Zen GEO – Generative Engine Optimization

Changelog

1.5.0

Crawler analytics release: it now records which AI assistant read which page, keeps three
months of history, and no longer mangles non-Latin URLs.

Per-crawler page attribution

• Page requests record the crawler that made them. The “Most requested pages” table gains a
  Read by column, so you can see that ClaudeBot read one page and Perplexity another — the
  half that tells you what each assistant has actually ingested, rather than just that it visited.
• Crawlers are ranked within each page, pages ranked by total, and 15 rows are shown.
• New Export pages (CSV) — one row per page per crawler, covering everything stored rather
  than only the rows on screen.

Longer, deeper history

• History is kept for 90 days instead of 30. A day costs one small integer per crawler, so
  three months is a few kilobytes — enough to see whether AI crawler interest is growing or fading.
• The activity chart spans 30 days rather than 14.
• New collapsible day-by-day history: every recorded day with its per-crawler breakdown.
• New Export daily history (CSV) — one row per day per crawler, so a copy outlives the
  retention window.
• The card states how far the history goes back and how long it is kept.

Fixes

• Non-Latin URLs were destroyed. sanitize_text_field() strips percent-encoded sequences, so
  a path of percent-encoded Georgian, Greek, Cyrillic, Arabic, Hebrew or CJK slugs was reduced to
  its punctuation — hundreds of distinct pages collapsing into meaningless rows like /glossary//
  and /glossary/-/. Paths are now decoded before sanitizing, so every page counts separately and
  reads properly (/glossary/აბჰაია/).
• The page list could never accept a new page once full. A new path was stored with a count of
  one, sorted to the bottom, and dropped by the very request that added it — so the list froze on
  whichever pages happened to fill it first and never changed again. Many more pages are now kept
  than are displayed, and the page seen in the current request is always retained so it can start
  accumulating.
• Empty state explained rather than hidden. The “Most requested pages” heading stays visible
  with a line saying pages appear on the next crawl, instead of the whole section silently
  vanishing and looking broken.
• Page rows recorded before this release are cleared. They were either corrupted by the encoding
  bug or carried no crawler, so they could not answer “which assistant read this page?” and would
  have sat permanently at the top of the report. Per-crawler totals, per-day history and the
  grand total are kept — the real history was never in those rows.

Scores, llms.txt, Markdown pages, structured data and crawler blocking were unaffected throughout;
every issue above was confined to the analytics display.

1.4.0

• Reads content from every Elementor widget, including ones it has never seen. The 1.3.0 accordion fix turned out to be one instance of a much wider problem: content was only read from a fixed list of control names, so anything using its own naming was invisible to the analyzer. Auditing Elementor free and Pro found 49 further text controls that were being skipped — Testimonial, Alert, Flip Box, Price Table, Blockquote, Call to Action, Animated Headline, Slides, Author Box, Image captions, Progress bars and more.
• Extraction is now generic: known names are used for their meaning (heading vs rich text) and everything else is kept when it reads like prose. Third-party and custom widgets are covered automatically, with no allowlist to maintain.
• Technical settings are still excluded — CSS ids and classes, URLs, colours, sizes, layout enums, aria labels and shortcode strings never reach the score.
• Affected sites will see scores rise, because pages built with these widgets were previously being measured as near-empty.

1.3.0

• Reads questions and answers straight out of accordions, toggles and Details blocks. Most FAQs are built in the page builder rather than typed into a plugin — those pairs now count towards the GEO score and publish as FAQPage structured data with nothing to retype. Covers Elementor Accordion, Toggle, Tabs and Nested Accordion, and the block editor’s Details block.
• Only headings that actually read as questions are treated as Q&A, so an accordion of plain section names (“Shipping”, “Returns”) is left alone rather than published as a false FAQ.
• Skips any Elementor Nested Accordion that already has its own FAQ schema switched on, so one page never carries two competing FAQPage graphs.
• Fixed: accordion, toggle and tab content was previously invisible to the analyzer entirely — a page whose whole body was an FAQ accordion scored as having almost no content.

1.2.0

• Zen GEO now appears as a tab inside the Elementor editor, beside Widgets, Components and Globals — the score, the fixes, the key takeaways, the Q&A pairs and the visibility switch, without leaving the builder.
• Built on Elementor’s own injectTab() panel API rather than by string-replacing Elementor’s rendered markup, so it does not break when Elementor changes a class name.
• New REST endpoint (zen-geo/v1/post/<id>) backing the panel, gated per post on edit_post.
• New WordPress Dashboard widget: average GEO score, what is published, AI crawler activity over the last 7 days, and the three weakest pages.
• Fixed: the score circle on the Zen GEO dashboard lost its styling in 1.1.0 and rendered as plain text.
• Fixed: dismissing the “Zen GEO is active” notice with the ✕ did not stick, so it reappeared on every admin page load for a week. It now saves the dismissal, and clears itself automatically once you have opened a Zen GEO screen.

1.1.0

• Automatic background scoring: existing pages and posts are scored on an hourly batch after activation, so the audit and the site average are meaningful without re-saving anything. Scores also refresh themselves as they age.
• Per-page control: every page and post can be included in or excluded from Zen GEO from the editor. An excluded page is left out of llms.txt, serves no Markdown version, gets no page-level structured data, and is dropped from the audit and the site average.
• Redesigned editor panel: a single headline verdict, a score ring, a to-fix / to-improve / good tally, the specific fixes first and the passing checks folded away.
• Post-list GEO column now shows excluded pages distinctly from pages not yet scored.

1.0.0

• Initial release.
• llms.txt and llms-full.txt, served dynamically and cached.
• Markdown representation of every indexed page at ?zengeo=md, advertised via rel=”alternate”.
• Per-crawler robots.txt policy for 19 AI crawlers, grouped by purpose, plus optional enforced 403 blocking and custom robots.txt directives.
• Local, IP-free AI crawler analytics with a 14-day chart, per-crawler totals and top paths.
• Consolidated JSON-LD graph: Organization/Person, WebSite, Article/WebPage, BreadcrumbList, FAQPage, speakable, LocalBusiness and WooCommerce Product, with automatic stand-down when an SEO plugin is active.
• Key takeaways and Q&A editors, published as visible text and structured data.
• GEO analyzer with twelve weighted checks, Elementor-aware content extraction, a sortable post-list column, a site-wide audit and CSV export.
• WP-CLI commands: audit, score, llms, crawlers.
• Optional weekly email report, Site Health integration, and settings import/export.