The vocabulary of brand intelligence, defined plainly.

The vocabulary of the record.

These are the working terms of brand intelligence: what Jinn extracts, checks, and enforces. Most are house terms, and each is defined against the machinery that implements it.

• Brand DNAJinn AgentsBrand DNA is a working record of a brand that an AI system can read and act on: named signals extracted from what the brand has actually made, not a prompt describing the vibe.Read the definition →
• Brand recordJinn AgentsA brand record is the artifact itself: the named, inspectable signals a brand is modeled as, versioned so they can change on purpose. It is the thing every Jinn product reads, distinct from Brand DNA, which is the idea of modeling a brand this way.Read the definition →
• Brand signalJinn AgentsA brand signal is one named, extracted fact about a brand, like its founding story, a primary color, or a banned word. It is the smallest unit of a brand record: addressable on its own, inspectable, and traceable to where it came from.Read the definition →
• Brand wearJinn AgentsBrand wear is the moment a generic AI system starts writing from a specific brand's record instead of its own defaults. The same brief, run twice: one run is what any model writes by default, the other is authored from the brand's signals.Read the definition →
• Voice checkGhostA voice check is an automated pass that scores a piece of generated content against a brand's real voice before it ships. It compares the draft to a profile built from the brand's own writing and returns how close the fit is.Read the definition →
• Banned listGhostA banned list is the set of words and claims a brand never uses, enforced as a hard check rather than a suggestion. A draft that contains a banned term does not just get a lower score. It fails, and it cannot be published.Read the definition →
• ProvenanceChartProvenance is the recorded origin and history of a piece of content or a claim: where it came from, who touched it, and what changed, kept so it can be audited later rather than taken on trust.Read the definition →
• Render scoringVermeerRender scoring is grading a generated visual against the brand's record instead of eyeballing it. A finished ad is checked against the real product and the brand's rules, so a wrong product, a distorted label, or copy sitting over a face gets caught before anyone ships it.Read the definition →
• Verified factFamaA verified fact is a claim about a brand that a human confirmed against a source and entered into the brand's fact canon. It starts as a proposal, becomes active only when someone approves it, and can be rejected or superseded. The opposite of a vibe.Read the definition →
• Fact canonFamaA fact canon is the maintained set of verified facts an AI is allowed to state about a brand. Each fact is proposed, confirmed by a human, and kept current, so generation draws on truths a person signed off rather than whatever a model assumed.Read the definition →
• Remediation playbookFamaA remediation playbook is a generated, gated set of fixes for a visibility gap: the step after measurement. Once an audit finds where AI engines get a brand wrong, the playbook turns each gap into publishable content, grounded in the brand's verified record.Read the definition →
• Topic gapFamaA topic gap is a subject your audience asks about where competitors have content and you do not. Jinn finds them by clustering competitor pages into topics, then checking which of those topics your own pages fail to cover.Read the definition →
• Citation sourceFamaA citation source is a domain an answer engine actually leans on when it composes an answer in your category. Not where you wish it looked: the real publisher pages behind the sentences an engine returns when a customer asks.Read the definition →
• CitabilityGhostCitability is how liftable a piece of content is for an answer engine: whether each section opens with a self-contained answer, carries a takeaways box and an FAQ, and backs every statistic with a date, a source, and attribution. A structural property, not a style opinion.Read the definition →
• Shoot bibleVermeerA shoot bible is the visual playbook a brand's image generation runs against: the scene, framing, lighting, and mood a render has to honor. Vermeer builds it from the brand's record, so every image is produced against the brand's own direction rather than a model's default.Read the definition →
• Render calibrationVermeerOn the launch roster. Its definition ships with its page.Definition in progress
• Edit driftVermeerEdit drift is how far a human's edit moved a generated image away from the product it started with. When someone re-directs a shot, the change to the scene is expected, but the product itself should stay identical. Edit drift measures whether it did.Read the definition →
• Disclosure-first engagementGhostDisclosure-first engagement is community participation with your affiliation stated plainly, where a reader actually sees it. It is the posture that refuses astroturfing and sockpuppets: you earn a community's trust by being useful and open about who you are, not by hiding it.Read the definition →
• Voice profileGhostA voice profile is the style half of a brand's record: how it writes, not what it writes about. Cadence, sentence rhythm, register, the words it reaches for and the ones it never uses, measured from real writing rather than described in adjectives.Read the definition →
• Generic tellsGhostGeneric tells are the phrases and shapes that mark writing as unedited AI output: the same three openers, headings phrased as questions, a rash of em dashes, tidy emoji bullet lists. None is wrong alone. Together they are why so much AI text reads as nobody's.Read the definition →
• Fail-closed verificationChartFail-closed verification is a checker built so a broken check can never count as a pass. If the verifier errors, times out, or returns something it cannot read, the safe default is fail, not succeed. A claim is treated as unsupported until it is actually shown to hold.Read the definition →
• AtomizerGhostAn atomizer turns one source, a talk, a video, a long post, into many distinct content angles. Not the same idea reworded seven ways, but genuinely different theses the source can support, each grounded in what the source actually said before anything is written from it.Read the definition →
• Content cadenceGhostOn the launch roster. Its definition ships with its page.Definition in progress
• brand.jsonJinn Agentsbrand.json is a machine-readable, public projection of a brand's record, built for other agents to read. It is a deliberately bounded slice: enough for an agent to represent the brand faithfully, with the proprietary strategy held back. A public face for the record, not the whole record.Read the definition →
• Prompt injectionJinn AgentsPrompt injection is content written to hijack an AI's instructions: hidden text that tries to make a model ignore its real task and follow the attacker's instead. We define it here to name a line, because how you treat it says what kind of system you are building.Read the definition →

Terms without links are the launch roster. Each gets its page as its definition ships.

The contested set, defined honestly.

The industry is still arguing about most of these. We define them the way we use them: plainly, without inflating what anyone can actually measure.

• GEOFamaGEO, generative engine optimization, is the practice of structuring content and facts so AI answer engines cite you when they compose an answer. It is an industry category, still loosely defined, and used almost interchangeably with AEO.Read the definition →
• AEOFamaAEO, answer engine optimization, is the practice of shaping your content and facts so an answer engine picks you when it responds to a question directly. It is the query-shaped cousin of GEO, and the two names are used almost interchangeably.Read the definition →
• AI visibilityFamaAI visibility is whether and how AI engines mention, cite, and describe your brand when customers ask about your category. It is the AI-era version of the question SEO used to answer: when someone looks, do you show up, and is what they see true?Read the definition →
• Citation vs mentionFamaA mention names your brand in an answer. A citation links your brand as the source the answer is built from. Engines treat the two differently, and so should you: being named is visibility, being cited is authority.Read the definition →
• Citation rateFamaCitation rate is how often an answer engine cites your domain as a source, measured across a fixed set of questions and tracked over time. A rate, not a count: the share of the answers where you are the source, not the raw number of citations.Read the definition →
• Share of voice, AIFamaShare of voice in AI answers is your slice of the answers engines give in your category, measured against your competitors. Not whether you appear at all, but how much of the conversation engines are having about your category is about you versus everyone else.Read the definition →
• Query fan-outFamaQuery fan-out is expanding one customer question into the many phrasings an engine actually sees. Nobody asks in exactly one way, so measuring or optimizing against a single query misses most of the answers being given about your category.Read the definition →
• llms.txtFamallms.txt is a plain-text file at the root of your site that tells AI agents what your site is and where the important pages are. Think robots.txt, but for answer engines and agents instead of search crawlers, pointing them at the truth rather than gating them out.Read the definition →
• AI crawlerFamaAn AI crawler is a bot an answer engine sends to read your site. Each one has its own name and its own behavior, and they do not all respect robots.txt the same way, so welcoming or blocking them is a per-bot decision, not one switch.Read the definition →
• GroundingChartGrounding is tying generated text to sources it can actually point to. An ungrounded claim is one the model produced with nothing real behind it, and ungrounded output is where hallucination starts. Grounded writing can show its work; ungrounded writing only sounds like it can.Read the definition →
• HallucinationFamaA hallucination is when an AI states something false with confidence. For a brand, the stakes are specific: the false thing an engine tells your customers about your price, your product, or your history, in the same certain voice it uses for the truth.Read the definition →
• Answer engineFamaAn answer engine is a system that responds to a question directly instead of returning a list of links. ChatGPT, Perplexity, Gemini, Google's AI Overviews: they read the web, compose an answer, and hand it back. It is the surface all of this optimizing is aimed at.Read the definition →
• AI OverviewsFamaAI Overviews is Google's answer layer: the AI-written summary that now sits above the blue links for many searches. It reads the web, composes an answer, and names a few sources. For most searchers it is the answer, and the links underneath are an afterthought.Read the definition →
• CrawlabilityFamaCrawlability is whether an AI bot can actually reach and read your pages, checked rather than assumed. AI crawlers do not run JavaScript, so a site that looks fine in a browser can be a blank shell to them. Crawlability is what they really see.Read the definition →
• Structured data (for AI)FamaStructured data for AI is machine-readable page facts, usually JSON-LD, that state plainly what a page is and what it claims, so an engine can quote you precisely instead of guessing from prose. It is the difference between hoping a crawler infers a fact and handing it the fact.Read the definition →