Zen GEO – Generative Engine Optimization
·
Score every page for AI answer engines, see which assistant read which page, then publish llms.txt, Markdown and structured data.
Search used to send people to your page. Increasingly, an assistant reads your page and answers on your behalf — and only sometimes cites you. Generative Engine Optimization is the work of making sure that when an assistant answers a question in your field, it can find your content, understand it, quote it accurately, and name you as the source.
Zen GEO does that work in WordPress. It is not an SEO plugin and does not replace one — it sits alongside Yoast, Rank Math or nothing at all, and stands down automatically where it would conflict.
What this does that an llms.txt generator does not
Publishing an llms.txt file is a few lines of code, and plenty of plugins stop there. Zen GEO treats that as one output of the job rather than the job:
- It scores your content and tells you what to change. Twelve weighted checks model how retrieval-augmented systems actually behave — they chunk a page, embed the chunks, retrieve a few, and quote whichever one answers the question on its own. Every failing check names a specific fix.
- It reads page builders. Content is extracted from Elementor’s stored element tree, so pages
built with Elementor, its Pro widgets, or third-party widgets are scored on their real text
instead of an empty
post_content— which is where most content analysers give up. Zen GEO also adds its own tab to the Elementor editor panel. - It shows which assistant read which page. Not just that AI crawlers visited: a per-page, per-crawler breakdown, stored locally with no IP addresses and 90 days of history.
- It finds the FAQ you already built. Question-and-answer pairs are read out of Elementor accordions, toggles and Details blocks and published as FAQPage data, with nothing retyped.
- It defers instead of competing. Yoast, Rank Math, AIOSEO, SEOPress and Slim SEO are detected and Zen GEO stands down where they already emit structured data.
Write content engines can actually quote
- Key takeaways — a short bulleted summary at the top of a page. Each bullet is a complete claim, positioned where a retrieval system chunks first.
- Questions & answers — published as visible text and as FAQPage data. Nothing hidden behind JavaScript, because text an engine cannot see is text it will not quote.
- GEO analyzer — scores every page 0–100 across twelve weighted checks that model how retrieval actually works: a direct answer up front, question-shaped headings, lists and tables, concrete figures, outbound citations, structured Q&A, depth, freshness, a named author, a hand-written summary, and a specific title. Every failing check comes with a specific instruction, not a colour.
- Site-wide audit with CSV export, and WP-CLI commands for bulk work.
Prove it is working
- AI crawler analytics, stored locally: which assistants visited, how often, and exactly which pages each one read. No IP addresses, no visitor data, nothing sent anywhere.
- A 30-day activity chart, per-crawler totals, the most-requested paths, and a day-by-day history kept for 90 days with CSV export.
- An optional weekly email summary to the site administrator.
Publish a machine-readable map of your site
- llms.txt — a Markdown index of your site at
/llms.txt, following the llmstxt.org convention: your site name, a summary, and annotated links to the pages that matter. - llms-full.txt — optionally, the same pages with their full text inlined, for assistants that fetch one document instead of crawling.
- Markdown pages — every page also available as clean Markdown at
?zengeo=md. No theme, no navigation, no cookie banner: just headings, prose, lists, tables and links. Advertised in each page’s<head>so crawlers discover it on their own.
Decide who may read your site
- Per-crawler allow or disallow for GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, CCBot, Bytespider, meta-externalagent, Amazonbot, AI2Bot, Diffbot, YouBot, DuckAssistBot, MistralAI-User, cohere-ai and the user-triggered fetchers.
- Crawlers are grouped by what they actually do — model training, answer-engine indexing, or a fetch triggered by a real person — because blocking the second kind removes you from AI answers entirely, and most people who say “block AI” do not mean that.
- Enforced blocking: robots.txt is a request. Optionally answer 403 instead, at the PHP layer.
- Custom robots.txt directives, validated on save.
Structured data that supports attribution
- One consolidated JSON-LD graph — Organization or Person, WebSite, Article/WebPage, BreadcrumbList — with every node cross-referenced so an engine reads one coherent statement rather than a pile of fragments.
- sameAs profile links, which is how an engine confirms the name on your site is the same entity it has seen elsewhere.
- FAQPage from your Q&A pairs, speakable from your key takeaways, LocalBusiness with address and opening hours, and Product for WooCommerce.
- Detects Yoast SEO, Rank Math, All in One SEO, SEOPress and Slim SEO, and stays out of their way by default.
Privacy
Zen GEO makes no external network requests of any kind. It does not phone home, has no license server, collects no telemetry, and loads no remote assets. Crawler analytics are stored in your own database and never leave it.