Turksem GEO SEO – AI Bot Tracker, llms.txt & GEO Audit

Turk SEM, an SEO agency that has been working on websites since 1999. We built this plugin for our own client work first, and every check in it started as a question we kept running into on real sites: why is the bot log empty, why does the crawler see a different page than the visitor, why did a fix change nothing. The full plugin page, including the bot identity note written for server administrators who find our user agent in their logs, is here.

Generative Engine Optimization is the work of getting your content read, understood, and used by AI systems such as ChatGPT, Claude, Perplexity, and Google's AI features. It has nothing to do with local or geographic SEO. GEO SEO covers the part of GEO you can measure from your own server: which AI crawlers come, what they can read, and what changed after each fix.

No, and it does not try to. GEO SEO runs next to them: it reads their meta descriptions for its audit and disables its own schema output when they are active. Use GEO SEO for what they do not do: AI bot measurement, AI access control, and proof of change.

Yes. The AI access matrix writes allow and block rules into a managed block of your robots.txt, one agent at a time. It also tells you when a block cannot work: user-triggered fetchers such as ChatGPT-User say in their own vendor documentation that they may ignore robots.txt, and GEO SEO will not pretend otherwise.

Yes, in one click, with a grader that scores the result. It also tells you plainly what llms.txt is unlikely to do for you today; the numbers are in the Honest by design section above.

Each front-end page view runs one early check on template_redirect. For a regular visitor that is a single precompiled pattern match, measured at about 10 microseconds. A database write happens only when a known crawler matches: one row for an AI agent, one daily counter update for any other crawler. On very busy sites you can sample (log 1 in 5 or 1 in 10 requests). The audit runs in the admin, in batches.

It is not. The AI bot log stores the bot name, page URL, timestamp, and a short one-way hash of the user agent string. For other crawlers, only the crawler's name, its category, the day and the visit counts are kept. IP addresses are never written to the database.

Two lists. The first holds 39 AI agents, each documented by its own vendor and linked to that documentation: 37 crawlers plus the two permission tokens Google-Extended and Applebot-Extended, which never fetch anything. The second adds 1,463 crawlers from the open crawler-user-agents list (see Credits), covering SEO tools, search engines, monitors, feed readers and scripts. Entries that duplicate a documented AI agent are left out, so no visit is counted twice. Your own watch list comes on top.

English, in this version. The admin copy ships from a reviewed, built-in table rather than translate.wordpress.org, because every audit sentence is written and checked alongside the check that produces it. More languages are planned; community translations may be adopted in a later version.

If a physical robots.txt or llms.txt file exists on your server, GEO SEO refuses to fight it and shows a warning instead. Delete or rename the physical file to let WordPress (and GEO SEO) serve it dynamically.

If a full-page cache (a CDN or a caching plugin) serves the request, WordPress never runs and the visit cannot be logged. The dashboard states this openly. Counts are a reliable floor, not an exact total.

For bugs, use the support forum here on WordPress.org, so the answer helps the next person too. For everything else, there is a two-minute, six-question feedback form. We read every answer, and it shapes the next version.