FocusWeb AI Visibility

By Uri N.

★ 5.0/5 (1 ratings)20+ active installationsVersion 1.6.0

AI SEO & GEO toolkit: llms.txt, schema & AI bot analytics — see which AI systems (ChatGPT, Gemini, Perplexity, Claude) read and cite your site.

Description

68% of Google searches now end without a click. Your future visitors are AI assistants and agents — and most WordPress sites are invisible to them. AI Visibility is an AI SEO / GEO (Generative Engine Optimization) plugin that makes your site readable, quotable and usable by ChatGPT, Gemini, Perplexity, Claude and every other AI system — and shows you exactly when they do:

  • Dynamic llms.txt — an always-up-to-date, machine-readable map of your key pages and content, following the llms.txt convention adopted by Anthropic, Vercel and Cloudflare.
  • Baseline structured data (JSON-LD) — WebSite and Article schema, automatically deferred when Yoast, Rank Math or AIOSEO is active.
  • Agent endpoint (read-only) — a clean JSON API (/wp-json/focusweb-ai-visibility/v1/ask) that lets AI agents search your published content, plus registration with the official WordPress Abilities API so MCP clients can discover it.
  • AI agent analytics — see which AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and more) visit your site, which pages they fetch, how often llms.txt is read, and what questions agents ask — with a daily color-coded chart.
  • Per-page schema control — every post/page shows its generated JSON-LD with an editable custom override, plus a site-wide force mode.
  • Google Analytics integration (optional, off by default) — fire an ai_referral event (source / medium / campaign) when visitors arrive from ChatGPT, Gemini, Perplexity, Claude and other AI chats. Uses your existing GA tag, or loads gtag.js with a Measurement ID you provide.
  • Bot Log export — a broader crawler feed (search engines, SEO tools, social link-preview bots, plus every AI bot above) exportable as a plain-text Combined Log Format .log file, for hosts where raw server access logs aren’t available. Never logs ordinary human visitors.
  • AI Insights Export — want more tailored recommendations for your site? Put your own AI agent to work: export a comprehensive data log (real questions asked, schema status, full per-visit AI bot fetch history) plus a ready-made instructions file, hand both to ChatGPT, Claude, Gemini or any AI agent you use, and get focused, actionable recommendations on content gaps, schema fixes, FAQ ideas, and pages AI bots aren’t finding. The plugin never sends your data anywhere itself; you decide which AI tool sees it.
  • Content Signals — declare your site’s preferences for search indexing, live AI-answer citations, and AI model training directly in robots.txt (the emerging contentsignals.org standard), with sensible defaults: be found and cited, opt out of training.
  • MCP tools for AI Site Manager (optional, off by default) — when FocusWeb’s AI Site Manager MCP gateway is active, offer AI agents you have connected to it nine tools: AI-bot analytics, per-page bot fetches, llms.txt (read and regenerate), page schema (read and edit), and the AI Insights data and analysis instructions — so an agent can analyse your site directly. Every call runs as the WordPress user the agent’s API key is bound to, and each tool can be switched off inside AI Site Manager.

  • Operating mode — choose Full optimization (the default: the plugin serves llms.txt, outputs schema, adds the Content-Signal line and answers AI agents) or Monitor only (monitoring only: the plugin records which AI systems visit and changes nothing on your public site; monitoring itself is always on in both modes). The current mode is shown on every plugin page.

Full RTL and Hebrew support.

External Services

This plugin connects to Google services only when the site administrator explicitly enables the Google Analytics integration (disabled by default):

  • gtag.js from googletagmanager.com — loaded on the front end only in “manual” mode, using the Measurement ID the administrator provides, to send ai_referral events to the site’s own Google Analytics property.
  • GA4 Measurement Protocol (google-analytics.com/mp/collect) — used server-side only when the administrator clicks “Fire sample event” on a page/source/title the administrator chooses, sending one or more test events (optionally session_start and page_view alongside ai_referral, to simulate a full visit) to the site’s own GA4 datastream using the Measurement ID and API Secret the administrator provides.

Data sent consists of the event name(s) and their parameters (source, medium, campaign, page URL, page title). No data is ever sent to the plugin author or any other third party. See Google’s Privacy Policy and Google Analytics Terms of Service.

FAQ

What is llms.txt, and what is GEO / AI SEO?

llms.txt is a plain-text convention (adopted by Anthropic, Vercel and Cloudflare) that gives AI systems a machine-readable map of your site. GEO (Generative Engine Optimization) — sometimes called AI SEO — is the practice of structuring your content so AI assistants like ChatGPT, Gemini, Perplexity and Claude can read, understand and cite it correctly. This plugin automates both.

Does this replace my SEO plugin?

No. It complements it. AI Visibility skips schema output automatically when a major SEO plugin is active, and adds the AI-specific layers no SEO plugin covers.

What is Monitor only mode?

A setting (AI Visibility → Settings & Status → Operating mode) for sites that want to see what AI systems do before changing anything. In Monitor only mode the plugin does not serve llms.txt (the address answers 404, but the request is still recorded), outputs no schema, adds nothing to robots.txt and does not register the agent endpoint or the Ability. AI bot analytics, the bot log, the dashboard and Google Analytics AI-referral events keep working. Switch to Full optimization at any time; no settings or data are lost. New installs start in Full optimization.

Is the plugin available in my language?

The admin screens are translated into Hebrew, Spanish, Portuguese (Brazil), German, French, Italian and Russian, and are shown automatically when your site uses one of those languages (or a regional variant of it). The initial translations were created with AI assistance and have not all been reviewed by native speakers, so some wording may need polishing. Corrections are very welcome: please post them in the plugin’s support forum, or join in at translate.wordpress.org.

Can I choose the schema type for a page?

Yes. Every page and post has an AI Visibility — Schema box with a Schema type menu (also available per row in the Schema Browser). Leave it on Automatic and the plugin picks the type for you; choose another to describe the page as an About page, a Contact page, a list page, an article, a product or a service. Product and Service describe what the page is about without inventing prices or reviews. For a FAQ, see the next question; for anything else the menu does not offer, use the Custom JSON-LD option in the same box.

How do I add FAQ schema to a page?

Open the page’s AI Visibility — Schema box, tick Add a FAQPage to this page, and either press Detect from the live page or write the questions and answers yourself. Detection reads a FAQPage the page already outputs (for example from a page-builder widget) or the page’s own accordion / blocks, and shows the questions for you to review before you save. The FAQPage is added next to the page’s other schema — the plugin’s own, your SEO plugin’s or your custom JSON-LD — and linked to the page. Use only questions and answers that are visible on the page, and if the page already outputs a FAQPage, switch that one off first so the page does not have two (Check live page tells you).

To confirm it worked, press Check live page: it states whether the page has a FAQPage, how many questions it contains, which plugin or builder outputs it, and whether it is complete (every question has its text and an answer).

Does the plugin work with Yoast SEO, Rank Math and All in One SEO?

Yes. The plugin does not replace their schema: it leaves it in place and only adds what you ask for (for example a FAQPage). It also reads the page type those plugins set, so the Automatic choice in the Schema type menu matches theirs, and it uses their titles, descriptions and noindex settings for llms.txt. It only reads from them; it never changes their settings.

How do I check the whole site at once?

Open AI Visibility → Schema Browser and press Scan site. The plugin loads each published page and post in turn (up to 100) and lists duplicate schema, broken FAQ schema, unreachable pages and pages that look like they have a FAQ but no FAQPage. Pages protected by a password or blocked by a firewall are reported as such, never as fine. It only reads your pages; it changes nothing.

How do I find duplicate schema on a page?

Open the page’s AI Visibility — Schema box (or the Schema Browser) and press Check live page. The plugin loads the published page, lists every schema block it finds and who outputs it, and flags a WebPage, FAQPage, Article, WebSite, BreadcrumbList or Organization that appears twice, or the same @id defined twice. Typical cause: an SEO plugin, a page-builder widget and this plugin all outputting the same thing. Switch the schema off in one of them, or use Custom JSON-LD to replace all schema on the page. The check only reads your page; it changes nothing.

Does the agent endpoint expose private content?

No. Only published, public content is ever returned. All endpoints are read-only in the free version.

Do I need to configure anything?

No. Activate the plugin and llms.txt, schema, the agent endpoint and bot logging work immediately.

Does the Bot Log store visitor IP addresses?

It stores the IP address of matched crawlers/bots only (search engines, SEO tools, social link-preview bots, AI bots) — it never logs ordinary human visitors. Disable it any time from AI Visibility → Bot Log, where you can also set the retention window or clear the log immediately.

What is the Content-Signal directive, and does it block AI crawlers?

It’s an emerging robots.txt extension (contentsignals.org) that separates crawl access from usage: instead of only saying “you may fetch this page,” it also says what the content may be used for — search indexing, live citation in AI answers, or AI model training. It’s a declared preference, not an access control — it doesn’t block any crawler by itself, and no major AI crawler is confirmed to enforce it yet, but some crawlers (including Cloudflare-managed bots) already read it. Configure it from AI Visibility → Settings & Status; defaults to allowing indexing and AI citations while opting out of training.

How does llms.txt handle multilingual sites?

Each page is tagged with a language, found in this order: Polylang or WPML if you use them; a language code you set on the page (the “AI Visibility — Language” box); a slug-suffix rule you define in Settings & Status (for example -fr=fr); a language code at the end of the slug (such as about-fr) when there is evidence it is right — the page text is in that language’s writing system, or a page with the same slug minus the suffix exists; the page’s own writing system for Armenian, Hebrew, Greek, Georgian and Thai; and finally the site language. When more than one language is in use, llms.txt lists them in its header, groups the page lists per language, and links each page to its translation. Pages that can’t be linked automatically can be paired by hand in Settings & Status.

Does the plugin support MCP?

Yes, in two ways. First, it registers its content search as an ability with the WordPress Abilities API (WordPress 6.9+), which WordPress’s MCP adapter can expose to MCP clients. Second, an optional setting (AI Visibility → Settings & Status → MCP) adds nine AI Visibility tools to FocusWeb’s AI Site Manager MCP gateway when that plugin is active: AI-bot analytics, per-page bot fetches, llms.txt (read and regenerate), page schema (read and edit) and the AI Insights data log and analysis instructions. It is off by default. The gateway handles connections and authentication; every tool call runs with the permissions of the WordPress user the agent’s API key is bound to, and individual tools can be switched off in AI Site Manager. This plugin never calls an AI service itself.

Screenshots

Changelog

1.6.0

  • New: the plugin follows your SEO plugin’s page type. When Yoast SEO, Rank Math or All in One SEO is active, the Schema type menu’s “Automatic” choice now shows what that plugin says the page is (for example AboutPage, or BlogPosting for a post) instead of guessing, with a note naming the plugin. Read-only: the plugin never changes your SEO plugin’s settings.
  • New: Site scan, in the Schema Browser. One button checks the live page of each published page and post (up to 100, filterable) one after another and lists what needs attention: duplicate schema, a broken or empty FAQPage, a page that could not be loaded, and pages that show questions (an accordion, details blocks) but have no FAQPage. Results are stored with each page and shown again in the Schema Browser rows. It only reads your pages.
  • New: “Check live page” now also shows what the page is described as on the live page and which plugin outputs it, and offers a FAQPage when it finds FAQ-style content without one.

1.5.0

  • New: a schema type for every page. The plugin now describes each page automatically (pages as WebPage, posts as Article, the blog index and archives as CollectionPage) and every node has an @id other schema can link to. In the page’s AI Visibility — Schema box and in the Schema Browser you can change the type by hand: WebPage, AboutPage, ContactPage, CollectionPage, Article, BlogPosting, NewsArticle, Product or Service. Product and Service add a node for the thing the page is about, with its brand or provider linked to your Organization. A type chosen by hand replaces your SEO plugin’s schema on that page; pages left on Automatic behave as before. Custom JSON-LD still overrides everything.
  • New: FAQ schema for any page. A FAQ section in the page’s AI Visibility — Schema box adds a FAQPage built from questions and answers you write or detect from the live page (the page’s own FAQPage schema, an accordion, blocks). It is an add-on printed next to the page’s other schema whatever its source, linked to the page by @id, and shown in the Schema Browser (Edit FAQ button).
  • New: duplicate-schema check. “Check live page” (in each page’s AI Visibility — Schema box and in the Schema Browser) loads the published page, reads every JSON-LD block on it — whether it comes from this plugin, Yoast SEO, Rank Math, All in One SEO, a page builder or a theme — and tells you when the same thing is described twice (two WebPage or FAQPage nodes, the same @id defined twice, an unreadable block) and which source outputs each one. The result is shown next to the page and cleared when its schema changes. A page that cannot be loaded (password, firewall, timeout) is reported as such, never as “no duplicates”.
  • New: Check live page also verifies FAQ schema — it reports whether the page has a FAQPage, how many questions it holds, who outputs it, and whether each question has its text and an answer. The Custom JSON-LD area of the page editor is now collapsed (it opens by itself when the page uses custom JSON-LD).
  • New: short explanations in the Schema Browser of how automatic types, a chosen type, several schemas in one custom @graph and duplicates work, with an example.
  • A site’s home page is always described as a WebPage, also when it shows the latest posts; the blog page and archives are CollectionPage. The plugin’s own schema block is now marked (class “focuaivi-schema”).
  • Fix: the translations bundled with the plugin (Hebrew, Spanish, Portuguese, German, French, Italian, Russian) were not being loaded, so the admin stayed in English unless WordPress.org had a language pack for your language. They now load automatically, including regional variants (for example es_MX uses Spanish, pt_PT uses Portuguese).
  • Hebrew: “Custom override” is now “מותאמת אישית” instead of “דריסה”.
  • The author link now points to the English FocusWeb site.
  • readme: new FAQ entry about the available languages, noting that the initial translations were AI-assisted and that corrections are welcome.

1.4.0

  • New: operating mode. Choose Full optimization (default, unchanged behaviour) or Monitor only, which records AI activity but publishes nothing: no llms.txt, no schema, no robots.txt line, no agent endpoint or Ability. The mode is chosen at the top of Settings & Status and shown as a badge on every plugin page; switching tells you exactly what changed and loses no data.
  • In Monitor only mode the dashboard hides the optimization score, the llms.txt and schema checks show “Paused” instead of failing, and the AI Site Manager tools that publish (regenerate llms.txt, edit schema) refuse with a clear message; aivis_status reports the mode.
  • Translations reviewed and completed in all seven languages (Hebrew, Spanish, Portuguese, German, French, Italian, Russian): the plugin description now matches the translated header, a wrong page title in Hebrew and Russian is fixed, the Russian text was rewritten in places, and German, Spanish and Hebrew now address you consistently. Unused strings were removed.

1.3.12

  • New: the Dashboard now states in one sentence whether AI systems can reach your site and, when something needs attention, links straight to the fix.
  • New: llms.txt is sent with an explicit one-hour Cache-Control lifetime (filterable with focuaivi_llms_txt_max_age). Some hosts and CDNs otherwise add a 30-day lifetime to .txt files, so crawlers could keep an outdated copy for a month.
  • Change: very long SEO titles (keyword-stuffed Yoast or Rank Math titles of 70+ characters) are shortened in llms.txt — the page’s own shorter title is used, or the title is cut at the last separator. Filterable with focuaivi_llms_title_max_chars.
  • Fix: llms.txt could cause a fatal error on hosts that do not have the PHP mbstring extension (WordPress only provides two of the functions it used). It now falls back safely and never splits non-Latin letters.
  • Fix: more probes for config and key files (.toml, .htaccess, .pem, .key, .properties, backups such as wp-config.php.bak) are no longer counted as AI-bot page visits, and the REST/API calls that older versions had logged are removed from the history once, on update.
  • Fix: uninstalling now removes everything the plugin stored (settings, language list, diagnostics, transients, per-page schema and language fields, and the review-notice flags), including on multisite.
  • New: one dismissible request for a review on WordPress.org, shown only to administrators on this plugin’s own pages, after two weeks and once AI activity has been recorded. “No thanks / already did” hides it for good.
  • Internal: the admin screens’ helper code moved into two small classes, and every automated check now runs from a single command.

1.3.11

  • New: MCP tools for AI agents, as a separate capability you can switch on or off (AI Visibility → Settings & Status → MCP; off by default). When enabled and FocusWeb’s AI Site Manager MCP gateway is active, agents connected to it get nine tools: aivis_status, aivis_llms_get, aivis_llms_regenerate, aivis_bot_activity, aivis_page_fetches, aivis_schema_get, aivis_schema_set, aivis_insights_data and aivis_insights_instructions. Analytics and llms.txt tools require the Administrator capability, schema tools require permission to edit that page, and none is destructive. The gateway’s own authentication, per-user permissions, per-tool switches and audit log apply.
  • New: language-aware llms.txt. Each page’s language is detected from Polylang or WPML, a per-page language field (new “AI Visibility — Language” box in the editor), a slug-suffix rule you define, a language code at the end of the slug (such as about-fr, only when the page text is in that language’s writing system or a page with the same slug minus the suffix exists), the page’s own writing system for Armenian, Hebrew, Greek, Georgian and Thai, and finally the site language. When more than one language is in use, llms.txt lists them in its header, groups Key pages and Recent content per language, and links each page to its translation (“Other language: “) from Polylang/WPML, matching slugs or slug pairs you define. The settings page lists the languages detected and draws its examples from them. /wp-json/focusweb-ai-visibility/v1/site-info now also returns the languages and each listed page’s language and translations.
  • Change: one description rule for every llms.txt entry, never combined: the SEO plugin’s meta description (Yoast, with template variables such as %%sitename%% resolved, else Rank Math, else AIOSEO); else the manual excerpt; else an automatic ~160-character excerpt of the rendered page. Page-builder noise no longer leaks in: style, script, svg and template blocks are removed with their contents, as are stray CSS rules, leftover shortcodes and repeated UI text such as “Close Close Close”. A meta description written in a different script from the page itself (for example an English description copied onto an Armenian page) is skipped in favour of the page’s own text.
  • Change: llms.txt titles use the SEO title (with the trailing site name removed — a setting, on by default), else a cleaned page title: HTML entities decoded, a language suffix such as “-hy” removed, and ALL-CAPS titles converted to Title Case.
  • New: “Regenerate llms.txt now” (Settings & Status) rebuilds llms.txt immediately instead of waiting for the 12-hour cache, through a real front-end request so page-builder content renders as it does for visitors.
  • Fix: llms.txt no longer lists password-protected pages or content marked noindex (Yoast including its per-post-type default, Rank Math, SEOPress), and the lists stay full after exclusions. Sites can add rules with the focuaivi_llms_exclude_post filter. The cache is also flushed when SEO meta, a password or a status changes without a full post save.
  • Fix: /llms.txt is no longer redirected to /llms.txt/ on sites whose permalinks end in a slash.
  • Fix: Abilities API registration on WordPress 6.9+ (correct hook, category, MCP meta). The search-content ability was registered on a hook name WordPress 6.9 never fires, so MCP clients saw nothing; it is now registered on wp_abilities_api_init, in its own category (focusweb-ai-visibility), with an output schema and the meta the WordPress MCP Adapter and other bridges need to expose it and to know it is read-only. Nothing changes on WordPress 6.4–6.8.
  • Fix: llms.txt could contain broken characters on sites in Hebrew, Armenian or Russian — a description trimmed for length could split a letter in half. WordPress then refused to store the file in its cache, so llms.txt was rebuilt (rendering every listed page) on every request. Descriptions are now trimmed by character, and llms.txt is checked for valid UTF-8 before it is cached.
  • Fix: AI agents’ REST and MCP API calls (for example /wp-json/…) and probes for config files (.yml, .yaml, .env, …) are no longer counted as AI-bot page visits, which had inflated “AI bot visits” and “most-fetched pages”.
  • Plugin Check: removed a redundant suppress_filters argument, documented the core and WPML hook names the plugin calls on purpose, and set “Tested up to” to 7.1.

1.3.3

  • Marketing refresh: banner and plugin description now lead with the plugin’s real differentiator — visibility, citation, and proof (real AI bot analytics), rather than a bare feature/jargon list.
  • Removed “NLWeb” from the WordPress.org banner — it was never an actual feature of this plugin.

1.3.2

  • “Fire sample event” (Settings & Status → Test the integration) now sends a full simulated visit — session_start + page_view + ai_referral sharing one client/session ID — so it shows up in Google Analytics under Traffic Acquisition, not only as an isolated custom event under Events. Configurable AI source, page URL and page title; the old single-event mode is still available via a checkbox.
  • Fix: self-test and sample-event results were stored as pre-translated text, so a diagnostic message could stay frozen in an old language after the site’s admin locale changed (e.g. an English-locale page showing a Hebrew message from a past test). Results are now stored as a status code and translated fresh on every page load, matching the rest of the page.

1.3.0

  • New Content Signals setting (AI Visibility → Settings & Status): adds a Content-Signal directive to robots.txt declaring this site’s preferences for search indexing, live AI-answer citations, and AI model training (the emerging contentsignals.org standard). A declared preference, not an access control — configurable per category, on by default with sensible defaults (allow indexing and citations, opt out of training), and automatically skipped when the site itself is marked non-public under Settings → Reading.
  • Settings page shows a live preview of the exact line that will be added to robots.txt, plus a fetch of your site’s actual current robots.txt (cache-busted, 5-minute cache) so you can confirm the line really appears — and catch the case where a physical robots.txt file on your server overrides WordPress entirely.

1.2.0

  • New “Insights Export” page, producing two separate Markdown files: a data log (real questions asked through the agent endpoint, per-page schema status, and — new — a full per-visit AI bot fetch history with individual timestamps, not just aggregated counts) and a separate instructions file (the analysis prompt for the AI assistant: site domain, explicit date range, which data sections are included). Hand both to ChatGPT, Claude, Gemini or any AI agent for content, FAQ, and schema optimization recommendations grounded in real evidence rather than a thin summary.
  • Insights Export never calls an AI API itself — it only produces files the site owner hands to whichever AI tool they already use.
  • Page scope (radio buttons): top 10/25/50 pages by AI bot activity, pages modified in the last 30/90 days, or all published content. Data sections (checkboxes): real questions asked, per-page schema status, full AI bot visit history — pick any combination.
  • Plugin Check: rebuilt the AI-instructions text without heredoc syntax (<<<), which Plugin Check disallows outright — now built as an array of lines joined with implode(), same output, no functional change.

1.1.0

  • New “Bot Log” page: a broader crawler/bot detector (search engines like Googlebot/Bingbot/DuckDuckBot/YandexBot/Baiduspider, SEO tools like AhrefsBot/SemrushBot, social link-preview bots, plus every AI bot already tracked elsewhere) logging full request details — IP, method, path, protocol, status, bytes, referrer, raw user-agent — to a dedicated table, exportable as a plain-text .log file in Combined Log Format. Built for sites where the host doesn’t expose raw server access logs, so external log-analysis tools can still get a bot-traffic feed without server access. Deliberately bot/crawler-only: never logs ordinary human visitors, keeping the privacy footprint far smaller than a full access-log capture would be.
  • Bot Log capture only starts an output buffer (to measure response size) for requests already matched to a known crawler — zero added overhead for regular visitor traffic.
  • Bot Log has its own configurable retention window (default 30 days, independent of the 7-day free-tier window on the AI-bot analytics table), a manual “Clear log now”, and an off-by-default “trust X-Forwarded-For” option for sites behind a reverse proxy/CDN.
  • Bot Log: added a “Bytes transferred” column to the crawler breakdown, and the on-disk size of the log table to the retention summary — informational only, for keeping an eye on storage footprint.
  • Translations: all 26 new Bot Log strings translated into all 7 bundled languages, verified with the same tokenizer-based audit as 1.0.0 — zero missing, zero stale.
  • Plugin Check: restructured the Bot Log export query to build each SQL statement as a literal string passed directly to $wpdb->prepare() (was assembling it in an intermediate variable first, which the static analyzer can’t verify as safe even though it was); documented the remaining identifier-interpolation cases (table names can’t be bound via $wpdb->prepare(), only values can) as accepted, matching the same pattern already used elsewhere in the plugin; dropped a Tags: entry that pushed the readme over WordPress.org’s 5-tag limit.

1.0.0

  • Dashboard redesign: added an “AI Visibility Score” gauge (llms.txt reachability + schema validity + recent activity), a KPI stat-tile row (AI agent visits, llms.txt reads, referral events, questions asked), replaced the Status table with colored badge cards, moved to a two-column layout (charts left, tables right), shortened the intro paragraph with a “Learn more” details toggle, linked “Structured data” straight to Schema settings, added shortened links with Copy buttons for llms.txt/agent endpoint, and an “Upgrade for full history” link on the agent activity chart.
  • Schema Browser: added a real “Schema Type” column (the actual @type — Article, WebPage, etc. — the old “Type” column only ever showed the WordPress post type), a “Rich results” column flagging pages missing fields Google looks for (image, author, logo…), and radial % gauges at the top for Valid schema, Rich-result ready, and AI bot fetch coverage (replacing the old plain-text stats line).
  • Schema generator: the baseline Article/WebPage schema now includes the featured image (required for Google’s Article rich-result eligibility), and the Organization node includes the site icon as logo (recommended for Knowledge Graph) — both were silently missing before, which the new Rich-result checker would otherwise have flagged on every single page.
  • Fixed a real schema-correctness bug the new Rich-result checker caught: WebPage nodes (i.e. every plain “page”, as opposed to a post) were using the Article-specific headline property instead of the name property that schema.org’s WebPage type and Google’s own guidance actually expect. Since automatic-mode schema is rebuilt fresh on every page load (never cached), this self-corrects immediately on update — no re-saving needed. Only affects custom overrides someone copied from the old auto-generated output as a starting point.
  • Schema Browser: added an “AI bot fetches” column cross-referencing the agent log against each page’s URL — a bold count per page, with its time frame stated directly in the column header (“last N days”), expandable into a line-level log of every individual fetch (bot + timestamp, most recent first, with its own Bot/Date-time header). Since JSON-LD ships embedded in the page HTML, a logged fetch is direct proof that page’s schema was delivered to the bot — it intentionally does not claim the bot parsed or used it, since no plugin can observe that. Includes both a per-page “Delete history” reset and a site-wide “Delete history for all pages” reset (each with a confirmation prompt).
  • Fixed AI bot fetch logging counting static assets and infrastructure files (stylesheets, scripts, robots.txt, sitemap XML, etc.) as page visits on hosting setups that route every request through WordPress — these no longer inflate “Most-fetched URLs”, the AI-visit KPIs, or the Schema Browser’s per-page fetch counts.
  • Schema Browser: color-coded the Mode column’s “Automatic” state (previously plain text, now consistent with the existing Custom override/Invalid coloring).
  • Plugin Check: readme’s top-level plugin name now matches the Plugin Name header exactly (was carrying extra marketing text, which Plugin Check flags as a mismatch); documented the legacy-prefix migration’s table RENAME/INSERT/DROP statements as an accepted identifier-interpolation case (table names can’t be bound via $wpdb->prepare(), and both names are already esc_sql()‘d); documented the referral date-range form’s $_GET['page'] read as an accepted read-only case (same as the existing $_GET['focuaivi_ref_days'] filter on the same form — display-only, no state change).
  • Translations: all 7 bundled languages (Spanish, Brazilian Portuguese, German, French, Italian, Hebrew, Russian) fully updated for the ~30 new strings introduced by the Dashboard redesign and Schema Browser overhaul, and every remaining stale/orphaned string from older UI copy (pre-rename artifacts, retired Pro-teaser text) has been removed. Verified with a full tokenizer-based audit against the current source: zero missing, zero stale across all 7 languages.

0.9.5

  • Added Russian (ru_RU) translation, contributed via translate.wordpress.org and verified against the current source code (zero missing or stale strings — the most precise of the 7 bundled languages).

0.9.4

  • Translations: fixed two strings whose text still referenced the pre-rename aiv_sample_test campaign value instead of focuaivi_sample_test, and added the “Missing or invalid nonce.” string that was never translated, across all 6 languages. Verified with a full audit against the actual source code (zero missing strings) and cross-checked all languages cover the identical string set.

0.9.3

  • Added full German (de_DE), French (fr_FR) and Italian (it_IT) translations (134 strings each) — all 5 planned languages (Spanish, Brazilian Portuguese, German, French, Italian) are now complete, alongside the existing Hebrew translation.

0.9.2

  • Added full Brazilian Portuguese (pt_BR) translation (134 strings) — 2nd of 5 planned languages. Spanish and Hebrew translations updated to match (134 strings each).
  • Moved the AI referral events section (per-source breakdown, events log, reset control) from Settings & Status to the Dashboard, next to the other activity charts.
  • Added a color-coded daily chart for referral events (reusing the same chart style as the crawler-activity and llms.txt charts), a date-range selector (1/3/7 days), and colored swatches on the per-source breakdown so its colors match the chart’s legend. Expanded the shared chart palette from 10 to 15 colors for sites with many distinct referral sources.

0.9.1

  • Added full Spanish (es_ES) translation (132 strings) — the first of 5 planned languages (Spanish, Brazilian Portuguese, German, French, Italian) chosen by WordPress.org install-base ranking.

0.9.0

  • AI referral events (Settings & Status): added a per-source breakdown (chatgpt / gemini / perplexity / claude / etc. counts side by side, not just one combined total), a date-range selector (1 / 3 / 7 days), and a “Reset referral counter” button (with a confirmation prompt) to clear the logged history.

0.8.5

  • Marketing/SEO: rewrote the directory title, short description, tags and description opener around the search terms people actually type (AI SEO, GEO, llms.txt, ChatGPT/Gemini/Perplexity visibility) instead of the FocusWeb brand name; added an FAQ entry explaining llms.txt and GEO. No functional changes.

0.8.4

  • Important upgrade fix: sites that upgraded from a pre-0.8.0 build lost their settings, per-post schema overrides, and bot-visit history, because the WordPress.org-required rename of every option/meta-key/table name from the aiv prefix to focuaivi had no migration path. This release adds a one-time automatic migration that recovers settings, custom schema overrides, and analytics history from the old names on upgrade. If you already lost data on 0.8.0–0.8.3, updating to this version will restore it as long as the old data hasn’t been deleted from your database.

0.8.3

  • Plugin Check: prefixed the one remaining non-prefixed global variable ($table in uninstall.php) as $focuaivi_table.

0.8.2

  • Renamed every function, class, define, filter/action, option, transient, meta key, nonce, REST namespace and script handle from the 3-character aiv/AIV prefix to focuaivi/FOCUAIVI (WordPress.org requires 4+ characters to avoid collisions across the ~100,000 hosted plugins). The REST namespace changed from aiv/v1 to focusweb-ai-visibility/v1.
  • Fixed the Plugin URI (https://focusweb.co.il/ai-visibility), which went through a broken multi-hop redirect (https→http→https) that some crawlers refuse to follow; now points directly at the canonical https://www.focusweb.co.il/ai-visibility/.

0.8.1

  • Plugin Check: fixed two MissingTranslatorsComment errors in the Settings status checks — the translators: comment must sit immediately above the line with the __()/sprintf() call, not just earlier in the same statement.

0.8.0

  • Security hardening from WordPress.org plugin review: escaped </script> breakout risk in JSON-LD output, validated/canonicalized custom schema JSON before it’s ever stored (rather than saving the raw POST string), moved the AI-referral/gtag front-end script from an inline <script> tag to a properly enqueued file (wp_enqueue_script + wp_localize_script), and added a nonce check to the public referral-logging REST route so it can’t be trivially forged cross-origin.

0.7.3

  • Dashboard charts: added light gridlines and a max-value reference in the corner for scale, bolded the daily total label (previously blurred to near-illegible at small size), rounded the bars, switched to the system font stack, and added a subtle baseline tick for zero-activity days so the axis isn’t just blank space.

0.7.2

  • llms.txt: page entries now include a body-content synopsis alongside the SEO meta description, not just a single clipped sentence — much more substance for brochure-style sites with no blog. Also fixed HTML entities (e.g. &hellip;) leaking into the output unescaped.

0.7.1

  • llms.txt: pages now get a one-line description too (previously only blog posts did), so brochure-style sites without a blog no longer generate a bare list of titles and links. Prefers the page’s own Yoast/Rank Math/AIOSEO meta description when set, falling back to its excerpt.

0.7.0

  • Schema Browser: added a Validate button per row linking to validator.schema.org with the live page URL pre-filled, and inline Edit/Save (mode + custom JSON) directly on the page, sharing validation with the per-post metabox.
  • Settings page renamed to “Settings & Status”, with a new Status section showing live valid/invalid indicators for Analytics, LLMS access, and Schema enforcement (including a live llms.txt availability + well-formedness check).
  • Diagnostics: self-test failures/warnings now include a machine-readable error code and technical detail (DB error, HTTP status, cache header), plus a rolling log of the last 20 tests instead of just the most recent one.

0.6.3

  • Renamed to “FocusWeb AI Visibility” for a clean directory slug; text domain updated to focusweb-ai-visibility.

0.6.2

  • Plugin Check: removed load_plugin_textdomain (wordpress.org language packs load automatically); documented direct queries to the plugin’s own analytics table.

0.6.1

  • Plugin Check compliance: escaped gtag output, guarded Abilities API call for WP < 6.9, esc_sql on table names, prepared IN clauses, Tested up to 7.0.
  • readme: External Services disclosure and Screenshots sections.

0.6.0

  • Server-side sample event: fires ai_referral directly to the GA4 datastream via the Measurement Protocol (Measurement ID + API Secret setting).
  • Last sample-event result and timestamp snapshot in Settings.

0.5.0

  • GA integration: last 10 fired ai_referral events and 7-day total shown in Settings (server-side mirror beacon).
  • Diagnostics: persistent snapshot of the last self-test result and time.

0.4.1

  • Schema author/publisher is now always the site brand (Organization) — WordPress usernames are never exposed. New focuaivi_schema_brand filter.

0.4.0

  • Admin split into Dashboard / Schema Browser / Settings submenus.
  • llms.txt reads chart by agent per day; explanatory dashboard text.
  • Schema Browser: inspect the JSON-LD of every page, with mode indicators.

0.3.0

  • Daily analytics chart (color per agent), llms.txt read counter, agent questions log.
  • Optional Google Analytics ai_referral events with source/medium/UTM tagging.
  • Per-page schema editor with custom override; site-wide force mode.
  • Analytics self-test; self-healing DB upgrades.

0.1.0

  • Initial release: dynamic llms.txt, baseline JSON-LD, read-only agent endpoint with Abilities API registration, AI bot analytics (7-day window), admin dashboard.