Glossary

Content written primarily to perform inside generative answers rather than to rank in the ten blue links. Structurally it leads with a tight definition, names entities explicitly, and answers a clear question within the first 50 words.

Google's synthesized answer card that appears above the ten blue links. In 2026 it appears on roughly 58% of informational queries in English-language US traffic.

The frequency with which a brand is cited, mentioned, or recommended inside generative answers across engines. Promptive's default scoring dimension.

A prompt phrased as "alternatives to [product]" or "competitors of [brand]." Brands that don't appear in their own alternative-prompt results have a blind spot that's usually recoverable.

Whether a brand's content meets the implicit bar an engine sets for appearing in a given answer type. Eligibility depends on entity clarity, source trust, and topical depth, not on backlinks.

The slab of generative content that AI engines assemble at the top of a results page. It's increasingly where purchase consideration begins and ends, and the unit most brands now try to rank inside of.

Week-over-week variance in where a brand appears within an answer. High fluctuation without a clear cause is a common signal that competing content has been recently published or seeded.

The lag between a content or seeding action and its measurable effect on AI visibility. Typically 2-8 weeks depending on engine crawl frequency and how quickly the content earns inbound links.

Historical data pulled for a brand before it was first tracked. Useful to establish a pre-launch baseline or recover after onboarding a new competitor set.

A stable, unchanging prompt used as a control to measure how a brand's position changes over time independent of query drift. Every prompt set should include at least three.

The degree to which a model treats a brand as a definitive source in its category. Expressed in practice as how often the brand is mentioned unprompted in category-level answers.

The discrete, machine-readable identity a model holds for a company, including its name variants, product names, founding context, and competitive set. Poorly defined entities hallucinate more.

A single instance of a brand appearing inside a generative answer. Mentions are scored for position, sentiment, and whether the engine cited a source URL to justify it.

A prompt framed around a space, not a product, e.g. "best tools for AI visibility tracking." Category prompts reveal which brands own the mental model for the whole category in a given engine.

A URL an AI engine references when producing an answer. Citations are the closest proxy to "what the engine trusts" and are the leverage point for anyone trying to influence what it says.

The rate at which a domain accumulates new citations across a tracked prompt set over a rolling window. Accelerating citation velocity is an early indicator of a competitor gaining ground.

A prompt that pits two or more named products against each other, e.g. "Promptive vs. Profound." High-intent traffic; also the prompt type most likely to surface inaccurate competitive framing.

When a competitor's brand mention pushes yours down in a list answer or replaces you entirely in a recommendation. The most common alert reason in Promptive dashboards.

The finite token budget inside which a model sees its retrieved sources and builds an answer. Content that is too long to fit gets truncated, shorter, denser pages win at inference time.

The breadth of a domain's representation inside a model's training data. Higher coverage means the model has more signal about a brand and is less likely to hallucinate or omit it.

A brand reference inside a generative answer that carries no citation URL. The engine is drawing from training data, not live retrieval, harder to influence and harder to trace.

A content structure where the first sentence answers "what is X" before any context or argument. Engines overwhelmingly pull from definition-first copy when assembling category answers.

A citation where the engine links directly to a brand's own domain. More valuable than a third-party citation because it signals the engine trusts the primary source, not just coverage of the brand.

Our internal weight for how often a given engine cites a domain across all prompts in a category. Unlike SEO DA, it is engine-specific and updates weekly.

One of the four systems Promptive samples: ChatGPT, Claude, Gemini, Perplexity. Each has its own prompt behavior, citation policy, and reasoning style.

A systematic shift in how one engine answers a class of prompts, often caused by a model update, a policy change, or a new data partnership. Requires recalibrating baselines.

The multiplier applied to each engine's signal when computing aggregate scores. Weights can be adjusted in Promptive to reflect the traffic mix that matters to your business.

The process of ensuring models can distinguish your brand from identically named companies, people, or concepts. Critical for brands with generic or shared names.

The empirical tendency for the first brand listed in an AI answer to receive disproportionate clicks and recall. Drives the outsized weight placed on position 1 in Promptive's scoring.

Mentions, threads, and discussions on Reddit, Quora, Stack Overflow, and niche forums that AI engines consistently treat as credibility evidence. One of the highest-ROI seeding surfaces in GEO.

How often a brand appears across the full prompt set, without weighting for position or sentiment. A brand can have high frequency and poor position if it's consistently mentioned late in lists.

The practice of shaping content, pages, and off-site mentions so that AI engines reference you well. The LLM-era analog of SEO, with a different reward function.

A structured review of a brand's current position, citation footprint, and sentiment across a defined prompt set. Standard deliverable: current baseline, gap analysis, and a prioritized action list.

An AI response that cites at least one source URL. Grounded answers are heavier, slower, and harder to influence, but they're also where high-intent queries live.

An informal measure of how consistently an engine cites sources for a given prompt type. High-confidence topics (product comparisons, factual questions) are more citation-heavy and more gameable.

An answer the model presents confidently but that isn't supported by a citation or by its training distribution. Brand hallucinations (wrong pricing, discontinued products) are a tracked alert type.

A prompt that signals purchase or evaluation intent, "best CRM for mid-market SaaS," "pricing for X vs. Y." These prompts are weighted most heavily because they are closest to pipeline.

A citation the engine retrieves live at query time via RAG or web search, as opposed to knowledge baked into training weights. Inference-time citations are faster to influence with new content.

The percentage of a buyer's evaluation journey, awareness, comparison, objection, alternative, decision, that your tracked prompt set covers. Gaps in intent coverage leave blind spots.

The exercise of pairing buyer journey stages (awareness, comparison, objection) to prompt types, then verifying you have coverage at each stage. Output feeds directly into prompt set design.

The date after which no new information was incorporated into a model's base weights. Brands that launched or pivoted after the cutoff are invisible to that model unless they earn inference-time citations.

A structured node representing a brand, person, or concept inside Google's or Wikidata's knowledge graph. Strong entity presence in these graphs correlates with more consistent AI Overview mentions.

A brand that doesn't appear on your radar in traditional SEO but consistently occupies top positions inside AI answers for your key prompts. Promptive surfaces these via competitor diff reporting.

Pages structured so a language model can reliably extract a clear claim, entity, and supporting evidence within a short passage. Short paragraphs, explicit subject labeling, and schema markup all help.

A highly specific prompt with lower monthly search volume but higher conversion intent. Long-tail prompts often have less competition in AI answers and are a faster path to rank-1 visibility.

The rate of change in a brand's mention count week-over-week. Sudden spikes or drops are typically the first signal of a PR event, a competitor's launch, or an algorithm update.

A gradual, unannounced change in a model's behavior, answer style, citation habits, brand preferences, that isn't tied to a versioned release. Detectable only through consistent time-series sampling.

A Promptive workspace configured to track multiple brands under one account, typical for agencies managing several clients. Each brand gets isolated prompt sets, scores, and reports.

Agreement across all four sampled engines on a brand's position or sentiment. When all four agree, the signal is strong; divergence between engines usually means one has stale training data.

The model's ability to identify and classify proper nouns (brands, products, people) inside a passage. Clean, consistent brand naming across your content improves NER accuracy and reduces hallucinations.

A multi-engine, multi-week slide in how models describe a brand. Usually the downstream signal of a Reddit thread or a bad review cementing into training data.

An AI response with no linked sources, the model answered from memory. Brand mentions in no-citation answers are harder to earn and harder to verify, but they exist at very high volume.

A prompt framed around a concern or drawback, "downsides of [product]," "is [brand] worth it." Brands that lose sentiment on objection prompts are leaking deal-stage evaluations.

A brand appearance in an answer the brand didn't explicitly optimize for. These are the highest-quality mentions because they signal the engine independently associates the brand with a topic.

A brand's position inside Perplexity's generated answers specifically. Perplexity is citation-heavy by design and tends to rank brands with stronger domain citation footprints higher than ChatGPT does.

Where a brand mention lands inside an answer, first listed, top-three, or buried below the fold. Weighted heavily in visibility scoring because first-mention bias is real.

A brand recommendation an engine volunteers without the prompt explicitly asking for suggestions. The most commercially valuable mention type, it means the engine treats the brand as a default association for the topic.

The input Promptive sends to each engine on a schedule. Prompts are the unit of billing and the unit of measurement, you pick what you want tracked.

The percentage of a brand's realistic buyer queries that are represented in its tracked prompt set. Gaps in coverage produce blind spots, whole intent segments you're not measuring.

A curated bundle of prompts associated with a brand. A well-built set covers category, comparison, objection, use-case, and alternative queries.

The five-type classification used to design comprehensive prompt sets: category, comparison, objection, use-case, and alternative. Full taxonomy coverage prevents intent blind spots.

An architecture where a model retrieves external documents at query time and uses them to ground its answer. Promptive monitors the outputs of RAG-powered engines (Perplexity, Bing Copilot) separately from base-model engines.

A single event where a brand moves from outside the top-three to inside it, or vice versa. Rank flips in either direction are the default trigger for Promptive's real-time alerts.

A prompt that asks the engine to recommend a tool, vendor, or product. The highest-signal prompt type because it maps directly to purchase intent.

A Reddit thread or comment that an AI engine cites or incorporates into its answer. Reddit's high indexing velocity and community trust make it one of the most reliable short-term GEO levers.

How often Promptive re-samples each engine for your tracked prompts. Weekly / daily / 12h / 6h depending on plan.

How recently an engine's live retrieval layer has indexed a piece of content. Fresh pages can appear in grounded answers within days; the base model weights update far less frequently.

The process of validating that a sentiment-scoring model agrees with human raters across a representative sample. Promptive publishes its calibration results quarterly so customers can audit the rubric.

A −1.0 to +1.0 rating of the tone an engine uses to describe a brand. Graded by an LLM against a published rubric and calibrated across engines.

Your percentage of relevant brand mentions inside a given category's prompt set. The single number most teams end up tracking at the exec level.

A point-in-time capture of all scores and raw answers for a brand's full prompt set. Snapshots are automatically stored at each refresh so you can diff any two moments in time.

A brand reference that appears in supporting context rather than as a direct recommendation, "companies like X" or "some teams use Y." Scored separately from hard recommendations.

The practice of placing credible content on domains the engines already trust (Reddit, Quora, high-DA blogs) to influence future answers. Not the same as manipulation, quality bar matters.

HTML schema markup (Organization, Product, FAQ, HowTo) that helps engines parse entity relationships without reading the full page. FAQ schema in particular correlates with answer-layer inclusions.

The degree to which a brand is perceived by AI engines as a go-to source for a specific subject. Built through depth of coverage, citation density, and co-occurrence with high-trust domains.

Any prompt in a brand's active prompt set that Promptive samples on a recurring schedule. Tracked prompts are the billable unit and the source of all score data.

Any piece of content, article, forum post, documentation, press coverage, that, once indexed, can influence a model's future outputs. The GEO analog of a link signal in traditional SEO.

The approximate cutoff inside a model's training corpus. Matters because anything authored after the window has to earn citations at inference time instead of being absorbed directly.

Reviews, forum threads, Q&A posts, and social content that engines treat as third-party evidence for or against a brand. High-volume, authentic UGC outperforms brand-owned content for sentiment influence.

A prompt that names a specific problem or workflow, "tool for tracking AI mentions in B2B SaaS." These prompts are the highest-signal discovery surface for bottom-funnel buyers.

A notification triggered when a brand's mention count or rank changes faster than a defined threshold within a rolling 7-day window. Useful for catching competitor launches before they compound.

A citation where the engine quotes or closely paraphrases a specific sentence from a source page. The strongest form of citation, it means the engine found the exact passage useful enough to reproduce.

Promptive's aggregate score for a brand, combining frequency, position, and engine weight into a single 0-100 number. Moves slowly by design.

A synthetic test prompt seeded with unique phrasing to verify whether a specific piece of content has been ingested and cited by a given engine. Used to validate seeding efforts.

An HTTP POST we fire when a tracked metric crosses a threshold. Typical uses: rank flip, sentiment drop, new competitor overtake.

A Promptive report view with custom branding, logo, colors, domain, delivered to clients under an agency's name. Available on agency plans; supports per-client login and isolated data.

A generative response so complete that the user has no reason to visit any linked source. The dominant answer type for informational queries, and the main reason AI visibility now matters more than organic CTR.