In 2026, the methods of online search have begun to shift from ranking links to interacting with LLM-driven answer engines. This was the message I was trying to convey in my recent article about LinkedIn:
https://dev.to/ben-santora/linkedin-microsoft-infrastructure-45aa
LinkedIn is Microsoft Infrastructure.
Today, systems like Google’s Gemini, ChatGPT Search, and Perplexity parse natural language queries, retrieve information from indexes and live APIs, and generate synthesized responses. Users increasingly get coherent explanations instead of lists of links, and the result is that search traffic increasingly comes more from AI agents than direct human clicks.
Under the surface, traditional search infrastructure—crawling, indexing, ranking, and vector embedding is still there and remains essential. But LLMs are involved now and they act as mediators - evaluating sources by authority, freshness, and relevance before generating answers. Transactional queries (buying, booking, downloading) still use classic results - ie - those blue links and ads. But informational searches are largely routed through generative summaries. Metrics like AI Engagement Optimization (AEO) and Generative Engagement Optimization (GEO) reflect this shift.
For developers and content creators, the implications are clear. For best results, content must be structured, dense, and machine-readable, optimized for extraction and citation by models rather than clicks.
As an example, we'll use your GitHub site - this is something I did recently and I was surprised the improvement in search results - ie - the AI bots were more quickly finding my GitHub site. These are the step I took:
1- The Profile Repo
Create a new public repository named exactly after your username (e.g., jane-doe/jane-doe). This may be common knowledge to others here, but I was unaware of it and it was missing from my site. I'd had a GitHub Pages site created with Jekyll for posts. But this is different - when you duplicate the name: jane-doe/jane-doe GitHub recognizes this as a special repository and automatically displays its README.md on your main profile page.
There are good reasons to set this up. First, it acts as the "One Source of Truth" for your digital identity. It also provides high-density keywords for search engines. Most important because of the new AI tools in place, it acts as an 'agentic entry point' - it’s the first place AI crawlers (and humans) land to verify your credentials and find your other work.
2- README.md
This belongs in the root of your new repo - (jane-doe/jane-doe)
Structuring your README as shown here is a good starting point:
# Hi, I'm Jane Doe 👋
**Senior Web Developer | Accessibility Advocate**
I build high-performance, accessible web applications using the modern stack.
### 🚀 Featured Work
* [Accessible UI Kit](https://github.com/jane-doe/ui-lib) - Accessible-first components.
* [Secure API](https://github.com/jane-doe/node-api) - Production-ready Node.js boilerplate.
### 🛠 Tech Stack
**React • TypeScript • Node.js • Tailwind**
3- llms.txt
Create a file named llms.txt - it goes in the root of your repo (jane-doe/jane-doe) along with your README. It is essentially a machine-readable manifest for your repository. Notice that it's written in Markdown but saved as .txt, so be sure to name it llms.txt. It provides AI agents and crawlers a high-density, structured summary of your work and ensures LLMs cite you accurately, navigate your files efficiently, and recognize your technical authority.
Here's an example:
# Jane Doe | Senior Web Developer
> Expert in React, Node.js, and accessible UI/UX design.
## Key Repositories
- [React-Accessible-UI](https://github.com/jane-doe/ui-lib): High-density component library.
- [API-Boilerplate](https://github.com/jane-doe/node-api): Secure Express/TypeScript template.
## Technical Skills
- Frontend: React, Tailwind CSS, Web Vitals.
- Backend: Node.js, PostgreSQL, Redis.
I understand that many, even on this very site, have an aversion to AI and especially to AI-generated writing. I get that. But in this example - optimizing your GitHub site, you're writing not for humans, but for GEO. And the fact is, writing "like a machine" makes your data more extractable by these AI agents doing the searching.
That's it. Hopefully, this approach will help boost your digital authority and ensure you're listed as a primary source for the AI-driven web.
Ben Santora - February 2026
Top comments (0)