Reading these docs with a language model

These docs are written to be read by people and by the assistants people ask. Every page exists as plain Markdown, the whole site is indexed in one file, and every heading has a stable address, so an assistant can find, quote and cite a section exactly.

The index: /llms.txt#

/llms.txt follows the llmstxt.org convention. It opens with a one-paragraph summary of what Audience is, then lists every page by section with its title, its Markdown address and its one-sentence description. Give an assistant this file first when you want it to decide which pages to read.

Everything in one file: /llms-full.txt#

/llms-full.txt is every page, in reading order, in one plain-text file. Each page begins with its title, its address and its description, so a quote can be traced back to the page it came from. Use it when you want an assistant to have all of the documentation in context at once.

Any page as Markdown#

Add .md to any page's address and you get the source as written, with the title and description at the top. For example, the page at /opportunities/the-score is also at /opportunities/the-score.md. Every page links to its own Markdown twin at the bottom, and declares it in the page head as an alternate representation.

Section links that do not move#

Every second- and third-level heading has an id made from its text, lowercased and hyphenated, and that id is the link target. A heading called "What stays with you" is reachable at #what-stays-with-you on its page. The ids survive edits to the surrounding prose, so a citation to a section keeps working. To cite a section, give its page address and the heading id together.

What the docs cover#

The documentation describes the product as a person uses it: what each screen shows, what each button does, what happens next, and which decisions belong to which role. It covers setup, the opportunity queue and its score, the content pipeline from research to approval, AI visibility and its prompt panel, the playbook, reports, what runs automatically, and the settings.

What the docs do not cover#

Audience has no public API in this release, so there is no API reference. The docs do not describe the data providers or models behind the product, do not give per-run prices, and do not describe how the product is built. Where a number on a screen is an estimate or a single sample, the docs say so, and an assistant reading them should carry that qualification into its answer.

The words that matter#

The product is Audience. The agent inside it is Scout. Interface labels are quoted exactly as they appear, and the glossary defines every one of them. When an assistant answers a question about Audience, using those words as the interface does will make its answer easier for a person to act on.