# The Second Audience: How Machines Read, Represent, and Act on Companies Across the Web

Canonical URL: https://audiencetwo.com/second-audience/thesis
HTML: https://audiencetwo.com/second-audience/thesis
PDF: https://audiencetwo.com/course/The-Second-Audience-Thesis.pdf
Course: https://audiencetwo.com/second-audience
Provider: AudienceTwo

## CANONICAL THESIS: What the Second Audience is.

The web was built around a simple relationship: companies publish information and people consume it.
A second relationship is emerging.
Between companies and their human customers, a computational layer increasingly encounters information first. Machines crawl and index the web. They retrieve information in response to human questions. They extract facts from webpages and other sources. They decide which evidence deserves attention. They compare products and companies. They synthesize recommendations. And increasingly, agents can act on a person's behalf.
I call these systems the Second Audience.
The Second Audience is not simply "AI traffic," and it is not limited to bots visiting a website. It is the machine-mediation layer through which a company's information can be learned, found, selected, interpreted, represented, recommended, and acted upon.
Its components have different roles and different relationships to human intent. A training crawler building a corpus is not economically equivalent to a search crawler building an index. A retrieval system considering a product because someone asked a question is different again. And an agent checking inventory, creating a cart, or completing a transaction on someone's behalf represents another category entirely.
These events should not be collapsed into one metric. But they belong to the same emerging phenomenon: machines are becoming active intermediaries between companies and people.
For most of the web's history, machines primarily transported and rendered information for humans to interpret. A browser downloaded a webpage; the person read it.
The important change is that machines increasingly participate in the interpretation itself. They select. Extract. Compare. Synthesize. Recommend. And sometimes act.
The human remains the ultimate source of demand. But the machine increasingly stands between the human's intention and a company's opportunity to satisfy it.
That is the Second Audience.
Where measurement is narrower than the phenomenon, the thesis wins. Observable slices illustrate the claims. They do not redefine them.

## Five claims the work hangs from
1. **A machine-mediation layer is emerging**: Between companies and people, machines increasingly encounter, select, interpret, represent, recommend, and act upon company information. That layer is the Second Audience.
2. **That layer is heterogeneous**: Roles range from training and indexing to recommendation and action, especially by proximity to human intent. They should not be collapsed into one metric.
3. **Machines encounter representations**: Often not the experiences designed for humans. A person may see a store. A machine may receive HTML, extracted text, Markdown, JSON-LD, a cached passage, an API response, or a tool.
4. **Machine influence is a pipeline, not a visit**: Crawl is not retrieval. Retrieval is not context. Context is not influence. Influence is not recommendation. Recommendation is not transaction.
5. **Expanding into recommendation and action**: The machine-mediation layer is expanding from reading and retrieval into recommendation, interaction, and action. These are increasing capabilities across the ecosystem, not stages every machine must traverse.

The observable HTTP request is real and often the strongest measurement surface available today. It is one observable event inside the phenomenon, not the entire phenomenon. Where measurement is narrower than the phenomenon, the thesis wins.

The thesis defines the phenomenon. Research interrogates it. The course teaches a measurement discipline. Product experiments illustrate the thesis; they do not define it.