, URL or path, and add custom links. Items marked noindex in Rank Math, Yoast, SEOPress or All in One SEO, and password-protected items, are now always left out.
- Added: Custom robots.txt lines box under AI Crawler Access. Only valid robots.txt lines are kept.
- Added: settings now warn when a physical robots.txt or llms.txt file in the web root would override the generated one.
- Added: Question Heading Level (H2, H3, H4, H5 or plain paragraph) so the page outline stays in order under your FAQ title. H3 stays the default, so existing posts don't change.
- Changed: the llms.txt setting description no longer says AI crawlers use the file - support is not confirmed.
- Tested with WordPress 7.1.
- Added: Claude-SearchBot (Anthropic's search crawler) to the AI Crawler Control table - now 14 crawlers.
- Changed: removed outdated "Google Rich Results" wording from the SEO score card, the block description and the default AI prompt (Google retired FAQ rich results in 2026).
- Changed: description wording made more precise - no "only plugin" claims, no unsourced statistics, and llms.txt described as the proposed convention it is.
- Fixed: the internal version constant was stuck at 1.9.2, so browsers could keep serving old cached block scripts and styles after an update.
- Fixed: robots.txt rules are now also skipped when the "discourage search engines" setting is stored as an empty value.
- Fixed: llms.txt now uses your SEO plugin's meta description when a page has no excerpt. Pages used to fall back to trimmed page content, which could start with menu and button text.
- Plugin and author links now point to the plugin's own site, answerschema.com (guides on FAQ schema, llms.txt and AI crawlers).
- No functional changes.
- Fixed: the OpenAI model list still offered the retired 2024-2025 lineup (gpt-4o, gpt-4o-mini, gpt-4.1, o3, o4-mini...) - OpenAI shut these down during 2026, so Auto Generate and Fix with AI could fail with model-not-found errors. The dropdown now offers the current GPT-5.6 models (luna / terra / sol), the default is gpt-5.6-luna, and any model saved by an earlier plugin version is mapped automatically to its closest current cost-tier equivalent - no action needed.
- Fixed: untitled posts/pages appeared in llms.txt as an empty markdown link. They are now skipped.
- Added: llms.txt - the plugin now serves an auto-generated llms.txt (llmstxt.org convention) at your site root: a machine-readable index of your pages and latest 200 posts with summaries, so AI crawlers can discover and understand your content. Cached for 12 hours, rebuilt automatically when content changes, and never overrides a physical llms.txt file. No API key needed. Toggle under Settings - AI FAQ Schema.
- Added: AI Crawler Control - allow or block 13 known AI crawlers (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Perplexity-User, Google-Extended, Applebot-Extended, CCBot, Bytespider, meta-externalagent, Amazonbot) per bot via WordPress's virtual robots.txt. All allowed by default; each entry explains what blocking it actually does (e.g. blocking OAI-SearchBot removes you from ChatGPT Search). No API key needed.
- Changed: No API key required - repositioned the whole plugin around keyless use. The block now opens on the "I Have FAQ" tab when no key is configured (Auto Generate was a dead end without one), the Auto Generate tab explains that AI is optional, and the settings page marks the OpenAI key as optional with a clear list of everything that works without it.
- Added: a single, permanently-dismissible review request notice - shown only after FAQ sections have been published on 3+ posts, and only on the plugin's own screens.
- Fixed: The Customize Style panel (question/answer colors, backgrounds, border color and radius, font sizes, accordion default-open state, per-block custom CSS) and the FAQ Output Language setting silently reverted to defaults after saving the post. These attributes existed server-side but were missing from the client-side block registration, so the editor never serialized them. Both the main block and the legacy fsg/faq-schema block are fixed - re-save the post once and the settings now stick.
- Fixed: The FAQ schema cache was invalidated on every plugin update and never rebuilt until each post was manually re-saved, so all FAQ posts re-parsed their blocks on every page view. The cache version is now a separate format constant (only bumped when the cached structure actually changes) and stale caches self-heal on the first page view.
- Improved: The OpenAI API key is now encrypted at rest (AES-256-CBC, keyed off your site's AUTH_KEY) instead of stored as plain text in the options table. Existing keys are migrated to the encrypted form automatically on first use - no action needed. Note: if your site's AUTH_KEY salt is ever regenerated, simply re-enter the API key.
- Changed: Repositioned for AI search. Renamed to "AI FAQ Schema - AEO & GEO for AI Search" and refreshed tags (faq, faq schema, aeo, seo, ai). The plugin is now framed around Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) — getting cited by AI Overviews, ChatGPT, and Perplexity.
- Fixed: Corrected misleading FAQ entries that claimed the schema "passes Google's Rich Results Test" and helps qualify for "Google's FAQ rich results." Google retired FAQ rich results from Search (and removed FAQPage from the Rich Results Test) in 2026, so these claims are no longer accurate. The readme now states the real value: AI-search citation (AEO/GEO), page understanding, and schema consumed by Bing and Bing-indexed AI assistants.
- Added: New readme sections — "How to add FAQ schema in WordPress" and "Optimizing for AI Overviews and AI search."
- Improved: External Services disclosure now lists the Fix with AI (answer rewrite) call to OpenAI.
- Added: Fix with AI — a one-click button in the quality analyzer that rewrites every answer into a citation-ready answer capsule (direct Yes/No or subject-first opener, 40-80 words, at least one data point, fully self-contained). Questions and language are preserved; both SEO and GEO scores re-calculate instantly so you see the improvement. No other FAQ plugin scores quality AND fixes it.
- Improved: Plugin renamed to "AI FAQ Generator & Schema by MKJB" and directory tags refreshed (faq, ai, chatgpt, schema, seo) for better discovery — "rich snippets" removed as Google retired FAQ rich results in May 2026.
- Improved: Readme highlights that Auto Generate reads your post content directly from the editor — one click, no pasting.
- Fixed: Selecting the o3-mini, o3, or o4-mini model caused every generation to fail with an HTTP 400 error — reasoning models reject custom temperature values, so temperature is now only sent to standard chat models. Reasoning models also get a much larger completion budget (8,000 tokens) so internal reasoning never starves the visible answer.
- Fixed: Generating 7 or 10 FAQs could fail with "OpenAI returned an invalid response format" — the 2,000-token completion cap truncated the JSON mid-string. Cap raised to 4,000 tokens.
- Fixed: Custom CSS (Customize Style panel) containing more than one rule leaked outside the block — only the first selector was scoped to the block ID. Every rule and every comma-separated selector is now scoped.
- Fixed: Mentions of "the US" in answers were falsely flagged as first-person usage ("us") by the SEO analyzer.
- Improved: Default AI prompt rewritten around the GEO "answer capsule" format that measurably increases AI citation rates — answers target 40-80 words (never over 120), open with a direct Yes/No restatement or a subject-first definition sentence, must pass the "information island" test (fully understandable in isolation), include at least one specific data point, and questions are phrased the way people actually ask AI assistants.
- Improved: Quality Analyzer answer-length check tightened from 1,000 to 700 characters max, matching the answer-capsule guidance.
- Changed: Plugin description and readme repositioned for the AI-search era — Google removed FAQ rich results from Search entirely in May 2026, so AI citation (which this plugin has scored since 1.3.0) is now the primary value, with FAQPage JSON-LD retained for Bing and Bing-indexed AI assistants.
- Fixed: On WordPress 7.0 the block editor runs the FAQ block inside an iframe (a consequence of the apiVersion 3 upgrade shipped in 1.6.0). The iframe constrains the block to its real column width, and the editor's question/answer inputs were overflowing the rounded block wrapper and getting clipped on the right edge. Added strict
box-sizing: border-boxandwidth/max-widthsafety rules so every form control (TextControl, TextareaControl, SelectControl) stays inside the wrapper at any width. - Improved: Editor UI is now fully responsive — added tablet/mobile rules that tighten padding on the tabs, manual-input rows, and header so nothing is cramped or clipped when the iframed editor is narrow.
- Compatibility: Tested up to WordPress 7.0.
- Added: FAQ Output Language control — Auto Generate now produces FAQs in the same language as your post content by default (auto-detect). Previously, the AI always output English regardless of content language, which broke multilingual sites where posts in Korean / Spanish / Japanese / etc. ended up with English FAQs.
- Added: 18 explicit language overrides in the block sidebar (English, Korean, Spanish, French, German, Japanese, Chinese, Portuguese, Italian, Russian, Arabic, Hindi, Vietnamese, Thai, Indonesian, Turkish, Polish, Dutch) — useful when you want FAQs translated into a different language than the source content.
- Improved: Language directive in the prompt is injected with highest priority and explicitly overrides the default prompt's "US market" framing, so question starters (What/How/Why), Yes/No answer starters, and cultural references all match the target language.
- Compatibility: Existing blocks default to "auto" — they will continue to work, and on the next regenerate they'll match your post's language automatically.
- WP 7.0 ready: Block API bumped from version 2 to version 3 (in both the JS
registerBlockTypeand the PHPregister_block_typecalls). WordPress 7.0 enforces full iframe isolation for the block editor and deprecates apiVersion 1 and 2; this release pre-emptively switches the FAQ block to apiVersion 3 so it keeps working when 7.0 lands. No user-visible behaviour change — the block already useduseBlockPropsand onlywp.*globals (nodocument.querySelectoretc.), so iframe isolation has no impact.
- Fixed: Inline CSS fallback now also covers FAQ markup that was saved as static HTML in the post body (a common state on multilingual sites where translation plugins copy the rendered HTML from the original post into the translated copy, stripping the block markup along the way). The 1.5.0 fix only covered dynamic block rendering via render_callback; this release adds a
the_contentfilter at priority 100 that scans for the FAQ block class and injects the inline stylesheet when needed. End result: FAQ styling now reaches the browser whether the block is dynamic, static HTML, translated, or pasted from another post.
- Fixed: Plugin CSS not loading on some pages — frontend FAQ blocks could appear unstyled when caching plugins (LiteSpeed Cache), translation/multilingual plugins, page builders, or path-obfuscation plugins (e.g. Hide My WP) suppressed the regular stylesheet enqueue. The block CSS is now emitted inline once per page render as a fallback, so styles always reach the browser regardless of the optimization stack.
- Added: Customize Style panel in the block sidebar with 10 per-block controls — Question/Answer text colors, Question/Answer backgrounds, Border color, Border radius, Question/Answer font sizes, Accordion default open state (all collapsed / first open / all open), and a Custom CSS textarea scoped to the block.
- Added: Per-block customize values render as inline CSS custom properties on the block wrapper, giving them the highest specificity — themes and caching plugins cannot override them.
- Improved: Accordion items can now start in an open state by default, useful for highlighting the most important question on landing pages.
- Renamed: Plugin display name changed from "MKJB AI SEO FAQ Schema" to "AI FAQ Schema by MKJB" for clearer discovery (keywords up front, brand preserved). Slug, settings, FAQs, API keys, and licensing are unchanged — this is a display-only rename and existing installs are unaffected.
- Improved: Updated tags and short description for better WordPress.org discoverability around AI search and rich-snippet keywords.
- Added: Three new FAQ entries — SEO vs GEO score difference, performance/caching behavior, and multilingual support.
- Cosmetic: All admin labels (Settings menu, error messages, on-screen hints) now read "AI FAQ Schema" instead of the old name.
- Improved: "I Have FAQ" tab now auto-detects multiple paste formats — no more strict Q:/A: only. Accepts Question:/Answer:, numbered lists ("1. ..."), bracketed labels ([Q]/[A]), markdown headings, bold-wrapped labels (Q:), and plain text where each question simply ends with "?".
- Improved: Pasting from HTML sources (blog posts, accordion widgets, <dt>/<dd> lists, <details>/<summary> elements) now strips tags and entities before parsing.
- Improved: Markdown decorations (#, ##, ###, >, -, , bold*) are stripped automatically so copy-pastes from Notion, Google Docs, and Markdown editors parse cleanly.
- Improved: Updated paste-tab help text and placeholder to show all supported formats with concrete examples.
- Fixed: Numbered style layout — answer text was breaking into single-word lines due to narrow CSS grid column width.
- Fixed: Updated OpenAI model list to currently available models (gpt-4o, gpt-4.1 series, o3, o4-mini). Replaced non-existent gpt-5.4 references that caused API errors.
- Fixed: Multi-line answers (numbered lists, bullet points) now preserve their line breaks on the frontend instead of being collapsed into a single line.
- Fixed: Rate limiter now uses a proper fixed 60-second window instead of a rolling window that could extend indefinitely.
- Fixed: Accordion heading structure improved for screen-reader accessibility.
- Fixed: Defensive length caps on individual question (300 chars) and answer (2,000 chars) to prevent runaway AI output.
- Improved: Schema injection is now cached per post on save — no more parsing blocks on every page view, significantly faster for high-traffic sites.
- Improved: Removed unnecessary admin-only data from the block editor script payload.
- Added: Dark mode support across all display styles — automatically follows visitor's OS preference.
- Added: Drag-and-drop reordering and Duplicate button in the Manual Input tab.
- Added: 2 new display styles — Chat Bubble (messenger-style) and Numbered (tutorial-style).
- Added: GEO Quality Analyzer with 100-point scoring for AI search citation optimization
- Added: Dual score display — SEO and GEO scores shown side by side with tabbed breakdown
- Added: 8 GEO checks: filler detection, citation length, self-contained answers, specificity, E-E-A-T signals, pronoun ambiguity, promotional fluff, structured format
- Improved: Default AI prompt now includes explicit GEO rules for AI citation-ready FAQ generation
- Improved: AI prompt separates SEO rules and GEO rules for clearer optimization
- Fixed: CSS not loading on frontend for posts using legacy block name
- Fixed: Broken unicode characters in CSS causing style rendering issues
- Added: 2 new display styles - Boxed and Timeline
- Improved: All existing styles (Minimal, Card, Accordion) with premium visuals
- Improved: Responsive design for mobile devices
- Fixed: Backward compatibility for posts created with plugin v1.2.1 or earlier (fsg/faq-schema block)
- Added: Legacy block registration so existing FAQ blocks continue to work after update
- Added: Schema injection now detects both old (fsg/) and new (mkjb/) block names
- Updated: OpenAI model list (Note: the gpt-5.4 references in this release were not real models and were corrected in 1.3.1 — please update to 1.3.1 or later if you are still on 1.2.2.)
- Improved: Plugin description, FAQ section, and tags for better discoverability
- Fixed: Block not appearing in WordPress block inserter when searching "/faq"
- Fixed: Block category updated from deprecated 'common' to 'text'
- Fixed: Added apiVersion 2 for proper useBlockProps support
- Fixed: Removed deprecated wp-editor script dependency
- Fixed: Removed all non-ASCII characters from JavaScript to prevent encoding issues on some server configurations
- Fixed: Updated deprecated isPrimary/isSecondary Button props to the current variant API
- Added: Block supports configuration (html: false, multiple: true)
- Improved: Version bump for reliable cache busting
- Added: "Remove All FAQs" button in I Have FAQ and Manual Input tabs
- Added: Score always displayed as X/100 (normalized when no keyword is set)
- Added: First-person usage detection (We/Our/Us/My) in Quality Analyzer
- Added: Sentence length check (20-word limit per sentence)
- Changed: Answer length requirements updated to 100-1,000 characters
- Changed: FAQ questions rendered as H3 for correct heading hierarchy
- Changed: Accordion questions use H3 inside summary element for accessibility
- Changed: Default AI prompt updated to enforce plain text formatting
- Fixed: Browser cache preventing updated JS/CSS from loading
- Improved: Quality Analyzer v2 rules aligned with SEO and GEO best practices
- Added: OpenAI model selection (gpt-4o-mini through o3)
- Added: Custom AI prompt with {count} placeholder
- Added: FAQ Quality Analyzer v2 with 100-point scoring
- Added: Target keyword input with AI auto-detect
- Added: FAQ title display option (H2/H3/H4)
- Added: Delete API Key button
- Added: Rate limiting (10 requests/minute/user)
- Added: Nested block support for schema injection
- Fixed: API key overwrite bug when saving other settings
- Fixed: Schema returning null on empty FAQ arrays
- Fixed: CSS double-loading on frontend pages
- Fixed: has_block() parameter error
- Improved: Modern CSS with custom properties
- Improved: Security hardening for public distribution
- Initial release