AI Brand Glossary
The Vocabulary of Brand in the AI era.
A
Accessibility Layer (a11y)
The embedded contrast and accessibility data within a colour token. An accessibility layer records WCAG contrast ratios against common surfaces, flags pass or fail against AA and AAA standards, and identifies which text colours are safe to use on that background. Without this layer, AI tools making colour decisions have no way to know whether their choices are accessible. With it, accessibility becomes a constraint the system enforces rather than a check the designer remembers to run.
AI Brand Assistant
A tool that evaluates creative assets against a brand's defined guidelines, rather than generating new content. An AI brand assistant checks whether a proposed design, image, or piece of copy is on-brand. It does not replace the creative. The distinction matters: evaluation tools empower designers; generative tools can threaten them.
AI Brand Execution
The process of using AI to produce creative outputs (copy, images, code, design) that conform to a brand's defined standards. AI brand execution is only as reliable as the structure of the brand data it has access to. Without a structured semantics layer, AI guesses at brand intent rather than reads it.
AI Hallucination (brand context)
When an AI generates content that is factually inconsistent with the brand's defined guidelines. This can mean using the wrong shade of a colour, applying a tone of voice the brand explicitly avoids, or inventing a visual style that was never approved. Brand hallucination is not a creative failure. It's a structural one. The AI lacked the right context.
AI-Native Brand System
A brand guidelines system designed from the ground up to be consumed by both humans and AI tools. An AI-native system doesn't just look good in a browser. It stores brand data in structured, queryable formats that AI can read and act on with precision. This is what Era 3 brand infrastructure looks like. See also: Era 3.
AI Semantics Gap
The disconnect between how brands document their guidelines (human-readable prose, PDFs, portals) and what AI needs to execute on them (structured, semantic, queryable data). Most brand guidelines today fall into this gap. The AI can read the words, but it cannot understand the intent. Closing the AI Semantics Gap is the core problem that AI-native brand systems are built to solve.
Anti-Descriptor
A negative instruction included in brand data or AI prompts that explicitly states what the brand is not. Anti-descriptors are often more useful than positive descriptors for AI tools, because they constrain the output range. "NOT corporate stock photography" is often clearer than "authentic imagery."
Asset Library
A collection of approved brand assets (logos, fonts, icons, photography, templates) stored and made accessible to the team. In AI-native systems, the asset library is connected to the semantic data layer, so AI tools know not just where files are, but when and how to use them.
AudioDNA
A structured description of a brand's audio character, formatted for AI comprehension. Audio DNA defines the sonic identity of a brand: its tempo range, instrumentation preferences, mood descriptors, and explicit anti-descriptors. Like Image DNA, it is designed to be injected directly into AI audio generation workflows to keep brand sound consistent across touchpoints. A brand that never thought about audio still has one once AI starts generating it.
B
Bilingual Brand
A brand system that communicates effectively to two audiences at once: humans who read and interpret guidelines, and AI systems that query and execute on them. A bilingual brand publishes a human-readable portal and a machine-readable data layer from the same source of truth. Neither is an afterthought.
Brand Architecture
The structure that defines the relationship between a parent brand and any sub-brands, products, or regional identities that sit beneath it. In AI-native systems, brand architecture has data implications: each level of the hierarchy may need its own structured export, with clear declarations of lineage and inheritance.
Brand Archetype
A universally recognised character type used to give a brand a consistent personality framework. Common archetypes include the Ruler, the Explorer, the Caregiver, the Creator, and the Jester. In AI-native brand systems, archetypes are encoded as structured data rather than left as prose descriptions. When an AI knows a brand is an Explorer archetype, it has a richer, more reliable framework for generating consistently characterful outputs than if it were only given a list of adjectives.
Brand Check
An evaluation of a creative asset (image, document, presentation, social post) against the brand's defined standards. A brand check can be run manually (a designer reviewing against the style guide) or automatically (an AI tool comparing the asset to structured brand data). Automated brand checks are only reliable when the underlying brand data is structured and semantic.
Brand Consistency
The quality of a brand that remains recognisable, coherent, and true to its defined standards across different contexts, tools, teams, and time. Consistency used to depend on people remembering the rules. In a distributed, AI-assisted workflow, it increasingly depends on the structure of the system.
Brand Data
The totality of a brand's defined elements, stored as structured, machine-readable information. Brand data is not a PDF or a Figma file. It's the underlying record that describes colours, typography, voice, imagery, motion, and their relationships, in formats that both humans and AI systems can access. See also: brand.json, design tokens.
Brand Document
A human-readable record of brand guidelines, typically a PDF, a Figma file, or a portal page. Brand documents are the dominant format of Era 1 and Era 2 brand systems. They describe the brand well for humans but carry almost no useful structure for AI tools.
Brand Equity
The value a brand holds in the minds of its audience, built through consistent experience over time. Brand equity is eroded by inconsistency. As AI generates more brand touchpoints, maintaining equity requires the same rigour that a strong brand system has always demanded, now applied to machine outputs as well as human ones.
Brand Governance
The policies, processes, and tools a brand uses to ensure its standards are applied correctly across all outputs and teams. In the AI era, brand governance extends to controlling how AI systems access and use brand data, not just how human teams do.
Brand Guidelines
A set of rules that define how a brand should look, sound, and behave. Brand guidelines have existed in some form since the era of print manuals. Their function has not changed. What has changed is who needs to read them. Humans have always been the primary audience. Now AI systems are too.
Brand Infrastructure
The technical and operational foundation that makes a brand system work at scale. Brand infrastructure includes the data formats, protocols, and integrations that allow brand data to flow consistently into every tool and workflow. In Era 3, brand infrastructure is as important as the visual identity itself.
Brand Intelligence
The capacity of an AI system to understand and apply a brand's guidelines with nuance and accuracy. Brand intelligence is a product of data quality: the more structured and semantic the brand data, the higher the brand intelligence of any AI tool consuming it.
Brand Personality
The human characteristics attributed to a brand: its tone, warmth, confidence, wit. Brand personality is easy to describe in prose. It is hard to encode in a way that AI can execute reliably. The most effective approach translates personality into specific behavioural rules: "use contractions," "keep sentences under 18 words," "lead with the benefit, not the feature."
Brand Portal
A web-based interface where a team can access, read, and reference brand guidelines. Brand portals are the primary output of Era 2 brand systems. They are well-designed for human navigation but do not expose brand data in structured formats that AI tools can query.
Brand Token
A named, reusable unit of brand data: a specific colour value, a type size, a spacing unit. Brand tokens are the building blocks of a structured brand system. They can be primitive (a raw value, like a hex code) or semantic (a named role, like "colour-background-interactive"). Semantic tokens are far more useful for AI, because they carry intent alongside value.
Brand Voice
The distinct way a brand uses language: its word choices, sentence patterns, tone, and personality. Brand voice is one of the most difficult guidelines to encode for AI, because it is inherently qualitative. The most effective encoding translates voice into behavioural rules and scored examples, not just adjectives.
brand.json
A structured, machine-readable file that encodes a brand's core guidelines in JSON format. Typically hosted at /.well-known/brand.json, it makes brand data discoverable and queryable by AI agents and tools. brand.json is the emerging standard for AI-readable brand identity, equivalent to what robots.txt is for web crawlers.
C
Canonical Block
In a structured brand system, the single authoritative source of truth for a specific brand element: the canonical colour palette, the canonical voice rules. All other instances derive from or reference the canonical block. This architecture ensures that changing a value in one place propagates correctly across the system.
Channel Adaptation
The defined way a brand's voice or visual identity shifts to suit different contexts: social media, product UI, long-form content, email. A well-structured brand system encodes channel adaptations explicitly, so AI tools can apply the right version of the brand for the right context without requiring a new prompt each time.
Color Ramp
A graduated scale of tints and shades derived from a single base colour. Color ramps give a brand flexibility across light and dark surfaces while staying within the defined palette. In a structured token system, a ramp is defined once and referenced everywhere it's used.
Color Token
A design token that stores a colour value alongside semantic metadata: its name, role, accessibility contrast data, and descriptive context. A colour token is more useful than a hex code because it carries intent alongside the value.
Component Fabrication
An AI failure mode where the model invents a new UI component or design element that doesn't exist in the approved system, rather than using what's available. Component fabrication happens when AI tools are not constrained to a defined component library. The fix is explicit enumeration: give the AI an authorised inventory and instruct it to use only those elements.
Context Window
The amount of information an AI model can hold in working memory during a single session. Brand guidelines passed as raw text to an AI tool compete for space in the context window alongside the task itself. Structured brand data, accessed via an MCP server or API, stays outside the context window and is available on demand. This is a more reliable architecture for brand consistency at scale.
D
Design Token
A named, structured unit that stores a single design decision: a colour, a type size, a spacing value, a motion duration. Design tokens are the bridge between a visual design and its implementation in code. In AI-native brand systems, tokens are extended with semantic metadata, usage guidelines, and accessibility data, so AI tools can use them correctly without manual instruction.
Drift Signal
An early warning sign that AI-generated copy is moving away from the brand's defined voice. Drift signals are specific linguistic patterns that indicate off-brand direction: using passive voice when the brand is active, reaching for jargon when the brand is plain, becoming formal when the brand is conversational. Encoding drift signals in a voice system gives AI evaluation tools something concrete to check against, rather than relying on a subjective sense that something "feels off."
Design Token Community Group (DTCG)
An open standards body defining a common format for design tokens across tools and platforms. The DTCG specification provides a shared vocabulary for how tokens should be structured, named, and exported. Compliance with DTCG makes brand data more portable across the AI tooling ecosystem.
Descriptive Name
A plain-language label for a brand element that helps AI tools understand its character, not just its value. For colours, a descriptive name like "dark forest green" or "warm sand" is more useful for image generation tools than a hex code, which they cannot interpret. Descriptive names are a core component of the semantics layer.
E
Era 1: Print Manuals
The first era of brand guidelines, roughly 1970s to 2000s. Authoritative, static, physical brand bibles designed for a world of print and human execution. IBM's identity standards, NASA's graphics manual, and similar documents are canonical examples. Built for permanence, not built for machines.
Era 2: Digital Documents and Portals
The second era of brand guidelines, roughly 2000s to 2023. This era has two phases. Era 2A: the PDF, the print manual digitised but otherwise unchanged. Era 2B: the brand portal, platforms like Frontify, Notion, and Brandpad that made guidelines easier to navigate and update for humans. Both phases remain fundamentally human-readable. AI can scrape the text, but it cannot understand the structure.
Era 3: AI-Native Systems
The third era of brand guidelines, from 2024 onwards. Structured, machine-readable, queryable brand data that both humans and AI systems can access from the same source of truth. The human reads the portal. The AI queries the API. This is the era that Sameness is built for.
F
Foundation Token
A primitive design token that stores a raw brand value: a specific hex code, a typeface name, a base spacing unit. Foundation tokens are the source layer of a token system. All semantic tokens and component decisions reference foundation tokens rather than introducing new values. This ensures that a single edit at the foundation level propagates correctly across the entire system.
Formality Calibration
An explicit setting in a brand's voice guidelines that defines where the brand sits on a formal-to-casual scale. Formality calibration is best expressed as a number (e.g. 0.4 on a 0–1 scale, where 0 is most casual and 1 is most formal) alongside a description. A numeric value gives AI tools something measurable. A description gives them context.
G
Guardrails (brand)
Explicit constraints built into brand data or AI system prompts that prevent the AI from producing outputs outside approved parameters. Brand guardrails define the edges of what's acceptable: which colour combinations are forbidden, which phrases are banned, which logo modifications are prohibited. Guardrails are the enforcement layer of a brand system. Without them, AI defaults to approximation.
H
Hierarchy (typographic)
The structured relationship between type levels (heading, subheading, body, caption) that guides the reader's eye and communicates relative importance. In a structured brand system, typographic hierarchy is defined not just as a visual rule but as a semantic relationship, so AI tools know which level to apply in which context.
I
Image DNA
A structured description of a brand's photographic and visual style, formatted for AI comprehension. Image DNA typically includes camera references, lighting profiles, colour grading descriptions, composition style, mood descriptors, and negative constraints. It's the foundation for consistent AI-generated or AI-curated imagery.
Identity Decisions
The recorded rationale behind key brand choices: why this typeface was chosen over another, why the brand sits at this position on the formality scale, why a specific colour was defined the way it was. Identity decisions are not just documentation for humans. In structured brand systems, the reasoning behind a decision is data. AI tools that have access to the why behind brand choices can apply them with more nuance and are less likely to override them when generating content.
Independent Block
In a structured brand system, an instance of a brand element that is defined locally for a specific context (a sub-brand, a campaign, a regional variant) using only the approved foundation tokens. An independent block declares its lineage explicitly: it's a defined variation, not a contradiction.
J
JSON-LD
JavaScript Object Notation for Linked Data. A format for encoding structured data in a way that is readable by both machines and search engines. In brand systems, JSON-LD is used to mark up brand information on web pages so that AI crawlers and search engines can understand and index the data accurately.
L
Lineage Declaration
An explicit statement in a brand's structured data that describes where a value comes from and why it differs from the canonical definition. Lineage declarations are what make brand extensions and sub-brands coherent. Without them, a different proportion of colours on a campaign page looks like an error. With a lineage declaration, it reads as an intentional, documented variation.
llms.txt
An emerging standard for making web content more discoverable and parseable by large language models. A brand's llms.txt file, hosted at the root of their domain, signals to AI agents what content is available and how to interpret it. It is to LLMs what robots.txt is to search engine crawlers.
Logo Clearspace
The protected area around a logo within which no other design elements may appear. Clearspace is typically defined as a ratio, for example a minimum equal to twice the height of the logomark on all sides. Defining clearspace as a ratio rather than a pixel value makes it AI-usable, because ratios scale proportionally with context.
M
Machine-Readable
A format or structure that can be interpreted and acted on by software without human translation. A PDF is not machine-readable in a useful sense. A structured JSON file with semantic metadata is. Machine-readability is the prerequisite for any brand system that needs to work with AI tools.
MCP (Model Context Protocol)
An open protocol that allows AI tools to query external data sources in real time. In brand systems, an MCP server exposes brand data (colour tokens, voice rules, logo variants, component guidelines) as a queryable endpoint. Any LLM or AI tool connected to the MCP server can request brand context on demand, without needing the guidelines in its context window.
MCP Server (brand)
A server that implements the Model Context Protocol for a brand's guidelines data. When an AI tool queries a brand MCP server, it receives structured brand data (precision values, semantic descriptions, and relationship logic) that it can use immediately to produce on-brand outputs. This is the primary integration point between a brand system and the AI tooling ecosystem.
Motion Token
A design token that defines a motion or animation parameter (a duration, an easing function, a delay rule) as a named, reusable value. Like colour and type tokens, motion tokens make animation decisions consistent and transferable across tools. In AI-native systems, motion tokens include descriptive metadata that helps AI tools understand the character of the motion, not just its technical parameters.
N
Nearest Established Name
A field in a brand's colour data that maps a specific brand colour to the closest widely-recognised colour name: "British racing green," "burnt sienna," "dusty rose." This field exists because AI image generation tools cannot interpret hex codes. They respond to descriptive language. A nearest established name gives them a precise, culturally legible anchor for the brand colour.
Negative Prompt
An instruction in an image generation tool that specifies what the output should not contain. Negative prompts are often as important as positive prompts for maintaining brand consistency. "NOT corporate stock photography, NOT cold lighting, NOT heavily retouched" prevents generic drift more reliably than positive descriptors alone.
O
OKLCH
A perceptual colour space that describes colour using three values: lightness, chroma (saturation), and hue. Unlike hex or RGB, OKLCH is designed so that equal numerical steps produce equal perceived differences in colour. This makes it significantly more useful for AI colour reasoning than hex codes, which have no perceptual structure. When a brand's colour tokens include OKLCH values alongside hex, AI tools can reason about colour relationships (lighter, darker, more saturated, complementary) with far greater accuracy.
P
Perceptual Descriptor
A qualitative, human-readable characteristic that describes a colour's visual character rather than its value. Perceptual descriptors use words like "deep," "muted," "cool-toned," "earthy," or "saturated." They exist because AI image generation tools respond to language, not numbers. A hex code carries no perceptual information. A set of perceptual descriptors tells an AI what a colour actually feels like, which is what it needs to reproduce the brand colour with fidelity.
Photography Harmony
A structured definition of how a brand's photography palette relates to its core colour system. Photography harmony encodes the colour grading, temperature, and tonal range that photographic content should share with the brand's defined colours. It ensures that imagery selected or generated by AI feels visually coherent with the rest of the brand, not like it came from a different visual world. Part of the colour system, not the imagery system, because the connection to the palette is what makes it meaningful.
Prompt Fragment
A pre-written, reusable piece of AI prompt language that encodes a specific brand parameter. Prompt fragments for image generation might describe a camera reference, a lighting profile, or a colour grading instruction. For text generation, they might encode a voice rule or a structural preference. Prompt fragments are a practical output of the semantics layer, ready to be injected into any AI workflow.
R
Reference Continuity
The degree to which an AI-generated asset should match a reference image's full visual language versus only its key motifs. Full reference continuity means the AI should replicate the overall composition, palette, lighting, and mood of the reference. Motif-only continuity means the AI should carry specific visual elements (a shape, a colour, a texture) into a new composition without reproducing the original. Defining which level of continuity is required prevents AI tools from either over-copying or under-referencing a brand's established visual language.
S
Semantic Drift
An AI failure mode where the output gradually shifts away from the specific, defined brand standard toward a more generic, commonly-occurring variant. A brand's specific deep green drifts toward a standard forest green. A precise, direct voice drifts toward generic professional warmth. Semantic drift happens because generic variants have stronger training data associations. The fix is explicit boundaries and descriptive anchors in the brand data.
Semantic Token
A design token that carries an explicit statement of intent alongside its value. Where a primitive token might be named green-700, a semantic token is named colour-background-interactive and includes a description: "Use for all primary call-to-action buttons." Semantic tokens are the correct abstraction for AI brand execution, because AI tools need to understand where a value belongs, not just what it is.
Semantics Layer
The layer of a brand data system that encodes meaning, intent, and context alongside raw values. A hex code without a semantics layer is just a number. The same hex code with a semantics layer carries a descriptive name, a nearest established name, accessibility contrast data, usage rules, and negative constraints. The semantics layer is what transforms brand data from a list of values into something AI can execute against with fidelity.
Structured Brand Data
Brand guidelines encoded in a machine-readable, semantically-enriched format. Structured brand data is not a Figma file or a PDF. It is the underlying information, stored in formats like JSON, design tokens, or MCP-queryable endpoints, that can be reliably read, queried, and acted on by AI tools. It is the material of Era 3 brand systems.
T
Three-Layer Export
An export architecture for brand data that produces three distinct layers from a single source: a precision layer (exact values for code and design tools), a semantics layer (descriptive context for AI tools), and a relationship layer (dependencies, selection logic, and contextual rules for AI orchestration). A three-layer export is more useful than a single-format export because different consumers need different representations of the same brand data.
Token Misuse
An AI failure mode where the model uses a primitive or incorrect token instead of the appropriate semantic token. For example: using a raw colour value directly in a component rather than the designated colour-background-interactive token. Token misuse happens when AI tools lack the semantic context to understand a token's intended role. The fix is structured token definitions with explicit usage guidelines.
Tone Modulation
The structured set of rules that define how a brand's voice adapts across different channels and contexts. Tone modulation is distinct from tone of voice: tone of voice defines the brand's core character, tone modulation defines how that character is expressed differently in a social post versus a product UI versus a long-form article. In a structured brand system, tone modulation is encoded as channel-specific rules (maximum sentence length, formality shift, register adjustment) that AI tools can apply without requiring manual re-briefing for each context.
Tone of Voice
The defined way a brand communicates in language: its register, rhythm, vocabulary preferences, and personality. Tone of voice guidelines are most useful for AI when they include specific behavioural rules ("use contractions," "lead with the insight, not the feature"), explicitly banned phrases, and scored examples of on-brand versus off-brand copy. Adjective descriptions alone are not enough.
Tone Samples
Scored, annotated examples of on-brand and off-brand copy, used to calibrate AI writing tools to a brand's defined voice. A good tone sample set includes the example text, a score (e.g. 1–5), and an explanation of why it is or isn't on-brand. Tone samples are the highest-leverage component of any voice encoding system. They demonstrate the brand's voice through contrast, not just description.
Type Guardrail
An explicit rule that prevents specific typographic misuses in AI-generated layouts or content. Type guardrails go beyond defining a type system: they name the failure modes the system must not produce. Common type guardrails include hierarchy flattening (using the same type level for elements of different importance), hierarchy inversion (using a smaller or lighter style where a dominant one is required), tier collapse (reducing a multi-level system to fewer levels than defined), and context mismatch (applying a display typeface in a body copy role). Encoding these as named rules gives AI tools something specific to avoid, not just something to aim for.
V
Verbal Identity
The full linguistic expression of a brand: its name, tone of voice, vocabulary rules, messaging hierarchy, and the language patterns that make it recognisable. Verbal identity is as structured as visual identity, but it is often less rigorously documented. In AI-native brand systems, verbal identity receives the same structured treatment as colour or typography.
Visual Identity
The visual expression of a brand: its logo, colour palette, typography, imagery style, and layout principles. Visual identity has traditionally been the most rigorously documented part of a brand system. In AI-native systems, the data structure behind visual identity is as important as the visual decisions themselves.
Voice Dimension
A defined axis along which a brand's voice can be measured, for example: formal versus casual, warm versus direct, simple versus detailed. Voice dimensions give a brand's tone of voice system quantitative structure. Each dimension is assigned a position (often a value between 0 and 1), which AI tools can use to calibrate their outputs more precisely than adjective descriptions alone permit.
Video DNA
A structured description of a brand's video and motion visual character, formatted for AI comprehension. Video DNA defines the cinematic language of the brand: pacing, shot types, colour treatment, lighting style, and audio direction. Where Image DNA governs still photography, Video DNA governs moving image. As AI video generation tools become standard in content workflows, Video DNA is what keeps brand-generated video from looking like it came from a template.
W
Well-Known URI
A standardised location on a domain where specific files are hosted to make information discoverable by automated systems. /.well-known/brand.jsonis the proposed location for a brand's machine-readable data, following the same convention as/.well-known/security.txt` and other established standards.