Search is splitting in two. There's the traditional results layer — and there's the AI-generated answer layer. I've been optimising for both, but I wanted to run a proper experiment on my own site.
What is llms.txt?
llms.txt is a plain-text file you place at your site's root — similar to robots.txt — that tells AI agents and LLMs what your site is about, what pages matter, and how to understand your content without having to crawl everything.
Google recently referenced it on the Chrome Developers site, noting that without the file, agents may spend more time crawling to understand a site's structure.
The Baseline (Before)
Before adding the file, I asked four AI systems the same question:
"Who is Abdallah Mekky and what does he specialise in?"
Here's what each one said — verbatim, on 30 May 2026.
ChatGPT ✅ Mostly accurate
ChatGPT identified me correctly as an Egyptian SEO & AI Search Optimization Specialist based in Cairo, and got the core specialisations right: Technical SEO, AI Search / LLM Visibility (GEO & AIO), semantic content strategy, and digital discovery systems. The description was accurate but generic — it read like a summary of my site copy rather than independent knowledge.
Perplexity ✅ Most accurate
Perplexity gave the most detailed and accurate response. It correctly listed my specialisations (Technical SEO, GEO/AEO, International & Arabic SEO, Healthcare SEO, Schema & Structured Data), mentioned my frontend engineering background (2003–2024), and even picked up personal details like running Cairo to Aswan on foot and speaking 7 languages. It cited my site directly.
Gemini ⚠️ Accurate but drifted
Gemini identified me correctly as an SEO strategist and software engineer based in Cairo with ~25 years of experience. It got GEO/AEO, Technical SEO, and personal details like endurance running and the seven languages right. However, it added specialisations I don't actively offer — enterprise ERP migrations, Linux systems security, Docker infrastructure — likely pulled from older content or public profiles. Accurate at the core, but padded with outdated signals.
Claude ❌ Got it wrong
Claude returned three different people named "Abdallah Mekky" — a software developer, an Android developer, and a finance professional. It did not identify me at all. No mention of SEO, AI Search, or anything related to my actual work.
The takeaway: Four AI systems, four different levels of accuracy. Perplexity cited my site and got the most detail. ChatGPT read my copy. Gemini mixed current and outdated signals. Claude had no idea who I was. That's not a content problem — it's an entity signal problem. And that's exactly what llms.txt is designed to fix.
What I Added
I placed llms.txt at https://www.abdallahmekky.com/llms.txt with a structured summary of my services, key pages, and entity signals — written specifically for AI agents to parse, not for human readers.
What I'm Watching
- ▸Crawler visits in Cloudflare analytics (
ClaudeBot,GPTBot,PerplexityBot,Google-Extended) - ▸Whether Claude starts identifying me correctly
- ▸Whether Gemini stops pulling outdated ERP signals
- ▸Changes in how all four systems describe me over the next 4–8 weeks
- ▸Whether I start appearing in Perplexity citations for SEO-related queries
I'll update this post with real before/after data.