Marketing glossary.

Emerging manifest file at the site root that declares an AI agent contract: supported tasks, content endpoints, authentication, MCP server location, license. The agentic-web companion to robots.txt and llms.txt.

Related: MCP · llms.txt · Agentic Browsing

Browsing performed by an AI agent on behalf of a user. Differs from traditional crawling in that the agent reasons about task completion (booking a call, comparing options) and may execute multi-step interactions. Lighthouse now scores sites on agentic-browsing readiness.

Related: MCP · agent.json

Google Search's AI-generated summary that appears above traditional search results. Pulls from pages already ranking in positions 1-5 and favors content with Featured Snippet eligibility. Now consumes around 60% of clicks on high-intent queries.

A 2-3 sentence self-contained paragraph in the first 200 words of a page that directly answers the title's question. The most likely passage to be quoted by AI engines and Google's Featured Snippets. Must be under 70 words and name the entity explicitly.

Related: GEO · AEO · Featured Snippet

Optimizing content for direct extraction by AI answer engines and voice search. Focuses on question-form headings, atomic facts, FAQ schema, and concise definitions in the first 200 words. Sister discipline to GEO.

Related: GEO · Answer Capsule · FAQPage Schema

A single-idea sentence that makes complete sense without surrounding context. AI engines preferentially quote atomic facts because they can be lifted into an answer without follow-up explanation. Statistics, thresholds, and timelines are the highest-value atomic-fact patterns.

schema.org type that declares a page's location in the site hierarchy. Renders breadcrumb trails in search results and helps AI engines understand site structure.

Related: Schema.org

Content elements specifically designed to be quoted by AI engines. The four highest-ROI patterns are original data tables, named thresholds, before/after examples, and process timelines. Each article should include at least one.

Median time before an AI citation drops out of typical response sets unless the source is refreshed. In the Xpand Q1 2026 corpus, citation half-life across the five tracked engines was 91 days.

Related: GEO · Citation Bait

Anthropic's web crawler for Claude. Identified by User-Agent ClaudeBot/1.0 and anthropic-ai. Allowing it in robots.txt is required to be cited in Claude-generated answers.

Open-source web archive that captures billions of pages monthly and ships them as a public corpus. Used to bootstrap most foundation model training datasets. The CCBot user-agent identifies its crawler.

Related: GPTBot · ClaudeBot

schema.org type for declaring a published dataset. Carries name, description, temporalCoverage, variableMeasured, creator, license, and distribution. Eligibility signal for Google Dataset Search and citation surfaces in AI research products.

Related: Schema.org · GEO

schema.org type for defining a glossary or technical term with a name, description, and optional containing DefinedTermSet. Cited frequently by AI engines for glossary-style queries. Underpins the /topics and /glossary indices.

Related: Schema.org · FAQPage Schema

Numerical vector representation of a piece of text, image, or other content. Two embeddings that are close in vector space represent semantically similar content. Foundation of vector search and RAG.

Related: Vector Search · RAG

The strength of an organization's presence across structured data sources (Wikidata, Crunchbase, LinkedIn, schema.org Organization). AI engines weight this heavily before deciding to cite a brand. Brands present on 4+ authoritative platforms are roughly 2.8× more likely to appear in AI answers.

Related: sameAs · Wikidata · Knowledge Graph

JSON-LD structured data that marks a page's FAQ section as machine-readable Question/Answer pairs. Generates accordion-style rich results in Google and is one of the highest-leverage AEO patterns. Pages with FAQPage schema get cited in Perplexity around 40% more often.

Related: JSON-LD · Schema.org · Rich Results

The practice of engineering content, schema, and entity authority so a brand is cited inside generative AI answers. Targets ChatGPT, Gemini, Perplexity, Claude, and Microsoft Copilot. Built on entity disambiguation, structured data, citation-bait content, and llms.txt.

Related: AEO · Answer Capsule · Entity Authority · FAQPage Schema

Google's User-Agent for AI training (Gemini, Vertex AI). Separate from Googlebot. You can allow Googlebot for Search ranking but block Google-Extended to keep your content out of training data, though that also limits Gemini citation.

OpenAI's web crawler used to populate ChatGPT's training data and live citations. Identified by the User-Agent string GPTBot/1.0. Sites that block it via robots.txt are excluded from ChatGPT-generated answers.

Related: ClaudeBot · PerplexityBot · robots.txt

When an AI model generates fluent text that is factually wrong or fabricated. Strong entity authority and structured data reduce hallucination risk by giving the model verifiable anchors. The reason brands invest in GEO infrastructure.

Related: GEO · Entity Authority

schema.org type for step-by-step instructional content. Each step is wrapped in a HowToStep object. Surfaces in Google and Gemini answer cards for procedural queries.

Related: Schema.org · FAQPage Schema

Google's database of entities (people, companies, places) and the relationships between them. AI engines like Gemini ground their answers against the Knowledge Graph before crawling fresh content. Strong sameAs links and Wikidata entries feed the Knowledge Graph.

Google's branded sidebar that surfaces for entity queries. Pulled from Knowledge Graph and Wikidata. Indicates Google considers the entity well-established.

Related: Knowledge Graph · Wikidata

Long-form companion to llms.txt. Concatenates page content into a single text file intended for offline LLM ingestion or local model fine-tuning. Lives at the site root alongside llms.txt.

Related: llms.txt · GEO

A plain-text file at the root of a domain (/llms.txt) that describes the site's purpose, key pages, and entity definitions for AI crawlers. The AI equivalent of robots.txt, early adoption signals organizational maturity to crawlers.

Model Context Protocol. An open standard for connecting AI assistants to external data sources and tools through a structured server interface. Lets a website publish endpoints that AI agents can query without browser scraping.

Related: agent.json · Agentic Browsing

Microsoft's AI assistant, integrated into Bing and Edge. Uses the Bing index for web citations. Sites must be indexed in Bing Webmaster Tools to be eligible for Copilot citations. Bing indexing is separate from Google indexing.

OpenAI's dedicated search-grounding crawler. Separate from GPTBot (which collects training data), OAI-SearchBot reads pages in real time when a ChatGPT user asks something that requires live information.

Related: GPTBot

Google's ability to rank individual H2 sections of a page independently of the page as a whole. Means each section needs its own atomic fact, named entity, and standalone-readable structure to compete for queries.

Perplexity AI's web crawler. Two variants: PerplexityBot (training) and Perplexity-User (live citation lookup). Citation rates correlate strongly with named-author bylines, visible publish dates, and high domain authority.

Retrieval-Augmented Generation. Pattern where an LLM first retrieves relevant passages from a knowledge base, then generates an answer grounded in those passages. Underpins Perplexity, AI Overviews, and most enterprise AI search products.

Related: Embedding · Vector Search

schema.org SpeakableSpecification that marks text on a page as suitable for voice assistant readouts. Tells voice surfaces (Alexa, Google Assistant, Siri) which sentences to read aloud for query answers.

Related: Schema.org · AEO

Search that compares query and document embeddings rather than keywords. Returns semantically similar passages even when no exact word overlap exists. Powers most modern AI retrieval stacks.

Related: Embedding · RAG

Free knowledge base maintained by the Wikimedia Foundation that stores structured data about entities (people, organizations, places, concepts). The single strongest correlate with citation in AI engines: 71% of cited entities in the Xpand Q1 2026 corpus had a Wikidata record, versus 11% in the non-cited control.

Related: Knowledge Graph · Entity Authority · GEO

The visible clickable text of a hyperlink. Search engines use anchor text as a relevance signal for the destination page. Over-optimization (exact-match keyword stuffing) can trigger penalties.

Related: Internal Linking

Inbound link from another website. Historically the strongest ranking signal in SEO. Still important, though structured data and entity signals have closed some of the gap in 2026.

Related: Domain Authority

The single authoritative URL for a piece of content, declared via <link rel="canonical">. Tells search engines which version to index when multiple URLs serve the same or similar content. Prevents duplicate content penalties and consolidates ranking signals.

Google's three performance metrics: LCP (Largest Contentful Paint), INP (Interaction to Next Paint), CLS (Cumulative Layout Shift). Direct ranking factor. Targets: LCP under 2.5s, INP under 200ms, CLS under 0.1.

The number of URLs a search engine will crawl on a site in a given window. Constrained by server capacity and the engine's allocation. Large sites optimize by ensuring high-priority URLs are easy to discover and not blocked by thin or duplicate pages.

Related: Indexing · robots.txt

Third-party score (Moz, Ahrefs equivalents) estimating a domain's overall ranking strength based on backlink profile and historical performance. Not a Google signal directly, but correlates with ranking outcomes.

Related: Backlink

Google's quality framework: Experience, Expertise, Authoritativeness, Trust. Used by quality raters and the algorithm to evaluate content. Real engagement examples, named author credentials, sourced statistics, and authoritative inbound links all signal E-E-A-T.

A direct-answer box at the top of Google search results, pulled from one ranked page. Eligibility requires answering the query in 40-60 words within the first H2 section. Featured-snippet pages are the primary candidates for AI Overviews.

An HTML attribute that tells search engines which language and regional variant of a page to show users in different countries. Specified as <link rel="alternate" hreflang="en-za" href="...">. Critical for sites with country-specific or city-specific content.

The process by which a search engine adds a page to its searchable database after crawling. A page is only available to rank once indexed. Google Search Console reports indexing status under Coverage.

Related: Crawl Budget · Sitemap

Links from one page of a site to another page on the same site. The strongest tool for distributing topical authority across a domain. Hub-and-spoke patterns (topic hub linking to detail pages) outperform flat structures.

Related: Topic Cluster

JavaScript Object Notation for Linked Data, the recommended format for embedding schema.org structured data in HTML. Lives in a <script type="application/ld+json"> tag and is parsed by search engines and AI crawlers but invisible to users.

Google's policy of using the mobile version of a page as the primary version for ranking and indexing. In effect for all new domains since 2023. Sites must serve full content on mobile, not stripped versions.

Related: Core Web Vitals

Enhanced search results powered by structured data. Examples: FAQ accordions in SERP, review stars, recipe cards, video carousels, breadcrumbs. Tested via Google's Rich Results Test.

Related: Schema.org · JSON-LD

A plain-text file at the domain root that tells crawlers which paths they can access. Can target specific User-Agent strings (Googlebot, GPTBot, ClaudeBot). Blocking AI crawlers here removes the site from AI-generated answers.

A shared vocabulary for structured data on the web, jointly maintained by Google, Microsoft, Yahoo, and Yandex. Implemented via JSON-LD blocks. The most cited types for marketing sites: Organization, Service, FAQPage, Article, BreadcrumbList, LocalBusiness.

The practice of structuring a site and its content to rank in traditional search engine results pages. Built on three pillars: technical (crawlability, speed, schema), on-page (titles, headings, internal links), and off-page (backlinks, citations).

An XML file that lists all public URLs on a site for search engine discovery. Should include lastmod timestamps, changefreq, and priority. Submit to Google Search Console and Bing Webmaster Tools after major content changes.

A content architecture where one comprehensive pillar page (2,500+ words) links to many supporting articles, and each supporting article links back to the pillar. Builds topical authority and signals depth to both Google and AI engines.

Average Order Value. Total revenue divided by total orders in a period. Levers to lift AOV: upsells, bundles, free-shipping thresholds, post-purchase modules. Each percentage point of AOV is leveraged across all marketing spend.

Customer Acquisition Cost, total marketing and sales cost divided by new customers acquired in the period. The denominator question is whether to include team salaries, software, and overhead, be consistent in how you define it.

Google's framework for collecting analytics data while respecting GDPR consent choices. Requires explicit user consent before sharing personal data with Google. Sites without Consent Mode v2 see significant tracking degradation in EU traffic.

Server-to-server tracking that sends conversion events directly from your server to ad platforms (Meta CAPI, LinkedIn CAPI, Google Enhanced Conversions). Bypasses iOS 14, Safari ITP, and ad blockers that broke client-side pixels. Required for accurate attribution in 2026.

Related: Server-Side Tracking · Consent Mode v2

Cost Per Action or Acquisition. Total spend divided by the number of completed conversions. Should be compared against allowed-CPA derived from LTV and target margin, not against vanity benchmarks.

Cost Per Click. Amount paid per click in an auction-based ad system. Manual CPC sets a max bid; Smart Bidding (Target CPA, Target ROAS) adjusts CPC algorithmically.

Cost Per Lead, ad spend divided by leads generated. Useful at the channel level but only meaningful when tied to lead quality. CPL with terrible MQL→SQL conversion is just a vanity metric.

Cost Per Mille (cost per 1,000 impressions). The auction unit on Meta, LinkedIn, and most paid social. Useful for benchmark comparison but the bottom-funnel metric (CPA, ROAS) is what you optimize against.

Click-Through Rate. Clicks divided by impressions, expressed as a percentage. Channel-dependent. Search 3-6%, Display 0.3-0.8%, LinkedIn 0.4-1.2%, Meta 1-3%.

Google Ads feature that uses first-party customer data (email addresses, phone numbers) to target ads and create lookalikes. Privacy-respecting alternative to third-party audiences in the cookieless era.

Related: First-Party Data

Audience built by an ad platform's algorithm to mirror the attributes of a seed audience (existing customers, high-value users). Available in Meta, Google, LinkedIn, TikTok. Best built from cleaned high-LTV seeds.

Lifetime Value. The total expected revenue from a customer over the duration of the relationship. Critical input to CAC payback calculations. SaaS LTV is typically modeled as ARPU divided by monthly churn rate.

Related: CAC · LTV:CAC Ratio

Ratio of customer lifetime value to customer acquisition cost. SaaS healthy band is 3-5x. Below 2x means the unit economics break; above 5x usually signals under-investment in growth.

Related: LTV · CAC

Meta's 1-10 rating of how well your CAPI data matches Facebook user accounts. Higher = better attribution. Score above 6 is the working minimum; below that, the optimization signal degrades. Requires hashed email, phone, and other PII fields.

Marketing Efficiency Ratio. Total revenue divided by total marketing spend across all channels in a period. Captures blended performance including organic uplift. The truth-teller metric ROAS pretends to be.

Related: ROAS · CAC

Keywords that prevent ads from showing for queries containing those terms. Essential cost-control lever in Search. A weekly negative keyword review typically saves 10-20% of wasted spend.

Google Ads' AI-driven campaign type that auto-allocates budget across Search, Shopping, YouTube, Display, Discover, and Gmail. Best for e-commerce with strong shopping feed and conversion data. Risk: spend can drift to low-quality placements without proper exclusions.

Return On Ad Spend, revenue divided by ad spend over a defined attribution window. The top-line efficiency metric for performance marketing. Defended ROAS requires accurate revenue attribution, which is why server-side tracking matters.

Running Google Tag Manager on a server (Google Cloud Run, Stape) instead of in the browser. The server sees first-party data and forwards events to ad platforms via CAPI. Restores ROAS visibility lost to browser privacy changes.

Google Ads automated bidding strategies that use machine learning to set CPCs in real time. Strategies include Target CPA, Target ROAS, Maximize Conversions, and Maximize Conversion Value.

Related: Performance Max

Conversion attributed to an impression that was seen but not clicked, within a defined window (typically 1-30 days). Common in YouTube, Display, and Meta reporting. Treat with caution; can overstate channel contribution.

Related: Attribution Window

An AI program that can take actions in pursuit of a goal, not just generate text. Marketing agents typically research prospects, draft outreach, manage workflows, and report findings. Differ from chatbots in autonomy.

Related: Agentic Browsing

Using an LLM to score inbound replies or form submissions for fit and intent before routing to a human. Output is structured JSON: fit (1-10), intent (1-10), summary. Lifts MQL→SQL conversion when paired with real-time enrichment.

Automated assignment of new leads to the right sales rep based on territory, company size, ICP fit, or product interest. Sub-2-minute routing dramatically improves contact rate compared to manual queue assignment.

Visual workflow automation (formerly Integromat). Real-time triggers, data transforms, and API connectors. Best for marketing operations, lead routing, and data sync. SaaS pricing model (per-operation) makes it pricier than n8n at scale.

Open-source workflow automation platform with 400+ integrations. Self-hostable. Native AI nodes for OpenAI and Anthropic. Our primary engine for complex multi-step automations because it scales without per-execution pricing.

An AI model's ability to call external functions or APIs to retrieve data or execute actions. Underpins most agent architectures. Anthropic's Tool Use API, OpenAI's Function Calling, and the MCP standard are the primary surfaces.

Related: AI Agent · MCP

HTTP callback that one system sends to another when an event occurs. Glues together SaaS tools in marketing workflows. Most automation platforms (Zapier, Make, n8n) build on webhooks under the hood.

Coordination of multiple steps across multiple systems into a single automated process. Marketing automation platforms (HubSpot, Marketo) orchestrate within their walls; n8n and Zapier orchestrate across vendors.

Quick-deploy automation across 5,000+ apps. Best for simple trigger-action workflows. Multi-step Zaps and conditional Paths handle moderate complexity. Pricing scales with task count, so Zapier is the right choice for low-volume orchestration.

B2B prospecting database covering ~280M contacts with verified emails, phone numbers, and firmographic data. Used for ICP list-building and enrichment. The depth advantage over alternatives shows up in mid-market and enterprise tiers.

AI-powered enrichment platform that combines 50+ data sources with natural-language transformations. Best for building hyper-targeted lists and personalized outbound sequences. Premium pricing reflects depth of coverage.

Gradually ramping up sending volume on a new domain to build sender reputation with email providers. Typically 4-6 weeks from 0 to full volume. Required before any cold outreach to avoid landing in spam.

Process of gradually increasing send volume from a new sender domain over 3-6 weeks while exchanging emails with seeded inboxes that mark them as not-spam. Required before any cold outbound from a fresh domain.

Related: Domain Warming

Ideal Customer Profile, the company shape that's most likely to buy and most profitable to serve. Defined by industry, company size, tech stack, geography, and trigger events. Sharper ICP definition is the single biggest lift for outbound conversion.

Sending cold email from multiple email addresses across multiple domains in parallel to distribute volume and reduce per-account flagging risk. Standard practice for outreach at scale.

Cold-email automation platform with built-in inbox rotation, warmup, and AI personalization. Popular among growth teams running 3-10K sends per week.

Related: Smartlead · Domain Warming

Signals that indicate a company is actively researching solutions in your category. Sources: G2 visits, content downloads, technographic changes, hiring patterns. Intent + ICP fit is the qualifier for cold outreach.

Percentage of cold emails that receive a reply (positive or negative). Strongest single signal of campaign quality. Well-built cold campaigns hit 8-15% reply on tight ICP; broken campaigns sit at 1-3%.

LinkedIn's premium search and outreach product. Used for ICP-aligned prospect list building and account-based targeting. Most teams pair it with an enrichment tool to extract email addresses from profile data.

Related: Apollo

Reputation metric (0-100) assigned to an email sending IP or domain by inbox providers, based on historical complaint, bounce, and engagement rates. Drives whether email lands in inbox or spam.

Cold-email automation platform competitive with Instantly. Strong on multi-inbox sending, master inbox view, and integration with deliverability-monitoring tools.

Related: Instantly

Three DNS records that authenticate your sending domain. SPF lists allowed senders, DKIM signs each email cryptographically, DMARC tells receivers what to do with failures. All three must be configured for B2B email to land in inbox.

Enterprise B2B data platform with deep firmographic, technographic, and contact data. Best-in-class for large account-based targeting at the enterprise tier. Expensive relative to Apollo and Clay.

Related: Apollo · Clay

Splitting traffic between two or more page variants to measure which converts better. Statistical significance requires sample size based on baseline conversion rate and minimum detectable lift, usually 1,000+ conversions per variant.

The discipline of increasing the percentage of visitors who take a desired action. Built on hypothesis-driven A/B testing, heatmaps, session recordings, and funnel analysis. Compound CRO improvements often beat new traffic acquisition for ROI.

Behavior detected when a visitor's cursor moves toward the browser close or back button. Used to trigger last-chance modals or offers. Effectiveness has declined as users learn to dismiss them.

Anything in a conversion flow that slows or stops the visitor from completing the action: long forms, surprise costs, unclear copy, slow load. Reducing friction is the highest-ROI CRO discipline.

A visualization of where users click, hover, and scroll on a page. Tools: Hotjar, Microsoft Clarity, FullStory. Best for diagnosing UX issues and identifying ignored CTAs or dead zones below the fold.

The principle that a landing page headline should mirror the ad copy or CTA that brought the user there. Mismatch causes immediate bounce. The single highest-leverage fix on most paid traffic landing pages.

The smallest improvement an A/B test is designed to detect at a given sample size and confidence level. A test with low traffic can only detect large lifts; trying to detect a 5% lift on a 1,000-visit page needs months.

Related: A/B Testing · Statistical Significance

Conversion-rate-optimization practice that designs and tests for mobile experiences first because mobile typically carries 60-75% of traffic. Desktop adaptations come second.

Playback of an individual user's session, every click, scroll, and keystroke (with PII redacted). Reveals friction points that quantitative metrics miss: confused form fields, broken interactions, abandonment patterns.

Confidence that an observed difference between A/B test variants is not due to random chance. Standard threshold is 95% confidence. Stopping a test before significance is reached produces false positives.

Related: A/B Testing

A primary call-to-action that follows the user as they scroll, typically a fixed bottom bar on mobile or a sticky sidebar on desktop. Increases bottom-funnel conversion by keeping the action one click away regardless of scroll depth.

Event-based product analytics platform, the main alternative to Mixpanel. Stronger on cohort visualization and retention curves; weaker on web analytics. Used by mature product-led growth teams.

Related: Mixpanel

The time period during which a touchpoint can be credited with a conversion. Common windows: 1-day click, 7-day click, 28-day click. Different windows produce dramatically different ROAS numbers, be consistent across reports.

Google Cloud's serverless data warehouse. Native sink for GA4 event-level data. Used as the source for Looker dashboards and custom attribution models. Costs scale with query volume, not seats.

Related: Looker

Grouping users by a shared attribute (sign-up week, acquisition channel, plan tier) and tracking their behavior over time. The honest way to read product retention and revenue retention.

Data collected directly by your own systems (CRM, server-side analytics, email opens). Not subject to third-party cookie restrictions. Increasingly the only reliable signal as iOS, Safari, and Chrome restrict cross-site tracking.

Examining drop-off rates at each step of a multi-step user flow. Identifies the single largest friction point. Different from cohort or retention analysis: funnels are cross-sectional, cohorts are longitudinal.

Google Analytics 4, the event-based successor to Universal Analytics. Tracks user journeys across web and app, integrates with Google Ads, and supports server-side data via Measurement Protocol. Required for modern attribution in Google's ad ecosystem.

Google's tag management system. Lets non-developers deploy and update tracking pixels (GA4, Meta, LinkedIn, etc.) via a web UI without site code changes. Server-side GTM containers run the same logic on a server for first-party data.

Google Cloud's enterprise BI tool. Connects to BigQuery, GA4, and most data warehouses. Used for the operator dashboard layer that sits above raw analytics tools. Heavier setup than Mixpanel or Amplitude but more flexible at scale.

Event-based product analytics. Best for SaaS funnel analysis, retention curves, and feature adoption metrics. Pairs with GA4 for marketing-side analytics, they answer different questions.

Distributing conversion credit across multiple touchpoints in the customer journey, rather than crediting only the last click. Common models: linear, time-decay, position-based. Requires unified data from CRM, ads platforms, and analytics.

The single metric that best captures the value a product delivers to users. Examples: Spotify's time-listening, Airbnb's nights booked, Slack's weekly active teams. Anchors prioritization across product and growth teams.

Tying marketing channel investment to opportunities generated and revenue closed downstream. Requires CRM integration, opportunity-source tagging, and a clean lead-to-opportunity funnel definition. The metric every CFO actually wants to see.