Nati Elimelech's Blog
Here I write about what I've learned in 20+ years at scale: how to build TECH SEO and GEO systems that work on millions of pages, how to solve problems no one else has solved, and how to translate SEO into language that product and engineering teams understand.
8 articles
Googlebot IP Addresses: 2,000+ IPs Updated Daily (All 5 IP Files)
Every Googlebot IP address and CIDR range, pulled directly from Google's official JSON endpoints. Updated daily. Covers all five IP files across Google's three crawler categories: Common Crawlers, Special-Case Crawlers, and User-Triggered Fetchers. Copy-paste ready for firewall rules.
web-bot-auth: The Bot Authentication Protocol Explained
The definitive guide to web-bot-auth, the IETF protocol for cryptographic bot identity. How it works, who built it (Cloudflare + Google), who's using it (OpenAI, AWS WAF, Akamai), and what it means for your site.
What is server response time, and how it affects Google's crawl rate
Server response time sets how much Googlebot crawls. On one large site, dropping from 500ms to 120ms took crawling from 15,000 to 100,000 URLs a day.
Platform properties in Search Console: Google starts measuring presence beyond your website
Platform properties in Search Console are now open to everyone: connect social and video accounts, and give SEO and social teams their first shared search data layer.
What is llms.txt, and why you probably don't need it
You don't need an llms.txt file for AI visibility. Google's docs say Search doesn't use it, and the logs of real sites show almost no crawler fetches it.
Google's Bots: Which Ones Feed Search, and Which Ones Feed AI
A guide to every Google crawler: Googlebot, Google-Extended, GoogleOther, Google-Agent and Google-GeminiNotebook. What each one does, which feed Search and which feed AI, how to spot them in your logs, and what each token actually controls.
query fan-out: what it is and how it shapes your AI visibility
query fan-out is how ChatGPT breaks one question into many parallel searches and assembles a single answer. Here's what it means for your visibility, and how to track it.
I tested if AI agents can read Lovable, Base44, Bolt, and v0 websites so you don't have to
I built the same page on Lovable, Base44, Bolt, and v0, then tested whether ChatGPT and Claude could retrieve it. All exposed the topic. Only some exposed enough HTML for reliable body retrieval.