Grounded in the work,
not the hype cycle.
Reach Works is one person's practice — built on a decade and a half of making content authoritative, structured and found, plus current research into how AI models decide what to cite.
Getting cited by AI isn't a growth hack. It's the problem newsrooms have always had — being authoritative, structured and legible to whatever decides what people see — pointed at a new reader that happens to be a machine.
I'm Leandro Oliva. My work has always come down to one question: how does content get found, and trusted? That has meant living inside the mechanics of discovery — SEO, site architecture, structured markup and the audience analytics that show what actually moves — at national newsrooms including The Wall Street Journal and The Intercept, where I led digital operations and audience engagement.
Now I point that expertise deliberately at machines. My research at the University of Amsterdam — a Research Master's in Cultural Data & AI — is on exactly this: how language models decide what to answer and cite. I co-authored an empirical audit of that behaviour, and I build the source-verification and claim-extraction tooling the structuring work runs on. Reach Works is where all of it becomes a service.
Three things AEO needs. I've done all three for years.
Editorial authority & trust
A decade holding accuracy, sourcing and standards across two national newsrooms. E-E-A-T — experience, expertise, authoritativeness, trust — isn't a checklist to me; it was the daily job.
Structured, legible content
Making dense material machine- and human-legible — from Brookings policy briefs to a claim-extraction pipeline that turns arguments into records a model can quote.
Primary research on models
A UvA research master's and published studies on how models answer, cite and represent — so the advice tracks how the systems actually behave, not how a blog post guesses they do.
Research & standards on how AI uses content
The parts most directly load-bearing for this work — with sources, because that's rather the point.
- [1] Co-researcherAn empirical audit of AI answer behaviour — 84 structured prompts across Gemini and Claude, classifying task-fulfilment into five outcome modes. Published in Appification in the Age of AI (ASI Sprint Report No. 3), App Studies Initiative, 2026.
- [2] ContributorContent Telemetry — an open standard for reporting how AI systems use content; authored a provenance-schema contribution adopted into the v0.1 specification.
- [3] AuthorThe Tautology of Recursivity: AI, "Context Rot," and the Society of the Spectacle — a peer-reviewed conference paper presented in the Artificial Intelligence panel at EUPOP 2026.
- [4] Independent analysisOngoing published work on the AI scraper economy, AI procurement and compute markets, and the EU Code of Practice on AI-generated content transparency — verified against the AI Act text and primary sources.
The machinery, not just the slide deck
The structuring and verification work Reach Works sells runs on tooling I designed and directed.
Source-verifying content OS
Verifies every source a draft cites against Crossref and OpenAlex, checks that each source actually supports its claim, and decomposes drafts into facts, quotations and argument beats. The engine behind claim-level structuring.
Local, read-only content tool
Open-source. Scores a writer's feed against their topics and drafts tailored angles — runs locally with the user's own model, reads only, collects nothing, never posts. The privacy stance the service inherits.
Retrieval & evaluation pipeline
Open-source retrieval-and-evaluation pipeline over EU legislation (EU AI Act, GDPR), with a Ragas-evaluated ablation of chunking strategies, hybrid retrieval and cross-encoder reranking. How I know what "machine-legible" actually measures.
Find the work: leandrooliva.com · github.com/landomo · linkedin.com/in/leandrooliva
Want that experience pointed at your visibility?
Start with an audit. See exactly where you stand in AI answers, and what the plan would be.