★ Glossary

Marketing and AI search, in plain English.

120 terms we use across the blog, the checker and client work, defined in a sentence or three. The first sentence is always the definition, so you can stop reading the moment you have what you came for.

A

Ads in AI answers

Sponsored placements inside AI-generated results: labeled links woven into answer text, sponsored follow-up suggestions, and paid entries in recommendation lists. Live in Google’s AI surfaces since 2026; the mechanics and your moves are in our ads analysis.

AI agent

Software that pursues a goal on a person’s behalf: it plans, browses, uses tools and takes actions, instead of just answering a question. Shopping agents that research and buy are the version businesses meet first; our agent-readiness audit covers what breaks them.

AI citation

A link or source credit inside an AI-generated answer, pointing at the page the answer drew from. Citations are the AI era’s version of ranking #1: present in the answer itself, and the slot you can’t buy.

AI contribution pilot

Google’s invitation-only program that pays a small group of publishers monthly when their content significantly feeds AI answers, with earnings shown in Search Console. There’s no application; citations alone don’t qualify, content used during generation does. Details in our breakdown.

AI Max (Google Ads)

Google Ads’ AI-driven targeting and creative layer for Search campaigns, mandatory for Search since late 2026, and one of the pipes through which ads become eligible for AI surfaces. You can’t book AI-answer placements directly; eligibility flows through AI Max and PMax.

AI Mode

Google’s conversational search surface: a full AI answer with follow-up questions, living in its own tab and increasingly fused with regular results. It fans your question into sub-searches and composes one cited response. See AI Mode vs AI Overviews.

AI Overview (AIO)

The AI-written answer box at the top of Google results, assembled from web sources and shown before any blue link. Since August 2026 it can open fully expanded for some queries, pushing organic results below the fold, covered in our analysis.

AI slop

Low-effort, mass-produced AI content published without verification or editing, generic, padded, often subtly wrong. It’s what quality systems and readers are learning to filter, and the opposite of the tested, first-person content that earns citations.

AI visibility

How findable, verifiable and quotable your business is to AI systems: whether assistants can confirm your basic facts and whether answers cite or recommend you. It’s what our free AI Visibility Checker scores.

Agentic commerce

Transactions where an AI agent acts as the buyer: it finds the product, checks specs and stock, fills the forms and completes payment through an authorized rail, while the human approves. Live today at small scale via systems like ChatGPT’s Instant Checkout.

Alt text

The written description attached to an image in HTML, read by screen readers and by machines that can’t or don’t render the picture. Good alt text describes what’s shown, not keyword soup, and it’s one of the places AI systems read your pages.

Anchor text

The clickable words of a link. It tells users, search engines and AI systems what the destination page is about, which is why “pricing for kitchen renovation” beats “click here” every time.

Answer engine

A search system whose output is a composed answer rather than a list of links: ChatGPT Search, Perplexity, Google’s AI surfaces. Optimizing for them is less about ranking and more about being the source the answer gets written from.

Applebot-Extended

Apple’s robots.txt token for opting out of its AI training, separate from regular Applebot, which keeps indexing for Siri and Spotlight. Block the Extended version and Apple’s search features still see you.

B

Backlink

A link from someone else’s site to yours, the classic vote of confidence in rankings. In the AI era its cousin matters too: the unlinked brand mention, which feeds the consensus answer engines check before recommending you.

Bytespider

ByteDance’s (TikTok’s) crawler, infamous for aggressive request bursts that spike server load, with no search-citation upside. It’s the one bot nobody regrets blocking; the rules are in our robots.txt guide.

C

CCBot (Common Crawl)

The crawler of Common Crawl, the open web archive whose datasets feed many AI training pipelines. Blocking it refuses that whole downstream, one line in robots.txt.

Canonical URL

The one address you declare as the “real” version of a page when several URLs show the same content, via the rel=canonical tag. It keeps duplicates from splitting your signals.

ChatGPT Search

OpenAI’s web-connected search inside ChatGPT: it runs live searches and answers with citations, drawing on Bing’s index plus OpenAI’s own crawling. There’s no submission route; the freshness levers are in this guide.

ChatGPT-User

OpenAI’s live-fetch agent: it grabs a specific page in real time when a user’s question needs it. Separate from GPTBot (training) and OAI-SearchBot (search index), and worth leaving unblocked.

ClaudeBot

Anthropic’s web crawler, used for model training and crawling; its sibling Claude-User fetches pages live when a person’s question needs them. Each is controlled separately in robots.txt.

Comparison (versus) page

A page on your own domain answering “[competitor] vs you” or “[competitor] alternatives” with an honest side-by-side. Own these or AI answers ground on an affiliate blog’s version; the playbook is in the competitor fix.

Context window

How much text an AI model can hold in working memory at once, the conversation, documents and instructions combined. Bigger windows mean the model can read more of your material before answering about it.

Crawl budget

The amount of crawling a search engine is willing to spend on your site in a period. Mostly a big-site concern; small sites waste more visibility through blocked or slow pages than through budget.

Crawler (bot)

An automated program that fetches web pages, for search indexing, AI training, or live answering. The 2026 skill is telling them apart: training bots are blockable without consequence, search and user-fetch bots are your visibility.

CTR (click-through rate)

Clicks divided by impressions: of everyone who saw your result, how many clicked. AI answers compress it brutally, Ahrefs measured a 58% CTR cut at position 1 under an AI Overview, which is why impressions-with-citations became the metric to watch.

Citation (local SEO sense)

A mention of your business name, address and phone on directories and other sites, the classic local trust signal. Not to be confused with an AI citation; confusingly, consistent local citations now help earn those too.

Core Web Vitals

Google’s page-experience metrics, loading, interactivity, visual stability, measured on real users. A minor ranking input and a major user-experience one; fix them for people, collect the SEO as a side effect.

Keyword cannibalization

When two of your own pages compete for the same query, splitting signals so neither wins. The fix is usually merging into one stronger page, or sharpening each page’s distinct question.

D

Deep research (AI)

An AI mode that runs a long, multi-step investigation, dozens of searches and sources, and returns a structured report instead of a quick answer. Another surface where being the citable source pays.

Discover (Google)

Google’s scrolling feed of recommended content on mobile, driven by interests rather than queries. It has its own AI features now, and its own Generative AI report in Search Console.

Domain authority

A third-party score (Moz’s DA, Ahrefs’ DR and similar) estimating a site’s link strength. Useful for rough comparison; not a Google metric, and visibly not what decides AI citations, where liftable facts beat famous domains.

E

E-E-A-T

Experience, Expertise, Authoritativeness, Trust: Google’s framework for judging whether content deserves belief, from its quality guidelines. The first E, demonstrated first-hand experience, is the one AI answers visibly reward, and the one most corporate content lacks.

Embeddings

Numerical representations of text that capture meaning, letting machines measure how similar two passages are. They’re why AI search matches your content to questions that never contain your exact keywords.

Entity

A distinct, identifiable thing in a search engine’s understanding, a business, person, product or place, with known attributes and relationships. Being a clear entity (consistent name, facts, structured data, profiles) is what lets machines confidently recommend you.

Exclusion dividend

The opportunity created when a competitor opts out of Google’s AI features: their citations don’t disappear, they get reassigned to the next adequate source. How to spot the vacancies and inherit them: the audit.

Experience signals

Evidence in the content that a real person did the thing: testing notes with dates and durations, original screenshots, exact prices, named failure cases, admitted trade-offs. The reason forums out-cite brands, and the fix in the Reddit-citations playbook.

F

FAQPage schema

Structured data marking up question-and-answer pairs you authored on your own page. Used truthfully (matching the visible content), it helps parsers find your answers; it’s the schema type behind every FAQ section on this site.

Featured snippet

The extracted-text answer box of the pre-AI era, one page’s passage, quoted and linked. AI Overviews are its successor: many sources, synthesized, which changed the craft from winning the box to being quotable inside it.

G

GA4 (Google Analytics 4)

Google’s current analytics platform, measuring users and events across your site. One AI-era quirk to know: clicks from AI Overviews and AI Mode arrive labeled as plain Organic Search, invisible as a separate channel.

GEO (Generative Engine Optimization)

The practice of structuring content and off-site presence so AI engines retrieve, cite and recommend you in generated answers. In working terms: liftable answer-first pages, real experience markers, clean entity data, and third-party consensus.

GPTBot

OpenAI’s model-training crawler. Blocking it in robots.txt keeps your content out of future training sets and has zero effect on Google rankings or on ChatGPT Search, which runs on a different bot (see OAI-SearchBot).

Gemini

Google’s AI model family and assistant, the brain behind AI Overviews, AI Mode, and the assistant replacing Google Assistant on phones and home speakers. When your customer “asks Google” anything in 2026, Gemini is usually who answers.

Generative AI report (Search Console)

Search Console’s report showing how often your links appear inside AI Overviews and AI Mode, worldwide since August 2026, impressions only, with no queries, clicks or CTR. How to read it without fooling yourself: our guide.

Google Ads

Google’s advertising platform: you bid for placements across Search, Shopping, YouTube, Display and, increasingly, inside AI answers. Advertisers currently can’t opt out of, target, or separately measure the AI placements.

Google Business Profile (GBP)

Your business’s listing in Google Search and Maps: hours, photos, reviews, services. In the AI era it’s shifting from a display you control to source material a generated answer is written from, the shift mapped in our local guide.

Google-Extended

A robots.txt token that opts your content out of training Google’s Gemini models, and nothing else: indexing, rankings and AI Overview appearance are untouched, since those run on Googlebot. The lowest-risk block there is.

Googlebot

Google’s main indexing crawler, the one behind Search AND AI Overviews. The bot you never block, and the reason “block the AI” robots.txt mistakes can de-index whole sites.

Grounding

When an AI system bases its answer on retrieved, current sources instead of only its training memory. Being “grounding-grade”, the page an answer can’t be written without, is what earns citations, and what Google’s contribution pilot pays for.

H

Hallucination

When an AI states something false with full confidence, invented facts, prices, features or sources. For businesses the practical defense is leaving no gaps: if your own pages don’t state your facts plainly, the machine fills the silence, sometimes wrongly.

Heading structure (H1-H3)

The hierarchy of headings on a page: one H1 for the topic, H2s for sections, H3s beneath. Machines read it as the page’s outline, which is why question-shaped headings with immediate answers win AI citations.

Helpful content system

Google’s ranking systems rewarding people-first content, made for readers, showing real experience, over content made to game search. The same qualities, by no accident, are what AI answers quote.

Hreflang

Markup telling search engines which language and regional version of a page to show which users. Essential for multilingual sites; mis-set, it shows Hebrew pages to Texans and vice versa.

I

Impressions

How many times your link was shown, in results or inside an AI answer, regardless of clicks. In the AI era impressions can rise while clicks fall; our own Search Console export showed 39,310 impressions to 123 clicks over 87 days.

IndexNow

A free, open protocol that pushes new and updated URLs to Bing (and others) the instant you publish, instead of waiting to be crawled. Google ignores it; Bing acts on it, and Bing feeds ChatGPT Search, making it the fastest indirect freshness lever.

Indexing

A search engine storing your page in its database so it can be returned for queries; crawled but not indexed means seen and shelved. No index, no rankings, no citations, which is why indexing checks come before everything else.

Information gain

What your page adds that existing sources don’t: original data, a tested result, a firsthand detail. Pages that only remix what’s already written contribute nothing to an answer; pages with gain are what generation reaches for.

Instant Checkout

OpenAI’s in-chat purchasing: ChatGPT completes a buy from participating merchants (Etsy, Shopify sellers) without the user visiting the store, over Stripe-built payment rails. The first mainstream proof that agentic commerce is live.

Internal link

A link from one page of your site to another. It routes readers, spreads authority, and shows machines how your topics connect, the cheapest SEO lever most sites underuse.

Search intent

What the searcher actually wants, to learn, compare, buy, or reach a specific place, regardless of the words used. Matching intent beats matching keywords, in classic SEO and doubly in AI answers that interpret the question first.

J

JavaScript SEO

The discipline of keeping content visible to machines when a site renders with JavaScript. Google eventually renders JS; many AI crawlers and agent pipelines read only the raw HTML response, so prices and facts that exist only after scripts run read as nothing.

K

Keyword

The word or phrase people type (or say, or prompt) when searching for something your business answers. Still the unit of research; in the AI era, the full questions and prompts around a keyword matter as much as the keyword itself.

Keyword difficulty

A tool-estimated score of how hard a term is to rank for, usually based on who currently ranks. A rough compass, not a verdict; tools disagree with each other and with reality.

Knowledge Graph

Google’s internal database of entities and the relationships between them, what powers knowledge panels and feeds facts into AI answers. Your structured data, profiles and consistent facts are how your business earns a reliable node in it.

Knowledge cutoff

The date after which an AI model’s built-in knowledge stops; anything newer requires live search or retrieval. It’s why a chat answer without web search can be confidently months out of date, and why forcing search mode matters.

L

Lastmod

The timestamp in your XML sitemap saying when each page last changed, which crawlers use to decide what’s worth re-reading. Keep it truthful: faking freshness is the oldest trick there is, and sitemaps that lie get trusted less.

Link building

Earning links from other sites to yours, through content worth citing, digital PR, partnerships. The modern version includes earning unlinked mentions and roundup placements, because AI consensus-checking counts those too.

LLM (large language model)

The AI model type behind ChatGPT, Gemini, Claude and the rest: trained on vast text to predict and compose language. Everything on this page exists because LLMs became the layer between your customers and your website.

llms.txt

A proposed file at yourdomain.com/llms.txt indexing your key content and policies for AI systems. Honest status: no major vendor documents actually reading it, so treat it as a cheap maybe, never a substitute for schema, static prices and crawler access.

Local pack

The map-plus-three-listings block Google shows for local searches. It still rules simple “pizza near me” queries, while question-shaped local searches increasingly get AI Overviews instead, the split that reshapes local SEO.

Local SEO

Visibility work for businesses serving a geographic area: Business Profile, reviews, local pages, citations. In 2026 it extends to being quotable in the AI answers that now cover 68% of local searches.

Log file analysis

Reading your server’s raw access logs to see exactly which bots and people fetched what, when, and with what result. The ground truth for crawler questions; our crawler guide shows the grep-level workflow.

Long-tail keyword

A longer, more specific query (“emergency plumber open Saturday Rishon”) with lower volume and clearer intent than head terms. AI search amplifies the long tail: fan-out and conversational prompts are long-tail by nature.

M

Mention (brand mention)

Your business named on another site without a link. Classic SEO shrugged at these; AI consensus-checking reads them, which quietly made unlinked mentions a visibility asset again.

Meta description

The page summary shown (sometimes) under your title in results. Not a ranking factor; a click factor, and one more plainly stated summary machines can read, so write it like the honest pitch it is.

Model training

Teaching an AI model from large datasets, including crawled web content, as opposed to live retrieval at answer time. Training and answering are separate pipelines with separate bots, the distinction every robots.txt decision rests on.

N

NAP

Name, Address, Phone, the trio that must match everywhere your business appears online. Inconsistent NAP confuses the Knowledge Graph and gives AI answers contradictions to trip on.

Negative keyword (PPC)

A term you tell Google Ads NOT to show your ads for, blocking irrelevant searches from spending your budget. The highest-ROI ten minutes in most ad accounts we audit.

Nofollow

A link attribute (rel=”nofollow”, plus “sponsored” and “ugc” variants) telling search engines not to pass ranking credit through the link. The link still works for humans, and still counts as a mention.

Noindex

A meta tag or header telling search engines to keep a page out of results entirely. Powerful and dangerous: a noindex left on by accident is the classic “why did our traffic die” finding.

Nosnippet

Markup limiting or forbidding search engines from quoting your page’s text in results and AI features. A blunt tool: it protects content from being excerpted by also removing you from the surfaces excerpts win.

O

OAI-SearchBot

OpenAI’s search-indexing crawler: the one that must be allowed for your site to appear in ChatGPT Search answers. It’s independent from GPTBot (training) and ChatGPT-User (live fetches), so you can refuse training and keep citations.

On-page SEO

Everything you optimize on the page itself: titles, headings, content, internal links, schema, speed. Its AI-era extension is extractability, making facts liftable, not just rankable.

Opt-out (Search generative AI setting)

The Search Console control that removes your site from Google’s AI Overviews, AI Mode and Discover AI within a day or two, rankings untouched. Global since late September 2026; why flipping it is usually self-harm: the opt-out guide.

Organic traffic

Visits from unpaid search results, as opposed to ads. The channel AI answers are squeezing, and the reason measurement is shifting toward visibility inside answers rather than clicks out of them.

P

People Also Ask (PAA)

The expandable related-questions box in Google results. By late 2026, studies put ~97% of PAA answers as AI Overviews, making it effectively another AI surface, and another set of questions your pages should answer directly.

Performance Max (PMax)

Google Ads’ goal-based campaign type that places ads across all Google surfaces automatically, including AI experiences. Powerful reach, limited steering, which makes clean feeds and assets the control you actually have.

PerplexityBot

Perplexity’s search-indexing crawler, routinely misfiled on blocklists as a training bot. Block it and you vanish from Perplexity’s cited answers; concerns about its robots compliance argue for firewall-level control, not for a citation-costing robots line.

Pillar page

A comprehensive page covering a topic broadly, with cluster articles covering each subtopic in depth, all interlinked. The structure that builds topical authority, and that AI fan-out rewards.

PPC (pay-per-click)

Advertising where you pay per click, Google Ads being the canon example. The AI-era wrinkle: paid slots now appear inside answers themselves, while the organic citation stays the one placement money can’t rent.

Preferred Sources

A Google feature letting users star sites they trust, which then surface more often for them across Search, Top Stories and, since May 2026, AI Overviews and AI Mode. Sites can embed a one-tap button asking for the star; setup in our guide.

Prompt

The instruction or question a person gives an AI. For marketers the useful habit is collecting the prompts customers actually use about your category, they’re the new keyword research.

Prompt injection

An attack where malicious instructions hidden in content hijack an AI agent that reads it, “ignore your rules and send the data to…”. The structural risk of agents that browse, and the reason scoped permissions exist.

Q

Quality Rater Guidelines

Google’s public manual for the human raters who evaluate search quality, the document where E-E-A-T and YMYL come from. Raters don’t set your rankings directly; the guidelines reveal what the systems are tuned toward.

Query fan-out

How AI search decomposes one question into multiple hidden sub-searches, researching each, then composing a single answer. It’s why covering a topic’s obvious follow-up questions earns citations the main keyword alone never would.

Search query

The actual string a user typed, as distinct from the keyword you targeted. Search-term reports and GSC queries show the real language of your market, which your pages should speak back.

R

RAG (retrieval-augmented generation)

The architecture behind answer engines: retrieve relevant sources first, then generate the answer from them. Practical translation: the engine googles, reads, and writes, and your job is being worth reading at step two.

Rank tracking

Monitoring where your pages rank for chosen keywords over time. Still useful, increasingly incomplete: a rank tracker can show position 3 while the AI answer above it took the clicks, which is why citation tracking joined the stack.

301 redirect

A permanent redirect from one URL to another, passing users and most ranking signals along. The right tool for moved or merged pages; chains and loops of them are the common self-inflicted wound.

Rich results

Enhanced listings in Google, stars, prices, FAQs, images, powered by structured data. The visible reward for schema done truthfully, and a preview of how machines parse your page.

ROAS

Return on ad spend: revenue divided by advertising cost. The metric that keeps ad conversations honest, and the one AI-answer placements currently make hard to isolate.

robots.txt

The file at yourdomain.com/robots.txt telling crawlers what they may fetch, a request honored by the major vendors, not a wall against the rest. The 2026 version blocks training bots by name while keeping search bots in; exact rules in the copy-paste guide.

S

Schema / structured data (JSON-LD)

Machine-readable markup describing what a page IS, a product, a business, an FAQ, with explicit facts like price and availability. It’s how agents and answer engines verify you without guessing; used truthfully, it clarifies, it never reclassifies.

Semantic search

Search that matches meaning rather than exact words, powered by embeddings. It’s why synonyms rank, why keyword stuffing died, and why writing naturally about a topic beats repeating a phrase.

SERP

Search engine results page, the page you land on after a search. Increasingly a misnomer: for many queries the “results page” is now one generated answer with results in the basement.

SGE (Search Generative Experience)

The 2023 experimental name for what became AI Overviews. You’ll still meet it in older articles and stale advice; if a guide says SGE, check its date before trusting its mechanics.

Sitemap (XML)

A machine-readable file listing your site’s URLs (with lastmod dates) so crawlers can find everything efficiently. Submitted once in Search Console, maintained automatically by your SEO plugin, checked whenever indexing misbehaves.

SynthID / content credentials (C2PA)

Watermarking and provenance standards that mark AI-generated media invisibly (SynthID) or with signed metadata (C2PA), so tools can verify origin. The honest rule of detection: a positive proves AI, a negative proves nothing; our photo-checking guide shows the ladder.

T

Technical SEO

The infrastructure layer of search visibility: crawlability, indexing, speed, rendering, redirects, sitemaps, structured data. In 2026 it absorbed a new client: AI crawlers and agents, with their own access and rendering quirks.

Title tag

The page title in the HTML head, shown as the headline in search results and browser tabs. The strongest few words of on-page SEO; lead with what people search, promise what the page delivers.

Token

The unit AI models read and write in, roughly word-fragments (about 3-4 characters of English on average). Pricing, limits and context windows are all counted in tokens, which is why “how long is it” became “how many tokens is it.”

Topical authority

Being the site that has covered a topic deeply and consistently enough that machines treat you as a go-to source for it. Built with pillar-and-cluster coverage, earned over months, and visible in how often answers reach for you.

Training data

The text, images and code an AI model learned from. Your published content is probably in several models’ training data already; the robots.txt tokens (GPTBot, Google-Extended, Applebot-Extended) govern only future collection.

U

udm=14 / udm=50

URL parameters that jump Google straight into a specific results mode: udm=14 is the classic web-links-only view, udm=50 opens AI Mode directly. Handy as browser shortcuts, and proof the old Google still exists one parameter away.

UGC (user-generated content)

Content made by users rather than the site: reviews, forum posts, comments, Q&A. The content class AI answers visibly over-quote, for its first-person honesty, and the style brands are relearning from it.

User agent

The identity string a browser or bot sends with each request (“Googlebot/2.1”, “OAI-SearchBot/1.4”). What robots.txt rules and firewall tiers key on, and trivially spoofable, which is why IP verification exists.

V

Verified bots

Crawlers whose identity can be cryptographically or IP-verified against the vendor’s published ranges, as opposed to anything merely claiming a famous user agent. The basis of modern firewall tiering: verified agents in, impostors out.

W

WAF (web application firewall)

The security layer (Cloudflare and friends) filtering traffic before it reaches your site, including bots. The 2026 WAF skill is tiering: block anonymous scrapers while letting verified search bots and buying agents through, because some automated traffic now carries a credit card.

Y

YMYL (Your Money or Your Life)

Google’s category for topics that can affect health, finances or safety, where quality standards and E-E-A-T scrutiny run highest. If you publish in a YMYL niche, credentials, sourcing and honesty aren’t optional garnish.

Z

Zero-click search

A search that ends on the results page, the user got their answer without visiting any site. US searches hit roughly 68% zero-click in early 2026, which is why visibility inside the answer, not just beneath it, became the job.

Terms defined. Now, how do the machines define you?

The free checker shows what AI can verify about your business, in about two minutes.