Free, open-source tools for Generative Engine Optimization (GEO) — making your site legible to ChatGPT, Gemini, Claude & Perplexity. Zero dependencies, no signup. Use them in your terminal, or right in the browser.
▶ tools.trygeosuite.it — the hub
| Tool | What it does | Try it | Code |
|---|---|---|---|
| 🤖 AI Crawl Check | See which AI crawlers (GPTBot, ClaudeBot, PerplexityBot…) your robots.txt allows or blocks, with an AI-visibility score. Plus a public bot-list API. |
open | ai-crawler-bots |
| 📄 llms.txt Generator | Turn any site's sitemap.xml into an llms.txt — the index that tells AI models which pages matter. |
open | llms-txt-generator |
| 🔖 Schema Templates | Copy-paste schema.org JSON-LD templates (Organization, Product, FAQPage…) and validate your own. | open | schema-templates |
| 🗺️ Sitemap Builder | Crawl a site and build a sitemap.xml — quick in the browser, or a full crawl from the CLI. |
open | sitemap-builder |
| 🧑 Poppy Check | Check and generate the /.well-known/poppy.json of the Personal Agent Protocol, the file people's own AI agents read to find a business and sign in for them. |
open | poppy-json |
| ✅ AI Readiness Check | All four checks against one site: AI crawlers, llms.txt, schema.org JSON-LD and sitemap, with one score out of 100 and the free tool that fixes each gap. Plus, outside the score, the Personal Agent Protocol file (/.well-known/poppy.json). |
open | this repo |
Every tool is also a zero-dependency CLI on npm:
npx @geosuite/ai-crawler-bots robots https://example.com
npx @geosuite/llms-txt-generator https://example.com/sitemap.xml
npx @geosuite/schema-templates show Organization
npx @geosuite/sitemap-builder https://example.com
npx @geosuite/poppy-json check https://example.comThe hub page itself — a small zero-dependency Cloudflare Worker that deploys to tools.trygeosuite.it. It does a bit more than serve static HTML:
- Bilingual (en/it):
/is content-negotiated onAccept-Language;/enand/itare explicit, with an EN·IT switch,hreflang/canonicalandx-default → /. - Live social proof: one header figure shows the five tools' combined npm downloads (all-time + last-month), fetched server-side with a last-good Cache fallback so a rate-limit blip never blanks it.
- Dogfoods GEO: serves
/robots.txt(welcomes AI crawlers → scores 100 on our own AI Crawl Check),/llms.txt,/sitemap.xml, and JSON-LD (Organization + an ItemList of the fiveSoftwareApplications) in the page head. - AI Readiness Check (
/check,/en/check,/it/check): runs the toolkit against one site throughGET /api/scan?url=(check.js). The/.well-known/poppy.jsonis checked like an agent would (draft 0.3 of the Personal Agent Protocol,poppy.js), shown but outside the score. The robots.txt verdict is asked to theai-crawl-checkWorker over a service binding, so both pages show the same score;llms.txt, JSON-LD and sitemap are plain fetches, no AI, no cost per scan. Results are cached per site for 10 minutes and counted in theai_readiness_checkAnalytics Engine dataset (target domain and scores, nothing about the visitor). A shared/check?url=…link re-runs the scan. The score is labelled as the site's technical readiness, not its visibility in AI, and the result ends on a GeoSuite demo, with signing up (an account comes with one audit included) as the smaller second option. - What does the AI answer? (
GET /api/preview?url=,preview.js): under the check, one real question to a model. A first call reads the home page and writes the question a customer would ask (in the site's language, without the brand); a second call answers it blind, twice in parallel with up to 3 web searches each, like ChatGPT does (OpenAI Responses API, ~15s and ~7 cents), names that come back both times ranked first; code decides whether the site is among the five names. Limits: 7-day cache per domain (free, not counted), 3 previews a day per visitor and 200 a day overall, counted exactly in thePreviewBudgetDurable Object (it stores only a day-salted hash of the IP). Provider and model are theLLM_*vars inwrangler.toml, any OpenAI-compatible endpoint; the key is a secret (npx wrangler secret put LLM_API_KEY, or.dev.varslocally). Without the key the preview hides itself and the check keeps working. - Shareable: Open Graph + Twitter card meta with a bundled 1200×630
/og.png(wranglerDatarule), plus GeoSuite's favicon (/favicon.png,/apple-touch-icon.png).
npx wrangler deploy # to the GeoSuite Cloudflare accountPushes to main auto-deploy via .github/workflows/deploy.yml
(needs CLOUDFLARE_API_TOKEN + CLOUDFLARE_ACCOUNT_ID repo secrets). If a tool's
worker name/subdomain changes, update the links in page.js.
MIT — by Matteo Perino, maintained under GeoSuite.
This is part of the open-source toolkit behind GeoSuite — the platform that measures and improves how AI engines (ChatGPT, Gemini, Claude, Perplexity) cite your brand. Explore the platform →