AI marketing & GEO, defined.
Plain-English definitions of the terms an indie founder running their own growth actually runs into — generative engine optimization, human-in-the-loop, programmatic SEO, and more. Each entry is a quick definition with a pointer to the deeper guide.
GEO
AI citation
An AI citation is when an AI assistant or answer engine (such as ChatGPT, Perplexity, Google AI Overviews, or Gemini) names your page or brand as a source in its generated answer, usually as a linked reference or an in-text attribution. It is the AI-search equivalent of a ranking position: the moment your content is selected to support the model's answer.
AI crawler
An AI crawler is an automated client that fetches your pages on behalf of an AI system rather than on behalf of a search results page. The name covers at least three different jobs — collecting text that may train a future model, building the index an answer engine retrieves from, and fetching one page live because somebody just asked about it — and each job usually has its own bot with its own name. So blocking "AI crawlers" is not one decision. It is three, and only one of them is the one that decides whether an assistant can cite you.
Answer Engine Optimization (AEO)
Answer Engine Optimization (AEO) is the practice of structuring and formatting content so that answer engines — AI assistants, voice assistants, and AI-powered search features like Google's AI Overviews and Perplexity — can pull it directly as the answer to a user's question. Instead of optimizing to rank a blue link, you optimize to be the answer that gets read aloud, summarized, or cited.
Entity SEO
Entity SEO is the practice of helping search engines and AI systems unambiguously identify your brand, people, products, and topics as distinct "entities" in a knowledge graph, and understand how they relate to other entities. Instead of optimizing for keyword strings, it optimizes for the real-world things those keywords represent.
Featured snippet
A featured snippet is the boxed answer Google shows at the very top of search results, above the standard organic links, automatically pulled from a page Google judges to best answer the query. It is often called "position zero" because it sits above the #1 ranked result.
Generative Engine Optimization (GEO)
Generative Engine Optimization (GEO) is the practice of structuring your content and brand presence so AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude — cite you in their generated responses. Where SEO competes for blue links on a results page, GEO competes to be the source an AI quotes when it answers a question directly.
Google AI Overviews
Google AI Overviews are the AI-generated answer boxes that appear at the top of Google search results, where Gemini synthesizes a direct answer from multiple web pages and links out to a handful of cited sources. They replace the old "ten blue links" experience for many queries with a summary the user reads without clicking through.
llms.txt
llms.txt is a plain-text, Markdown-formatted file placed at a website's root (yoursite.com/llms.txt) that gives large language models a curated map of your most important pages, in clean prose stripped of HTML, navigation, and ads. It was proposed in 2024 by Jeremy Howard of Answer.AI as an open convention to help AI systems find and understand your key content.
Retrieval-Augmented Generation (RAG)
Retrieval-Augmented Generation (RAG) is an AI architecture that fetches relevant documents from an external knowledge source at query time and feeds them to a language model so its answer is grounded in those retrieved facts rather than its training data alone. It is the core mechanism behind how AI search engines like ChatGPT, Perplexity, and Google's AI Overviews pull in live web content and cite their sources.
Schema markup
Schema markup is structured data code (usually written in JSON-LD) that you add to a web page to label what its content means — for example, marking up a product's price, an article's author, or an FAQ's questions and answers. It gives search engines and AI answer engines machine-readable facts they can parse, trust, and cite, instead of guessing from raw text.
Speakable schema
Speakable schema is a Schema.org structured-data property (using the SpeakableSpecification type) that marks the specific sections of a web page best suited to be read aloud by text-to-speech engines, such as voice assistants and answer engines. You point it at the most quotable parts of your page using an HTML element id, a CSS selector, or an XPath.
Topical authority
Topical authority is the degree to which a search engine or AI model treats your site as a genuine expert on a whole subject area, not just a single page. You earn it by publishing deep, interconnected content that covers a topic comprehensively rather than chasing one keyword at a time.
Zero-click search
A zero-click search is a search where the user gets their answer directly on the results page (via an AI Overview, featured snippet, knowledge panel, or direct answer box) and never clicks through to a website. The query is resolved entirely inside the search interface, so no traffic reaches the underlying source.
AI agents
Agent autonomy spectrum
The agent autonomy spectrum is the range of how much independent decision-making and action an AI agent is allowed to take, from low-autonomy assistants that only suggest and need human sign-off on every step, to high-autonomy systems that plan and execute on their own. Where you place an agent on that spectrum is a deliberate design choice that trades speed for control and reversibility.
Agent orchestration
Agent orchestration is the coordination layer that decides which AI agent acts, when, on what data, and with what authority so that multiple specialized agents work as one governed system instead of in isolation. It handles task routing, context handoff, and the start/retry/finish lifecycle of each agent toward a shared goal.
Agent-washing
Agent-washing is the practice of marketing ordinary automation, scripted chatbots, or rule-based workflows as autonomous "AI agents" they are not. The term, popularized by Gartner in 2025, describes vendors who overstate how independent, adaptive, or intelligent their software actually is to ride the agentic-AI hype.
Agentic workflow
An agentic workflow is a goal-directed AI process where an AI agent decides the sequence of steps at runtime — planning, calling tools, observing the results, and adjusting — instead of running a fixed script you wrote in advance. The core loop is: set a goal, reason about it, act, observe the outcome, and repeat until the goal is met.
AI agent
An AI agent is a software system, usually powered by a large language model, that perceives a situation, reasons about a goal, and takes multi-step actions using tools (APIs, databases, browsers) to accomplish a task with limited human prompting. Unlike a chatbot that just answers, an agent plans, acts, observes the result, and adapts in a loop until the job is done.
AI copilot
An AI copilot is an AI assistant that works alongside a person inside their workflow, suggesting, drafting, and surfacing insights in real time while the human stays in control of every final decision. Unlike a fully autonomous AI agent, a copilot assists rather than acts on its own.
AI Growth Officer
An AI Growth Officer is an AI agent that acts as the lead, orchestrating role for a startup's growth function: it takes a goal, breaks it into marketing work across channels like SEO, social, email, and PR, and coordinates specialist agents to draft and run that work. Think of it as a head-of-growth layer in software that plans and delegates, rather than a single tool that does one task.
AI marketing agent
An AI marketing agent is a software agent powered by a large language model that carries out marketing work — researching, drafting content, running outreach, and analyzing results — by planning and taking actions through connected tools rather than just answering prompts. In practice the most reliable ones operate inside an approval gate: they propose and draft, and a human signs off before anything goes live.
AI teammate
An AI teammate is an AI system that works alongside people as a collaborator you direct and review, rather than a tool you operate manually or an autonomous worker you set loose. It takes on scoped tasks, proposes work for you to approve, and stays accountable to a human owner.
Approval gate
An approval gate is a checkpoint in an automated or AI-driven workflow where a human must review and explicitly approve a proposed action before it can run. It splits the work into "draft" and "do" so a person stays in control of anything consequential or irreversible.
Evidence chain
An evidence chain is the linked trail of sources, data, and reasoning steps that backs a claim or an agent's action, so anyone can trace the conclusion back to where it came from. In AI marketing, it means every recommendation or draft cites the underlying source, making the work auditable instead of a black box.
Hallucination (AI)
A hallucination is output an AI model presents as fact that is not grounded in anything it was actually given — an invented statistic, a link to a page that does not exist, a customer quote nobody said, or a report of work the model never did. It is not a malfunction in the ordinary sense: a language model predicts plausible text, so a fluent wrong answer comes out of the same machinery as a fluent right one. That is why a hallucination arrives in the same confident voice as everything else, and why the defence is verification rather than firmer instructions.
Human-in-the-loop (HITL)
Human-in-the-loop (HITL) is a design pattern where a person reviews, approves, or corrects an AI system's output before it takes effect, so the model proposes and a human makes the final call. In AI agents it usually means the agent drafts an action and pauses for human sign-off instead of acting fully autonomously.
Model Context Protocol (MCP)
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to outside tools and data through one common interface, so a capability written once can be used by any application that speaks the protocol. It standardizes the wiring between a model and the systems it reads from and acts on — not the judgement about what it should do with them.
Prompt injection
Prompt injection is an attack in which instructions hidden inside content an AI agent reads — a web page, a PDF, a product review, an inbound email — are followed by the model as if you had typed them yourself. It only becomes dangerous once the agent has tools: the same sentence that is harmless in a chat window becomes an instruction to send, publish, or spend the moment the model can act.
Marketing
App Store Optimization (ASO)
App Store Optimization (ASO) is the practice of improving an app's visibility and conversion in the App Store and Google Play — tuning the keywords, title, icon, screenshots, and ratings so the app ranks for what users search and turns store visits into installs. It is the app-store equivalent of SEO: half of it is being found, and half of it is being chosen once you are.
Brand Awareness
Brand awareness is the extent to which your target market recognizes your brand and recalls it when they have the problem you solve. It is a lagging, top-of-funnel measure — a result of being repeatedly visible and useful rather than a channel you run directly — and it ranges from aided recall, recognizing you when prompted, to unaided recall, thinking of you unprompted.
Brand Voice
Brand voice is the consistent personality your company's writing carries — the vocabulary it reaches for, the claims it refuses to make, how formal it is, how its sentences run — held steady across every channel and every person who writes for you. Tone shifts with the situation; voice is what stays the same underneath it.
Content cluster
A content cluster (or topic cluster) is a group of interlinked pages on one subject: a broad pillar page that gives the overview, plus several cluster pages that each go deep on a specific subtopic, all linking to and from the pillar. The structure signals to search engines and AI answer engines that you cover a topic thoroughly.
Conversion Rate Optimization (CRO)
Conversion Rate Optimization (CRO) is the practice of increasing the share of visitors who take a desired action (sign up, start a trial, buy) by improving page copy, design, flow, and offer rather than just driving more traffic. It is the disciplined cycle of analyzing user behavior, forming hypotheses, testing changes, and keeping the ones that lift conversion.
Dark social
Dark social is sharing that happens in private channels — direct messages, group chats, Slack, email, private communities — where the click arrives with no referrer, so genuine word-of-mouth traffic is recorded as Direct. The visits are real and often the best-converting ones you get; only the source is invisible.
Domain Rating (DR)
Domain Rating (DR) is an Ahrefs metric, scored 0 to 100, that measures the relative strength of a website's backlink profile based on the quantity and quality of other domains linking to it. A higher DR signals a stronger, more authoritative link profile relative to other sites.
Ideal Customer Profile (ICP)
An Ideal Customer Profile (ICP) is a description of the exact type of company or person who gets the most value from your product and is the most profitable, retainable customer to acquire. It defines who you sell to (firmographics, role, pain, budget, behavior) so every marketing and sales effort targets the right buyer instead of everyone.
Internal linking
Internal linking is the practice of connecting pages within your own website using hyperlinks, so that related content references related content. It helps search engines and AI engines crawl your site, understand how topics fit together, and distribute ranking signals (link equity) across pages.
Jobs to Be Done (JTBD)
Jobs to Be Done (JTBD) is a way of describing demand by the progress a customer is trying to make, rather than by who they are or which features they ask for. The unit of analysis is the job — the situation someone is in, the struggle that pushes them to look for something new, and the outcome they want — which is why two buyers with identical firmographics routinely hire completely different products for the same job.
Lead magnet
A lead magnet is something useful you give away in exchange for a contact detail, almost always an email address. The trade is the entire mechanic: the visitor gets a template, checklist, calculator, or report immediately, and you get permission to reach them again — which converts anonymous traffic into an audience you own rather than rent from a search engine or a feed.
Negative keyword
A negative keyword is a term you tell an ad platform to exclude, so your ad is not shown for a search containing it. It is the control that decides who does not see your ad — and on a small budget it shapes results more than the keywords you add, because a wasted click is paid for out of the same money as a real one.
Owned media
Owned media is any channel where you control both what gets published and your access to the people who receive it: a website on a domain registered to you, an email list you can export, your own app. It is one part of the owned, earned and paid model, alongside earned media — attention other people give you through press, reviews, links and shares — and paid media, the attention you buy. A social media profile is often counted as owned, but it behaves more like rented space: you write the posts, and the platform decides who sees them and on what terms.
Positioning
Positioning is the deliberate choice of how you want your market to understand what your product is, who it is for, and why it is the better option for them. It sets the frame — the category you compete in and the alternative you beat — so your messaging, pricing, and channels all have something true to point at. Strong positioning makes the right prospect think "this is for me" in seconds; weak positioning leaves them guessing what you even are.
Programmatic SEO (pSEO)
Programmatic SEO (pSEO) is the practice of generating hundreds or thousands of search-optimized pages automatically by combining a single reusable template with a structured dataset, so each page targets a distinct long-tail query (like "[tool] integration" or "best X in [city]"). Instead of writing every page by hand, you build the template once and let data fill it, producing pages at scale.
Retargeting
Retargeting is paid advertising shown only to people who have already interacted with you — visited a page, used the app, or appeared on a list you uploaded. A tag on your site drops those visitors into an audience, and your ads are then served to that audience instead of to strangers. It is the cheapest-looking line in a media plan and the one an early-stage startup is least likely to be able to run.
Search intent
Search intent (also called user intent) is the underlying goal behind a search query: what the person actually wants to accomplish when they type or speak it. It is usually sorted into four types: informational (learning something), navigational (reaching a specific site), commercial (researching before a purchase), and transactional (ready to buy or act).
Social Listening
Social listening is watching public conversations — on social platforms, forums, review sites and communities — for mentions of your brand, your competitors and the problem you solve, then reading across many of them for a pattern you can act on. It is usually separated from social monitoring by what happens next: monitoring answers the individual post, listening aggregates. The binding constraint in practice is access rather than analysis, because you can only listen where a platform lets your tool read, and the places worth hearing are often the hardest to read.
Social Proof
Social proof is the effect where people look to the actions and opinions of others to decide what to do — in marketing, it is the testimonials, reviews, ratings, usage numbers, and endorsements that reassure a hesitant buyer that people like them already trust you. It works because a stranger cannot verify your claims directly, so evidence that others got value becomes the shortcut they use to say yes.
UTM Parameters
UTM parameters are five optional tags you append to a link's query string — utm_source, utm_medium, utm_campaign, utm_term and utm_content — so your analytics files the visit under a named source instead of inferring one. They are labels you write yourself, not a tracking technology: nothing verifies them, they describe the link you built rather than the journey somebody took, and they exist on exactly one URL before they are gone.
Growth
A/B Testing
A/B testing is running two versions of the same page, email or flow at the same time, splitting visitors between them at random, and keeping whichever version measurably wins. The randomisation and the simultaneity are the method: they are what stop a difference in the audience, the day or the season being mistaken for a difference in the design. Its unstated requirement is volume — a test only answers anything once enough people have converted inside each arm, which is why most early-stage sites cannot honestly run one.
Affiliate Marketing
Affiliate marketing is a performance-based channel where a third party — a publisher, creator, review site or other business — earns a commission for each sale or signup they drive to you through a unique tracked link. It is a pay-for-results distribution model: you owe nothing until an affiliate produces the outcome you agreed to pay for, which is what separates it from advertising you pay for up front.
Aha Moment
The aha moment is the point when a new user first experiences the core value of your product — the instant the thing they signed up for actually clicks. It is the specific realization that turns a curious signup into someone with a reason to come back, and it is the target that good onboarding and activation are built around.
AI-native startup
An AI-native startup is a company built from the ground up with AI as a core part of the product, the team, and the day-to-day operating model, not bolted on as a feature later. The defining trait is leverage: a tiny team ships and grows like a much larger one because AI does work that used to require headcount.
Build in public
Build in public is the practice of openly sharing the journey of building a product or company — shipping updates, metrics, wins, and failures in real time, usually on social platforms like X/Twitter, LinkedIn, or a newsletter. The goal is to attract an audience, build trust, and turn followers into early users while you are still building.
Bullseye Framework
The Bullseye Framework is a method for choosing marketing channels, introduced in the 2015 book Traction by Gabriel Weinberg and Justin Mares. It works in three concentric rings — brainstorm every one of the nineteen traction channels, run cheap tests on the three most promising, then focus your effort on the single channel that testing proves is working — so a company tests broadly but commits narrowly.
Cold email
Cold email is a personalized, unsolicited outreach message sent to a prospect you have no prior relationship with, usually in B2B sales, to start a conversation rather than close a deal on the spot. The aim is a reply, call, or demo from a well-targeted contact who fits your ideal customer profile.
Community-Led Growth
Community-led growth is a go-to-market motion in which a group of users — rather than a sales team or an advertising budget — is the primary way people discover, adopt, and stay with a product. Members answer each other's questions, publish templates and tutorials, and recommend the product to peers, so the community acts as both an acquisition channel and part of the retention loop. Having a Discord is not the same thing: in community-led growth the community is the mechanism, not a support annex bolted onto a funnel.
Content Velocity
Content velocity is the rate at which you publish new content over time — for example, eight blog posts or twenty landing pages per month. It measures consistent output, not a single burst, and it matters because both search engines and AI answer engines reward sites that cover a topic deeply and update it regularly. Higher sustainable velocity, paired with quality, is how a small site builds topical authority faster than competitors.
Digital PR
Digital PR is the practice of earning coverage, mentions, and backlinks from online publications, blogs, podcasts, and newsletters by creating something worth writing about — original data, a strong story, expert commentary, or a useful free tool. For a founder it is modern link building: the earned links raise search authority, and the mentions build brand and the entity signals that AI engines trust when they decide who to cite.
Dogfooding
Dogfooding is the practice of running your own business on the product you sell, so that you hit its problems before your customers do. The name comes from the older phrase "eating your own dog food". For a small software company it doubles as a growth practice: the daily use produces bug reports, feature priorities and honest material to write about, all from work you were going to do anyway.
Founder-Led Sales
Founder-led sales is the early-stage practice of the founder personally running the sales process — sourcing conversations, giving demos, handling objections, and closing — rather than delegating it to a sales hire. Its purpose is as much learning as revenue: because the founder can change the product and the pitch in response to what buyers say, these first deals surface who actually buys, what convinces them, and which objection repeats — the knowledge a salesperson would later need to sell the product at all.
Fractional CMO
A fractional CMO is a senior marketing executive who leads a company's marketing strategy on a part-time, contract basis, giving startups and small teams head-of-marketing expertise without the cost or commitment of a full-time hire. The word "fractional" describes the time commitment, not the depth of involvement.
Freemium
Freemium is a pricing model that gives a usable core version of a product away for free, indefinitely, and charges for advanced features, higher usage limits, or team capabilities. It works as a growth engine only when the free tier is generous enough to attract many users yet limited enough that a meaningful share of them hit a wall worth paying to remove.
Go-to-Market (GTM) Strategy
A go-to-market (GTM) strategy is the plan for how a company reaches its target customers and turns them into users or buyers — who the product is for, how it is positioned, which channels and pricing move it, and how a prospect actually becomes a customer. For an early-stage startup it is less a big document than a sequence of bets: a narrow ideal customer profile, a sharp message, and one or two channels to prove before scaling.
Growth loop
A growth loop is a self-reinforcing system where the output of one cycle of users or activity feeds back in as the input for the next cycle, so growth compounds on itself instead of relying on a one-time burst of acquisition. Unlike a linear funnel that ends when a user converts, a loop reinvests each result — a new signup, a piece of content, a referral — to drive the next round of growth.
Influencer Marketing (Creator Marketing)
Influencer marketing — now more often called creator marketing — is the practice of partnering with people who have earned an audience's trust so their recommendation carries your product to that audience. What you are really paying for is not reach but borrowed credibility: the endorsement converts because the followers already believe the person making it, which is why a niche creator with a few thousand engaged followers often outperforms a celebrity with millions of passive ones.
Lifecycle email
Lifecycle email is a sequence of automated, behavior-triggered emails that move a user through each stage of their relationship with your product, from welcome and onboarding to activation, upgrade, retention, and win-back. Unlike one-off broadcasts, each message fires in response to where someone is in their journey.
Minimum Viable Product (MVP)
A minimum viable product (MVP) is the smallest version of a product that still delivers real value to a real user and produces a decision-grade answer to the riskiest question you have. It is defined by the question it settles, not by how few features it contains — an MVP nobody can actually use teaches you nothing, and a version one with everything in it is no longer minimum.
One-person company
A one-person company is a business founded, owned, and run by a single person who stays the sole strategic operator while leaning on automation, AI tools, and a few contractors instead of hiring a full team. The founder makes every product, marketing, and business decision; "one-person" describes who runs it, not a vow to never grow.
Open core
Open core is a business model in which the heart of a product is published under an open-source licence that anyone can read, run and fork, while the company sells what sits around it — usually a hosted version, team and enterprise features, and support. The software is genuinely open; the revenue comes from the parts that are deliberately not in the repository.
Pirate Metrics (AARRR)
Pirate metrics, or AARRR, is a five-stage model of a startup's funnel: acquisition, activation, retention, referral and revenue. Introduced by investor Dave McClure in a 2007 talk called Startup Metrics for Pirates, its value is not the list of stages but the discipline it imposes — one number per stage, so you can see which stage is leaking instead of assuming the answer is more traffic.
Product Hunt launch
A Product Hunt launch is the act of listing a new product on Product Hunt, a community site where makers post products and users upvote and comment on launch day. Listings compete for daily, weekly, and monthly rankings, and a strong launch drives a burst of early-adopter traffic, signups, and backlinks.
Product-Led Growth (PLG)
Product-led growth (PLG) is a go-to-market strategy where the product itself is the primary driver of acquisition, conversion, and expansion. Users experience the product's value first — through a free trial, a freemium tier, or an interactive demo — and upgrade themselves, rather than being sold to by a sales team. Instead of marketing and sales pushing people toward the product, the product pulls them in and does much of the selling.
Product-Market Fit (PMF)
Product-market fit (PMF) is the point at which a product satisfies strong, sustained demand in a good market — customers buy it, keep using it, and tell others without being pushed. The term was popularized by Marc Andreessen in a 2007 essay, which defined it as "being in a good market with a product that can satisfy that market," building on a concept Andy Rachleff of Benchmark had framed earlier.
Product-Qualified Lead (PQL)
A product-qualified lead (PQL) is a user who has already experienced real value in your product — usually on a free trial or free plan — and shown behaviour that signals they are ready to buy, such as hitting a usage limit or using a core feature repeatedly. Unlike a marketing- or sales-qualified lead judged on who someone is or what they clicked, a PQL is qualified by what they actually did inside the product.
Referral program
A referral program is a structured growth channel that rewards your existing customers for recommending your product to new people, usually with a shareable link and a reward (account credit, a free month, or cash) paid when the referred person signs up or pays. Because the recommendation comes from a trusted source, referred users convert at higher rates and tend to stick around longer.
Share of Voice (SOV)
Share of voice (SOV) is the percentage of total visibility in your market that your brand owns compared to competitors, measured within a channel — search, social, paid ads, PR, or, increasingly, AI-generated answers. If ten brands compete for a topic and your content earns three of every ten mentions or rankings, your share of voice is 30%.
Usage-based pricing
Usage-based pricing charges a customer for what they consume — API calls, messages sent, generated images, gigabytes stored — instead of for a seat or a flat monthly plan. It is the default shape for products with a real marginal cost per action, because the invoice then moves with the cost of serving that account rather than with its headcount.
Growth metrics
Activation Rate
Activation rate is the percentage of new signups who reach a product's "aha" moment — the first action that delivers real value — within a defined window. It measures whether people who sign up actually start using the product, bridging the gap between acquisition (getting signups) and retention (keeping them). A common formula is activated users ÷ total signups over the same period, where "activated" is your own clearly-defined milestone.
CAC payback period
CAC payback period is how many months of gross profit from a new customer it takes to recover what you spent acquiring them. It answers a timing question that customer acquisition cost alone cannot — not what a customer costs, but how long your cash is committed before that customer starts paying for the next one.
Churn Rate
Churn rate is the share of customers — or of recurring revenue — lost over a given period, usually a month or a year. It is the hole in the bucket: at 5% monthly customer churn only about 54% of a cohort is still there twelve months later, so churn decides whether new signups compound into growth or merely replace what left.
Cohort analysis
Cohort analysis is the practice of grouping customers by something fixed about their arrival — the week they signed up, the channel they came from — and following each group separately over time, instead of reading one blended number for everybody. It answers a question a total cannot: not "is our retention good", but "is it better for the people who joined this month than for the people who joined six months ago".
Conversion lag
Conversion lag is the delay between someone clicking your ad and completing the action you count as a conversion. Because ad platforms credit a conversion back to the date of the click, a recent day's conversion total keeps rising for days after that day has ended — so the newest days in any report are systematically under-reported, which makes cost per acquisition look worse and ROAS look lower than it really was.
Customer Acquisition Cost (CAC)
Customer acquisition cost (CAC) is the total sales and marketing spend required to win one new paying customer, calculated by dividing all acquisition costs over a period by the number of new customers gained in that period. It tells a founder whether growth is actually profitable, and it only makes sense read next to the lifetime value (LTV) each customer returns.
Customer Lifetime Value (LTV)
Customer lifetime value (LTV, also written CLV or CLTV) is the total gross profit a business expects to earn from one customer across the entire relationship — from the first payment to the day they churn. It answers "how much is a customer actually worth?", and paired with customer acquisition cost it tells you whether your growth math works: you have to earn more from a customer than you spent to win them.
Incrementality
Incrementality is the share of a result that would not have happened without the marketing you are measuring. It is a causal question rather than a credit-splitting one: attribution asks which touch deserves the conversion, while incrementality asks whether that conversion needed the touch at all.
Marketing attribution
Marketing attribution is how you decide which marketing touchpoints get credit for a conversion — which ad, post, page, or email counts as the reason somebody signed up or paid. The model you pick does not change what happened; it changes which channel looks like it worked, which is why the same week of data can justify two opposite budget decisions.
Monthly recurring revenue (MRR)
Monthly recurring revenue (MRR) is the predictable subscription revenue a business expects to receive again next month, normalized to a monthly figure — an annual plan counts as one twelfth of its price every month rather than as a lump in the month it was paid. It is a run rate rather than cash collected, which is why one-time fees, setup charges and discretionary usage are excluded: MRR exists to answer what the business earns next month without selling anything new.
Net promoter score (NPS)
Net promoter score (NPS) is a customer-loyalty metric built from a single question — how likely you are to recommend this product to a friend or colleague, answered on a scale of 0 to 10. The percentage of respondents scoring 0 to 6 is subtracted from the percentage scoring 9 to 10, producing one number between −100 and +100. It was introduced by Fred Reichheld in the Harvard Business Review in 2003 and became widespread because it compresses satisfaction into a figure a whole company can be pointed at.
Net Revenue Retention (NRR)
Net revenue retention (NRR) measures how much recurring revenue a fixed group of existing customers produces in this period compared with what the same group produced a period earlier, counting upgrades, downgrades and cancellations, and excluding anyone new. Above 100% means the customers you already have grow revenue on their own; below 100% means every new sale has to refill a leaking bucket before it adds anything.
North Star Metric
A North Star Metric (NSM) is the single measure that best captures the core value your product delivers to customers — the one number a whole team can rally behind because moving it means users are getting more of what they came for. Good examples are nights booked (Airbnb) or messages sent (a chat app), not revenue or signups. It is a leading indicator of sustainable growth, not a vanity count.
Quality Score
Quality Score is Google Ads' 1-10 diagnostic estimate of how relevant your keyword, your ad and your landing page are to someone searching, built from three reported components: expected click-through rate, ad relevance and landing page experience. It is a directional health check on a keyword — not the number used in the live auction, and not a dial you can turn directly.
Return on Ad Spend (ROAS)
Return on ad spend (ROAS) is the revenue a campaign generated divided by what the campaign cost — $4 of revenue from $1 of spend is a 4× ROAS. It measures how efficiently ad money turns into revenue, which is a narrower question than whether the business made any money.
Server-side tracking
Server-side tracking is recording an event on a server you control — your own backend — and sending it on to analytics or ad platforms from there, instead of having JavaScript in the visitor's browser send it. It changes who has to cooperate for the number to exist: a browser is free to decline the request, and your own server cannot decline itself. It does not make data more accurate by itself and it does not settle the consent question — it moves where the write happens, along with what you can still see and what you now have to carry yourself.
Time to Value (TTV)
Time to value (TTV) is how long it takes a new user to reach their first real benefit from your product — the moment it visibly solves the problem they signed up for. A shorter time to value means people experience the payoff before they lose interest, which is why TTV is one of the strongest levers on activation rate and retention.
SEO
Canonical Tag
A canonical tag is a link element in a page's head that names the preferred URL when several URLs serve near-identical content, so search engines consolidate ranking signals onto one address instead of splitting them across variants. It is a hint rather than an instruction: Google weighs it against your redirects, internal links and sitemap, and can still choose a different URL as the canonical one.
Content decay
Content decay is the slow loss of value in a page you already published — traffic sliding as rankings erode and the search result changes shape around you, or the page quietly becoming wrong as the world it described moves on. The second kind is the more expensive one, because a page can keep ranking perfectly while the sentences on it stop being true.
Core Web Vitals
Core Web Vitals are the three page-experience metrics Google measures from real Chrome visitors: Largest Contentful Paint (LCP) for loading, Interaction to Next Paint (INP) for responsiveness, and Cumulative Layout Shift (CLS) for visual stability. A page is assessed at the 75th percentile of real visits — three quarters of sessions must clear the threshold — and the result acts as a small tiebreaker in ranking, not as a substitute for content that deserves to rank.
Crawl budget
Crawl budget is how many URLs a search engine is willing and able to fetch from your site in a given period. Google defines it as the product of two things — a crawl capacity limit, the most it can fetch without degrading your server, and crawl demand, how much it actually wants your pages — and treats each hostname as having a budget of its own.
E-E-A-T (Experience, Expertise, Authoritativeness, Trust)
E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trust — the content-quality signals Google's human Search Quality Raters use to judge how much a page and its creator can be relied on. It is not a direct ranking score you can tune; it is a framework describing the qualities Google's ranking systems try to reward, and increasingly the same qualities that make AI engines willing to cite you.
hreflang
hreflang is an annotation that tells search engines which language and region a version of a page is written for, so a searcher is served the version meant for them. It is a targeting signal, not a ranking boost — it decides which of your translations appears, never whether the page deserves to appear at all.
IndexNow
IndexNow is an open protocol for telling participating search engines that a URL has been added, updated, or deleted, instead of waiting for a crawler to come back and notice on its own. You host a key file at your domain root to prove you control the site, then submit the changed URLs to one participating endpoint, which shares them with every other engine that supports the protocol. It changes when your pages get discovered — not how well they do once they are.
Keyword cannibalization
Keyword cannibalization is when two or more pages on the same site compete for the same search intent, so a search engine has to choose between them — splitting links, clicks and ranking signals across pages instead of concentrating them on one. It is a site-architecture problem rather than a writing problem, and the fix is almost always to differentiate the pages by intent or merge them into one.
Long-Tail Keyword
A long-tail keyword is a longer, more specific search phrase that gets little traffic on its own but carries unmistakable intent — "best CRM for a two-person agency" rather than "CRM". The name comes from the shape of search demand: a few head terms take enormous volume, and millions of specific phrases each take a handful of searches, which together outweigh the head.
AI & building
Content Security Policy (CSP)
A Content Security Policy (CSP) is a set of rules a website sends to the browser, usually as a Content-Security-Policy response header, that says which sources a page may load scripts, styles, images, fonts and frames from, and where it may send data. Anything the policy does not allow is refused by the browser before it runs. The page itself still loads, which is why a CSP mistake rarely looks like an error: it looks like a page that is quietly missing something.
Design system
A design system is the set of shared decisions a product's interfaces are built from — colour, type, spacing, components and the rules for using them — together with the enforcement that makes those decisions hold. The artefacts are the visible half. The enforceable half is what separates a system from a document, because a rule nothing checks is only a suggestion, and a suggestion that silently does nothing is worse than never having written one.
Prompt caching
Prompt caching is a model provider reusing its work on the opening of a prompt it has just processed, so the next request that starts with exactly the same text is billed at a steep discount and answered faster. It rewards one habit above all others: keep the beginning of every request identical, and put whatever changes at the end.
Rate limit
A rate limit is a cap on how many requests a service will accept from one caller in a window of time — a few per second, or a few hundred per minute — enforced by refusing the excess, conventionally with HTTP status 429 Too Many Requests. It is often confused with a quota, which caps total usage over a longer period such as a day or a month. The two fail differently: a rate limit clears on its own within seconds or minutes, while a used-up quota stays exhausted until it resets.
Vibe Coding
Vibe coding is building software by describing what you want in natural language and letting an AI coding tool — such as Cursor, Claude Code, Lovable, Replit, or v0 — write, run, and revise the code, so the human steers by intent and feedback rather than typing most lines by hand. The term was coined by AI researcher Andrej Karpathy in February 2025 to describe 'giving in to the vibes' and largely trusting the model's output.
Vibe Marketing
Vibe marketing is running your marketing the way vibe coders build software: you describe the outcome you want in natural language, AI agents produce the campaigns, content, and outreach, and you steer by reviewing and approving the output instead of executing every task by hand. The term emerged in 2025 as the marketing-side sibling of vibe coding, and it describes a workflow, not a specific tool.
An AI growth team that runs this for you
AgentCeres is a managed AI marketing team — you approve what ships. 14-day free trial, from $39/month.