Glossary

The AI brand glossary.

The vocabulary of AI-era brand work is settling fast, and most of it gets defined by engineers. These are the terms in plain language, for the person who owns the brand.

AI import
Extracting a structured brand model from what you already have — a brand PDF or a website — using AI. A head start you review and correct, not an authority. In SPOKEPROS it's how a brand goes from document to system in one sitting.
Approved phrases
The words and formulations your company deliberately uses — product names, category terms, the way you describe what you do. Stored as data, they're applied by people and AI tools alike instead of living in one editor's memory.
Banned phrases
The inverse: words that never leave the building — competitor framings, legally risky claims, retired taglines. When they're structured rules rather than folklore, connected AI tools respect them by default.
Boilerplate
The official company description, maintained in fixed lengths and languages, used verbatim in press releases, listings and bios. Six competing versions of it is the classic symptom of an unstructured verbal identity.
Brand check
Scoring a draft — or a live page — against the brand model, with specific fixes. Turns review from something only the brand owner can do into something anyone can run.
Brand drift
The slow divergence between what the brand is supposed to sound and look like and what actually ships. Accelerates sharply when AI multiplies content volume and every writer prompts differently.
brand.json
A machine-readable manifest of a brand, served at a stable URL so software can read the brand's facts directly. Every published SPOKEPROS brand serves one alongside its llms.txt.
Brand model
The brand as structured, versioned data — visual and verbal identity as usable parts rather than pages. The foundation of a living brand system; everything downstream (pages, servers, checks, generation) reads it.
Brand server
A live endpoint that serves your brand model to AI tools. Every published SPOKEPROS brand is one — Claude, ChatGPT and Cursor connect to it and read voice rules, boilerplates and banned phrases before writing. The deep-dive →
Living brand system
A brand maintained as a current, structured, machine-readable system rather than a static document — usable by people at one URL and readable by their tools. The full guide →
llms.txt
An emerging convention: a plain-text file at a stable URL that gives language models an authoritative summary of a site — or, for a published brand, of the brand itself. Anything that reads URLs can stay on-brand without a special integration.
MCP (Model Context Protocol)
An open protocol that lets AI tools call external sources of truth ("servers") for context while they work. It's how a brand server speaks to Claude, ChatGPT and Cursor: the tool asks, the brand answers.
Messaging house
The structure holding your positioning: the roof (the core claim), the pillars (the supporting messages), and the proof points under each. When it's data, every argument — human or AI — starts from the same foundation.
On-brand at the source
The upstream alternative to review: the brand is present where drafts are generated, so content starts on-brand instead of being corrected into shape afterwards. The core idea behind AI brand governance done humanely.
Tone by context
Rules for how the same voice flexes across situations — a crisis statement and a product launch shouldn't sound identical, and the difference should be a written rule, not a veteran's instinct.
Tone traits
The named qualities of your voice, each with do/don't pairs that show the line. "Confident" means nothing until the trait shows what confident sounds like here — and what it doesn't.
Verbal identity
The words half of the brand: messaging house, tone, boilerplates, phrases. Historically underspecified next to visual identity — the verbal identity gap — and now the half that decides consistency.
Version history
The record of how the brand model has changed, and what exactly changed between versions. What makes "which version is current?" a solved question instead of a Slack thread.
Voice rules
The subset of the verbal identity an AI tool applies while writing: tone traits with do/don'ts, context guidance, approved and banned phrases. The first thing a connected tool reads from a brand server.

Stop defining it. Start running it.

Every term above is something SPOKEPROS does on the Free plan — import, structure, publish, connect.

Free plan. We're onboarding in batches — invite codes go out weekly.