I spent a day making my portfolio readable by AI instead of by humans

2026-07-29 · Yusuf Gadelrab

Here's the thing I noticed. When a recruiter wants to know about a candidate now, a real number of them ask an assistant first. ChatGPT, Claude, Perplexity, whatever is in the search box. If the model can't parse you, you don't exist in that answer.

Treating models as a first-class audience

So I treated the models as a first-class audience:

What changed on the site

robots.txt now names 49 crawlers explicitly. This matters more than people think. Google-Extended and Applebot-Extended use separate user agents and separately gate whether your content can show up in AI generated answers. A wildcard Allow does not cover them.

llms.txt and llms-full.txt, a plain text profile written to be quoted, not crawled.

One canonical schema.org Person node with a persistent @id, injected across every page on the site (161 of them today), so I'm one entity instead of 161 fragments.

Visible FAQ sections with matching FAQPage schema. Answer-first, so the answer is liftable verbatim.

IndexNow, which pings Bing, Copilot and DuckDuckGo the moment something changes instead of waiting weeks for a crawl.

427 structured data blocks, all valid.

The actual first impression

None of this is a growth hack. It's just noticing that the index changed and the resume didn't.

https://yusuf-gadelrab.github.io/

If you're job hunting in 2026, go ask an assistant about yourself. Whatever it says back is your actual first impression.