Blog
Mert Özen's blog posts on software development, backend, frontend, and DevOps.
What Is GEO? How AI Answer Engines Find and Cite Sources
AI answer engines split a question into sub-searches, find pages in an index or in their own crawlers' data, put them into the LLM's prompt and tie the answer to sources; this live retrieval channel is completely separate from the model's training data. We cover how Google, Copilot, ChatGPT search, Perplexity and Claude find sources, and why GEO is still SEO.
Google's AI Search Guide: What to Do and What to Skip (llms.txt, Chunking and Special Markup Myths)
According to Google's May 2026 guide, AI Overviews and AI Mode rest on the core ranking systems; an indexed page eligible for a snippet is enough, and llms.txt, chunking, rewriting for AI and special schema aren't needed. We cover the guide's "you don't need to" list, its emphasis on non-commodity content, spam policies applied to AI answers, and why the guide applies to Google Search only.
Understanding and Managing AI Bots: Training vs Search vs User-Initiated Fetches
AI bots come in three classes: training crawlers such as GPTBot and ClaudeBot, search bots such as OAI-SearchBot and Claude-SearchBot, and user-initiated fetches such as ChatGPT-User; to opt out of training but stay citable, you block only the first class. We verify robots.txt scenarios against RFC 9309 rules and cover Google's snippet and Search Console controls, mertozen.com's robots.txt and separating AI bots in logs.
Citable Content: What the Research Actually Shows
The "up to 40%" gain in the GEO paper was measured in a single-actor benchmark where the source was already in the answer's context; C-SEO Bench found most methods ineffective under competition, and a 2026 survey found no tactic with a proven durable causal effect. We give the practices consistent with the evidence and review a post on this site.
llms.txt and Markdown Versions: The Spec, Who Uses It, and Implementing It in ASP.NET Core
llms.txt is a Markdown map at the site root made of an H1, a summary and link lists; Google Search ignores it, no major AI search engine has officially said it uses it, and its real use is coding assistants and agents. We cover spec v2, a Minimal API that generates llms.txt and per-post Markdown from EF Core data, Output Cache invalidation and mertozen.com's file.
Building Sites for AI Agents: Semantic HTML, the Accessibility Tree and Lighthouse Agentic Browsing
Browser agents read a page through screenshots, the DOM and the accessibility tree; a button built from a div doesn't show up there at all, and a button without a name doesn't say what it does. We measure two versions of an Angular 22 form with an accessibility tree dump and Lighthouse 13.6.0's Agentic Browsing category, showing what the tool catches, what it misses and WebMCP's experimental status.
Measuring GEO: Search Console, Bing AI Performance, Referral Traffic and Server Logs
There is no single measure of AI visibility: the Search Console generative AI report gives impressions only, AI Overviews clicks are merged with Web, Bing AI Performance shows citations and a grounding query sample, and on your side you have referrals and user-initiated fetch logs. We show code that records AI bot and referral events with ASP.NET Core middleware and produces a weekly summary with EF Core, and the limits of prompt panels.