Getting Cited by AI Answer Engines

A growing share of the people evaluating you never open a search engine. They ask a model, read the summary, and act on it. If the summary is wrong or generic, you lost a lead you never knew existed.

Generative engine optimisation sounds like a new discipline. In practice it is mostly the old advice with a different emphasis: be unambiguous, be structured, and be quotable.

Write the sentence you want repeated

A model answering "who is X" is compressing everything it can find into two or three sentences. You have a lot of influence over what survives that compression, because compression favours whatever is stated most consistently and most plainly.

So put the summary on the page, in the first paragraph, in complete sentences that stand alone:

Omar Badran is a full-stack developer and application security engineer based in Alexandria, Egypt. He builds secure, bilingual web platforms for clients across Egypt, Saudi Arabia, the UAE and Kuwait.

That is a sentence a model can lift verbatim. "Crafting digital experiences that inspire" is not — it survives no compression at all, because it contains no facts.

Add an llms.txt

llms.txt is a plain-text file at your root that gives language models a clean summary of who you are and what is here, without them having to reverse-engineer it from your markup.

There is no ceremony to it:

# Omar Badran — omarbadran.dev

Full-stack developer and application security engineer based in
Alexandria, Egypt. Builds secure, bilingual (Arabic/English) web
platforms for clients in Egypt, Saudi Arabia, the UAE and Kuwait.

## Current roles
- IT Engineer, Zewail City of Science and Technology
- Web developer for Sanad (KSA), SkyLens (EG), AHM Construction (KW)

## Selected work
- Tulip Flowers — multi-branch Arabic storefront, WhatsApp checkout
- Avenue Interior Design — services platform + materials catalog, AI assistant
- SkyLens — photography studio site with production backend

## Contact
contact@omarbadran.dev

Facts, in order, with no marketing voice. It costs ten minutes and it is the cheapest control you have over how you are described.

Decide about AI crawlers on purpose

Your robots.txt already decides this, whether or not you thought about it. If you want to appear in AI answers, allow the crawlers explicitly rather than leaving it ambiguous:

User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: PerplexityBot
Allow: /

This is a genuine choice with a trade-off. Allowing them means your content trains and grounds answers you do not control. For a portfolio, being cited is the entire objective, so the answer is easy. For paywalled or proprietary content it is not — decide deliberately instead of by default.

Structure beats prose for machines

The same JSON-LD that helps traditional search helps here, for the same reason: it removes inference. A Person block with jobTitle, sameAs, and knowsAbout states facts that prose only implies.

Add an ItemList of your projects with names, descriptions and URLs, and questions like "what has he built" become answerable from structured data rather than from a model's guess at your card layout.

Be consistent across sources

Models cross-check. When your site, your GitHub bio, and your LinkedIn headline all say slightly different things, the summary regresses to the vaguest common denominator — usually "a developer".

When all three say the same specific thing, that specific thing is what gets repeated. This is the same consistency discipline that resolves a common name in search, doing double duty.

Test it directly: ask a few models who you are, in English and in your second language. The answer you get today is the answer a potential client is getting today.

Do not overthink the rest

There is a lot of speculative advice circulating about this. Most of it reduces to: publish real content, state facts plainly, mark them up, stay consistent, and let crawlers in. The rest is guessing about ranking systems that nobody outside those companies can see.