Foreword and guide to the updated AI100 corpus

This is not a blog or a news feed. This is a research library — a structured corpus of texts about how brands exist in the responses of AI systems. Or don't exist.

Each text occupies a specific place: some introduce concepts, others examine specific phenomena, and others provide practical tools. All materials are cross-linked and organized into reading paths — ready-made sequences for a specific role or task.

The thematic field of the corpus: how a model represents a brand internally, where it gets its information, how it forms recommendations, why the same brand looks different across systems and languages, and what to do about it.

All materials in the corpus follow a uniform structure. This is intentional — consistency makes it easy to navigate and compare texts.

Attribute line

Above the title — a compact line: material type, ●● difficulty level, reading time, and the tasks the article helps solve. More on types and levels below.

Description and lead quote

Below the title — one or two sentences explaining the essence. Below that:

An italicized lead quote that sets the tone and formulates the central thesis of the article.

Research card

Most articles include a table with three fields:

Body text and table of contents

The text is divided into sections with subheadings. On the right — a table of contents for quick navigation between sections.

Three closing blocks

Research articles end with three blocks that capture the current state of knowledge on the topic:

Sources and related materials

At the end — a numbered list of sources. Numbers in square brackets [1], [2] appear throughout the text and link to the corresponding entry. After the sources — a block with links to other texts in the corpus on the same topic.

Each text is marked with a colored badge in the attribute line.

Foundational text — the foundation of the corpus. Introduces key concepts, explains mechanisms, builds the groundwork. If you're just starting — you'll most likely begin with foundational texts.

Research article — an analytical article backed by data. Examines a specific phenomenon, always contains a research card and three closing blocks.

Field note — an observation recorded in real time. Less formalized than a research article; captures a discovered effect or anomaly with primary data.

Update — an analysis of a specific platform change. Tied to a date, includes an assessment of how the change affects brand visibility.

Reference — material for constant reference. A glossary of all terms and metrics, convenient to return to while reading any other article.

Observation template — a practical tool. A card that can be used to record observations for each study.

Guide — a navigational text (including this one). Helps orient within the corpus and choose an entry point.

Each material is marked with colored dots — in the material list and in the attribute line on the article page.

● Introductory — an entry point. No prior knowledge required. Suitable for a first encounter with the topic.

●● Intermediate — assumes you've already read at least a couple of introductory texts and understand the corpus's core concepts.

●●● Advanced — for readers who have completed at least one reading path. Touches on methodological nuances, edge cases, and non-obvious relationships.

A path is a ready-made sequence of texts selected for a role or task. You don't have to complete a path in full: even the first two or three texts provide a working understanding of the topic.

Getting started — the minimum set. What AI visibility is, where the model gets information about a brand, and what the customer journey through an AI intermediary looks like.

For the business owner — focus on business decisions: the economics of invisibility, the competitive landscape, when and why to run a study.

For the marketer — sources, citations, what transfers from SEO and what doesn't, practical steps to grow visibility.

For the technical lead — infrastructure: machine-readable data, markup, access control, integrations.

For the researcher — methodology: how the benchmark works, which metrics and why, limitations, reproducibility.

For the agency and consultant — how to explain the topic to a client, which arguments work, how to structure diagnostics.

Full course — all materials in the corpus in recommended order, from introduction to advanced topics.

The knowledge base page is organized into four modes. Switch between them using the cards at the top of the page.

All materials

The full catalog in table form. Each row shows the article title with a brief description, type badge, colored difficulty dots, and reading time. The table can be sorted by any column (click the header) and filtered through the search bar — search works across titles and descriptions.

Find by task

Filtering mode. At the top — task cards: "Understand the problem", "Start diagnostics", "Assess risks", "Prepare implementation", and others. Select a task — get only the materials that address it. Additional filters let you narrow the selection by topic and difficulty level.

Reading path

Seven ready-made paths — sequences of texts selected for a role or task. Each path expands on click and shows numbered steps with types, levels, and reading times. Total time and expected outcome are indicated. The "Start reading" button opens the first text in the path.

Reference

Three reference blocks convenient to return to while reading:

• Concepts — glossary of all corpus terms. Each term includes a definition and an italicized practical meaning — what this term means for decision-making.
• Report metrics — weight tables for the main score and diagnostic metrics. Show what makes up the AI Visibility Score and how to interpret each indicator in the report.
• Research scenarios — types of questions AI100 asks the model during testing. Explain what each scenario tests and how it affects the final score.

The AI100 corpus is available in five languages: Russian, English, Spanish, French, and German.