Testing how content becomes visible inside AI answers.
Nexxus8 is a GEO research project. Each experiment tests a specific structural or semantic change to a web page, then tracks whether that page starts appearing inside AI-generated answers. Findings are published as field notes.
Empirical GEO experiments, published as they run.
Each test isolates one variable — content structure, semantic markup, source citations, definition format — and measures the effect on LLM retrieval. Visibility is tracked using LLMin8 across real prompts on ChatGPT, Perplexity, Claude, and Gemini.
questions
Open questions in GEO — tested with real data.
Each field note addresses one of these directly, using real pages and measured outcomes.
Which content structures increase citation frequency in ChatGPT and Perplexity?
Does semantic structure affect LLM retrieval more than inbound links?
How do definitions, comparisons, and cited sources affect what LLMs pull?
What changes — measurably — in retrieval rate after a content update?
experiments work
How a typical experiment runs.
Real content, real URLs. No synthetic environments.
Heading structure, schema markup, definition format, or source density — isolated per test.
Prompt-level visibility tracked across ChatGPT, Perplexity, Claude, and Gemini.
Findings published including null results — when a change produces no measurable retrieval effect.
“If GEO is going to become a real discipline, it needs to be tested — not just theorised.”— That’s what Nexxus8 is for.