Oct 2, 2026 · Victor Slompo
Everyone talks about GEO. I measured it.
SEO, GEO, AEO. No hype. Server logs.
The images are the original Instagram carousel, in Portuguese.
THE PROBLEM
GEO became a buzzword. Almost nobody shows results.
Most people stop at "add structured data". And neither OpenAI nor Google gives you a report on how often you show up in their answers. The result: everyone promises, nobody proves it.
THE CASE
The annual report of one of Brazil's largest cosmetics companies, rebuilt for AI to read.
- 01Real HTML, server-rendered, in 3 languages
- 02Schema.org/JSON-LD, llms.txt and 17 AI crawlers invited by name
- 03And most important: a log collector to measure everything
PROOF #1 · AI READ IT
1,427
reads of the content by AI crawlers
8 different engines · 15 days of measurement · 8,047 requests logged
PROOF #2 · SOMEONE ASKED
553
fetches triggered by a real question
Not a crawl. It's ChatGPT or Claude going to the site the moment a person asked.
PROOF #3 · IT BECAME A VISIT
60
people reached the site by clicking an AI citation
57 from ChatGPT · 113 downloads of the report PDF
WHAT YOU TAKE FROM THE WORKSHOP
The method behind these numbers. Step by step.
- 01Content that both AI and people understand: questions, FAQs, tables and an expert voice
- 02The technical layer: Schema, llms.txt and robots rules for AI agents
- 03How to measure: separate crawls, real questions and citation clicks
Want this working in your company?
I take AI off the slide deck and put it into production, with measured results.