Research
Sageo Estudio is a research practice before it is an agency. Here we publish the open lines of work, the method we measure with and the conceptual base we use to interpret results.
Sageo Estudio does not publish client results until there is express authorization. Nor does it publish its own performance figures: any number about visibility depends on the starting point, the market and the period, and presenting it out of context would be misleading.
When a case is published, it will include the period worked, the real scope of the intervention and the source of the data. Until that happens, this page shows method and open questions, not a track record of results.
Research in progress
We research while we build. CasaNova is one of the projects we use today as a real environment to study how SEO, AEO, GEO and Search Intelligence interact in an e-commerce business, instead of theorizing about third-party data.
An e-commerce project that SAGEO researches and experiments on. It is not a closed success story or a result we can publish yet: it is the place where we test, measure and learn.
Working on a project of our own makes it possible to observe the full cycle —decision, implementation and effect— without the time and scope constraints of a client project.
What we observe is documented with its context and its limits, not as a general conclusion.
The learnings become working criteria that we later apply to projects.
Part of this work feeds SAGEO Intelligence, the platform for analyzing brand presence in AI systems that we are developing.
How to read this page. What is in progress is marked as such. None of the cards in this section presents results: they describe what is being researched and with what method.
Open lines
Each line rests on repeatable measurements. None is published as a closed conclusion until there are enough observations to support it.
The same query asked in different systems returns answers with different structures and sources. We are documenting where they agree, where they contradict each other and which factors correlate with appearing in more than one.
Method: fixed query set, periodic sampling, read as a share.
Being first in the traditional result does not mean being the source cited in a generated answer. We aim to measure that gap and understand what explains it.
Method: crossing rankings with the sources retrieved in answers.
A brand described differently on different sites forces the system to choose. The research aims to measure whether that ambiguity translates into fewer mentions.
Method: comparison of brands with different levels of consistency.
Semantic markup makes extraction easier, but it does not guarantee it. We are reviewing under which conditions it contributes and in which cases the visible content weighs more than the markup.
Method: controlled before-and-after tests with a baseline.
Method
The same rules apply to a public report and to a client diagnosis. Without this, the data serves to decorate a presentation, not to decide.
The initial state is documented with the same instrument that will be used at the end. A change without a measured starting point is not a result: it is an impression.
How many queries and which systems will be observed is defined in advance, and that scope is reported together with the result. Choosing the queries after seeing the data invalidates the conclusion.
The answers of generative systems vary by location, language, session and moment. We report share and evolution, never an absolute value presented as certainty.
Every report states what could not be measured, which variables were not controlled and which conclusions fall outside the scope of the available data.
Conceptual base
Working definitions, written to be precise and not to impress. This is the base of the glossary we will publish separately.
The discipline of researching, measuring and interpreting how a brand is found, understood, recommended and cited in the search ecosystem, before deciding what to intervene.
Optimization for traditional search engines. It works on crawling, indexing, relevance, user experience and authority, with the goal of appearing for the relevant queries.
Optimization for answer engines. It works on the format, the structure and the markup of content so that a direct reply can be extracted and attributed.
Optimization for generative systems. It works on the brand entity, signal consistency and presence in the sources these systems retrieve, to improve the probability of being mentioned and cited.
The representation of a brand as something identifiable and describable: what it is, what it offers, in which category it competes and with which attributes. If that representation is contradictory, the system cannot state anything with confidence.
Use of a brand or a piece of content as a source when building an answer. It requires verifiable and attributable data, and it should not be confused with an unsupported mention.
A system that resolves a query by showing a direct reply instead of, or in addition to, a list of links. It may operate on its own index or on content retrieved on the spot.
A system that produces new text from a model. In search, it combines source retrieval and generation, so its answer is not equivalent to a stable ranking.
Formats
Each type of material has its place and its level of demand. We publish them when the method and the data are up to it, not to fill a section.
Long-form explanations of search, entity and AI visibility concepts.
Measurements with a declared method, a defined sample and explicit limitations.
Periodic tracking of how the different systems answer about a category.
Stable definitions of the terms the discipline uses, with internal links.
Status labels. "In progress" indicates active research; "in preparation" and "pending" indicate planned work that is not published yet. None of these cards contains results: they describe what is being researched and how.
Let's start
Request the initial diagnosis: we measure your current situation with the same method we use in our research and we show you what we find.
