glossary

Plain-language definitions for SEO and AI-search terminology.

A

AEO (Answer Engine Optimisation)↗: The practice of structuring content to be retrieved and cited by AI answer engines such as ChatGPT Search, Perplexity, and Google's AI Overviews. Often used interchangeably with GEO; the emphasis is on earning inclusion in synthesised answers rather than ranked positions.
Agent-readiness↗: The infrastructure layer that determines whether AI agents can discover, read, and act on a website. Sits below content quality: a well-structured page an agent cannot access is no more useful than a well-ranked page a crawler cannot reach.
Agentic commerce↗: AI agents completing a purchase on a user's behalf end to end, rather than surfacing a product for the user to buy themselves. The protocols exist (ACP, UCP, Stripe's Machine Payments Protocol); the transaction volume, for most merchants, does not yet.
Agentic Search↗: AI agents performing multi-step research across multiple sources to answer complex queries. Distinct from single-query retrieval; agents can break questions into sub-queries, browse multiple pages, and synthesise information autonomously.
Agentic SEO↗: Using AI agents to execute SEO workflows autonomously: keyword research, content auditing, technical fixes, and implementation, with minimal human input at each step. Distinct from agentic search (optimising content for retrieval by AI research agents).
AI agent↗: Software that pursues a goal autonomously across multi-step tasks, using tools, memory, and reasoning to plan and act without human approval at each step. Distinct from AI assistants (which respond to requests per session) and chatbots (which follow pre-defined rules).
AI content detector↗: A third-party tool that guesses whether text was machine-written by reading its style. Unreliable for any individual page, and prone to flagging human writing as AI. Distinct from a watermark, which the generating system inserts deliberately.
AI crawler↗: An automated user-agent operated by an AI company. Three distinct purposes: training (GPTBot, ClaudeBot, CCBot), search indexing (OAI-SearchBot, Claude-SearchBot, PerplexityBot), and live fetches made on a user's behalf (ChatGPT-User, Perplexity-User). Blocking a training crawler has no effect on that platform's search citations.
AI Mode↗: Google's conversational search interface, reached via a tab in Search, that replaces the standard SERP with a Gemini-powered chat experience. Unlike AI Overviews, which appear within normal results, AI Mode is a full alternative interface handling complex, multi-part queries through follow-up conversation.
AI Overview↗: A generative AI answer shown at the top of Google's search results, synthesised from multiple sources with citations. Often produces zero-click outcomes for informational queries.
AI share of voice↗: The proportion of brand mentions across AI-generated answers within a category that belong to one brand. A separate metric from SEO share of voice, measured over a prompt set rather than a keyword set, and calculated differently by every tool.
AI shopping agent↗: An assistant that discovers and recommends products inside a conversation, such as Google AI Mode, ChatGPT, Copilot or Perplexity. The signals they weight mirror ecommerce SEO: complete descriptions, accurate structured data, reviews, and return and shipping policies.
AI Traffic: Visits to a site from AI answer engines such as ChatGPT Search, Perplexity, Copilot, and Claude. Growing traffic source as these systems complement or replace traditional search.
AI Visibility↗: How frequently a site is cited or retrieved by AI answer engines across multiple queries. A complement to search rankings; two sites can have similar Google visibility with very different AI citation rates.
AI watermarking↗: A machine-readable signal embedded into AI output at the point it is generated, so its origin can later be established. Imperceptible to a reader and, in current designs, does not change what the content says. No platform has said watermarks affect ranking or AI citation.
Algorithm update↗: A change to how a search engine ranks results. Google deploys broad core updates several times per year; smaller updates are continuous.
Alt text↗: The text alternative for an image, used by screen readers and search engines. Should describe the image accurately and concisely.
Anchor text↗: The clickable text of a hyperlink. Used by search engines as a relevance signal for the linked page; over-optimised exact-match anchors can trigger penalties.
Anchor Text Profile↗: The distribution of anchor text across all backlinks pointing to a page. Over-optimised profiles (mostly exact-match keywords) can trigger penalties; natural profiles are diverse and mix branded, generic, and partial-match anchors.
Anonymised queries↗: Search queries Google omits from the Search Console Performance report to protect user privacy: those searched by only a few dozen users over a two to three month period, plus any containing personal or sensitive information regardless of volume. They count toward chart totals but get no row in the query table or API, so query rows never sum to the total, and applying a query filter drops them entirely.
AS (Authority Score)↗: Semrush's proprietary score (0-100) estimating a domain's overall SEO authority, combining link, traffic, and quality signals.
Author Box: A bio module displaying the author's name, credentials, and expertise at the top or bottom of an article. Strengthens E-E-A-T signals and is recommended by Google for YMYL content.

B

Backlink↗: A hyperlink from another website pointing to yours. One of the strongest ranking signals; quality matters more than quantity.
Bing Webmaster Tools↗: Microsoft's counterpart to Search Console: site verification, sitemap submission, URL inspection, crawl diagnostics, and Bing search performance data. Since February 2026 it also carries an AI Performance report showing citation data across Microsoft Copilot and ChatGPT Search.
Black hat SEO↗: Tactics that violate search engine guidelines: cloaking, sneaky redirects, automated content generation, link schemes. Carries penalty risk.
Bounce rate↗: The percentage of visits where the user leaves without further interaction. A weak SEO signal in isolation; context dependent. Replaced by engagement rate as GA4's primary engagement metric.
Brand mentions↗: Online references to a brand name, product, or person, whether linked or unlinked. Unlinked mentions can signal brand authority to search engines; linked mentions also pass link equity. Also called brand signals.
Branded search↗: Queries that include a company or product name. Reflects demand from users who already know the brand, so it is reported separately from non-branded search, which reflects the demand SEO earned. Branded traffic typically converts several times higher.
Breadcrumbs↗: A navigation aid showing the user's location within a site hierarchy. Often marked up with BreadcrumbList schema for SERP display.
Broken Link Building↗: A link-building tactic that finds broken links on relevant sites and offers replacement content. Lower barrier than pitching from scratch since the site owner already has a reason to link.

C

C2PA (Coalition for Content Provenance and Authenticity)↗: An open standard for recording who created a file and how it has been edited since, as cryptographically signed metadata attached to the file. Sold as Content Credentials. Complements watermarking: the metadata carries a fuller history but can be stripped, where a watermark travels with the content itself.
Canonical tag↗: An HTML element (`<link rel="canonical">`) declaring the preferred URL when multiple URLs serve substantially the same content.
cats.txt↗: A satirical web standard published by Mark Williams-Cook in September 2025 for declaring a domain's office cats: name, role, breed, and a PurrLevel scored out of ten. Written to test the evidence bar for llms.txt, and it cleared every bar that gets offered as proof. AI crawlers fetched it, Google indexed it, and AI Overviews repeated its fictional contents as fact.
ccTLD (Country-Code Top-Level Domain)↗: A domain extension assigned to a specific country: .co.uk for the UK, .de for Germany, .fr for France. A strong geotargeting signal, but limits flexibility if a business expands internationally.
Circle to Search↗: An Android feature that allows users to initiate a Google search from within any app by circling, highlighting, scribbling on, or tapping content on screen: text, images, or video frames. The search runs without leaving the current app and uses Google's standard ranking systems.
Citation (local)↗: An online mention of a business including its name, address, and phone number (NAP). Important for local pack ranking.
Cloud stacking↗: Hosting thin pages or documents on high-authority cloud platforms (AWS, Azure, Google Cloud) to manufacture backlinks that borrow the host domain's authority. A link scheme: the host's trust does not transfer to pages nobody references or visits. The Google-properties variant is entity stacking.
CLS (Cumulative Layout Shift)↗: A Core Web Vitals metric measuring unexpected layout shifts during page load. Good: under 0.1.
Cluster (content cluster)↗: A group of pages covering related sub-topics, linked back to a central pillar page. Builds topical authority.
Commodity content: Content on high-volume, low-differentiation topics where many sites publish similar information with minimal unique perspective or data. Difficult to rank because competition is intense and relevance signals are commoditised; requires exceptional depth, original research, or unique angle to break through.
Competitor gap analysis↗: Identifying queries where one or more competitors rank in the top 20 but a site does not appear at all. Distinct from content gap analysis in where it starts: competitor rankings rather than topic coverage. Gaps where two or more competitors rank are the stronger signal.
Consent Mode↗: Google's signalling framework for adjusting tag behaviour according to a user's cookie choices. When consent is denied, tags send cookieless pings Google uses for modelled conversions rather than dropping the event entirely.
Content audit↗: A systematic review of a site's existing pages, assessing each for performance, accuracy, and strategic value. Informs decisions on whether to keep, update, merge, or remove pages.
Content brief↗: A document defining a piece's target query, reader intent, structural requirements, and scope before writing begins. Reduces revision cycles and keeps output consistent across writers. The common failure is specifying word count and keyword density while saying nothing about what the reader needs.
Content Credentials↗: See C2PA (Coalition for Content Provenance and Authenticity).
Content Decay↗: The tendency for page traffic to decline over time as competition increases, information becomes outdated, or search intent shifts. Addressed through content refreshes or updates.
Content depth↗: The degree to which a page answers the full set of questions a searcher might have about a topic. Google does not reward word count; it rewards coverage of the sub-questions that distinguish a useful resource from a shallow one.
Content gap analysis↗: The process of identifying topics or keywords that competitors rank for which a site does not cover. Used to prioritise new content that can capture existing search demand.
Content pruning↗: The practice of removing or consolidating low-value, thin, or underperforming pages to improve a site's overall quality signals and concentrate crawl budget on content that earns traffic.
Content refresh↗: Updating an existing page to maintain or improve its ranking performance. Usually faster and more effective than writing a new page for the same query, because the existing page already carries index history and accumulated ranking signals.
Context Window: The maximum length of text (measured in tokens) that a language model can process in a single request. Larger context windows allow LLMs to consider more information when generating responses.
Core update↗: A broad change to Google's ranking systems, released roughly three to four times a year and announced on the Search Status Dashboard. Not a penalty: a drop means competing pages now look more relevant or trustworthy, not that a rule was broken. Recovery usually requires the next update cycle.
Core Web Vitals↗: Google's set of three user experience metrics used as ranking signals: LCP (load speed), INP (interactivity), and CLS (visual stability). Introduced as a ranking factor in 2021.
Crawl budget↗: The number of URLs a search engine crawler will fetch on a site within a given period. A constraint primarily for very large sites.
Crawl error↗: A failure that stops a search engine fetching a URL it tried to reach: server errors (5xx), redirect problems, soft 404s, and blocked requests. Crawl problems surface in Search Console's Page indexing report, though DNS and connectivity failures appear through URL Inspection and Crawl Stats instead. Only 5xx and 429 responses actually slow Googlebot down.
Crawlability↗: The degree to which search engine bots can access and fetch the pages of a site.
Crawler directive↗: An instruction telling a search engine bot how it may crawl or index a site. Delivered three ways: robots.txt rules govern crawl access, the meta robots tag governs indexing in the HTML, and the X-Robots-Tag header governs indexing for non-HTML files. A robots.txt Disallow controls access, not indexing.
Crawling↗: A search engine bot fetching pages from the web by following links and reading sitemaps. A separate step from indexing, which is an editorial decision made afterwards: a page can be crawled daily and never indexed.
CSR (Client-Side Rendering)↗: Rendering HTML in the user's browser after JavaScript executes. Can delay content visibility to crawlers unless the site prerenders or hydrates correctly. Contrasts with server-side rendering.
CTR (Click-Through Rate)↗: The percentage of impressions that result in a click. A primary measure of SERP performance. Not a direct Google ranking factor, and there is no universal benchmark to aim at, because Google aggregates click data per query and document rather than comparing a page's rate against an average.

D

DA (Domain Authority)↗: Moz's proprietary score (0-100) estimating a domain's link-based authority. Not used by Google directly.
Dark traffic↗: Real human traffic whose source analytics cannot identify, so it defaults to (direct)/(none). Sources include untagged links, HTTPS-to-HTTP referral stripping, in-app browsers, and private social shares. Some ChatGPT and Perplexity visits arrive with no referrer at all.
Deindexing↗: The removal of pages Google had previously indexed, distinct from a page that never got indexed. Common causes are a manual action, an algorithmic quality reassessment during a core or spam update, content decay, or a technical regression such as an accidental site-wide noindex. The fix depends on which one applies.
Digital PR↗: An off-page tactic that earns editorial links from news sites, magazines, and authoritative publishers by creating content genuinely worth covering. The link is a by-product of the coverage rather than the direct goal.
Disavow file↗: A file uploaded via Search Console listing backlinks Google should ignore. Useful for manual action recovery; widely overused otherwise.
Dofollow↗: Practitioner shorthand for a standard link carrying no `rel` restriction. It is not a real attribute and is never written in markup. The default state of any link.
DR (Domain Rating)↗: Ahrefs' proprietary score (0-100) for a domain's link profile strength. Logarithmic; not used by Google directly.
Duplicate content↗: Substantially identical or very similar content appearing at multiple URLs. Dilutes ranking signals across pages; resolved with canonical tags, redirects, or consolidation.

E

E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness)↗: Google's quality framework, applied through its search quality rater guidelines and particularly stringent for YMYL content.
Editorial Calendar: A publishing schedule that tracks content assignments, deadlines, and publication dates. Essential for maintaining consistent topical coverage and distributing work across a team.
Editorial Link: A backlink earned without solicitation, placed by another site owner or editor because they found the content valuable and relevant. The highest-quality link type because it indicates genuine endorsement.
Embedding: A numerical vector representation of content (word, passage, or document) that captures its semantic meaning. Used by language models and search systems to calculate similarity and relevance.
Engagement rate↗: A Google Analytics 4 metric: the percentage of sessions that were engaged sessions. A session is engaged if it lasted longer than 10 seconds, included a conversion event, or had two or more pageviews. Replaced bounce rate as the primary engagement metric in GA4.
Entity SEO↗: Optimising for search engines and AI systems that reason about real-world entities (people, organisations, products) rather than keywords.
Entity stacking↗: A link-building tactic that interlinks Google-owned properties (Docs, Sheets, Sites, custom Maps) and points them at a target site, claiming to build entity authority. Marketed as white-hat because it uses Google's platforms, but it is a link scheme: the links carry no editorial endorsement and are devalued or ignored. Also called Google entity stacking or authority stacking. The platform-agnostic version is cloud stacking.
Expired domain abuse↗: Buying a lapsed domain name and repurposing it primarily to exploit the ranking signals (backlinks, age, residual trust) it built up under its previous owner. A named Google spam policy since March 2024; the tell is a mismatch between the domain's history and the unrelated content now hosted on it.

F

Faceted navigation↗: A filtering system common on e-commerce and catalogue sites that lets users narrow results by attribute: size, colour, brand, price. Can generate large numbers of thin, near-duplicate URL combinations that dilute crawl budget if not controlled with noindex or canonical tags.
Featured snippet↗: A search result format where Google extracts and displays a passage from a ranking page directly in the SERP, above the standard organic results.
FID (First Input Delay): A former Core Web Vital measuring responsiveness to the first user interaction. Replaced by INP in March 2024.
First-Hand Experience: Personal, direct involvement with a topic: using a product, living through an event, practising a skill. Signals genuine authority and is now explicitly part of Google's E-E-A-T framework (the first E).

G

GA4 (Google Analytics 4)↗: Google's analytics platform and the on-site layer of SEO measurement: what organic visitors do after they arrive. GA4 organic sessions and Search Console clicks measure different things and rarely agree; the size of the gap is set by your own consent rate and audience.
GBP (Google Business Profile)↗: Google's free local business listing, the primary asset for local pack and Google Maps visibility. Formerly Google My Business.
GEO (Generative Engine Optimisation)↗: The practice of structuring content for retrieval and citation by AI answer engines. Overlaps with traditional SEO; emphasises entity clarity and direct-answer formats.
Geotargeting↗: Telling search engines which country or region a page is meant for. Google reads the target locale from four signal groups: ccTLDs, hreflang, server location, and market signals such as local addresses, currency, and local links. The Search Console International Targeting setting was removed in 2022.
Google Discover↗: A personalised content feed in the Google app and mobile Chrome that surfaces content by user interest rather than by query. There are no keywords to rank for and no submission process; content surfaces through engagement signals, editorial quality, and large images.
Google Lens↗: Google's visual search product. Searches by pointing a camera at a subject, uploading an image, or selecting content on screen. Can identify objects, read text in images, surface similar products, and return local business information. Integrated into the Google app, Google Images, and AI Mode.
Google Sandbox↗: The widely believed idea that Google withholds rankings from new sites for a fixed probationary period. Google has consistently denied a formal sandbox, though the 2024 Content Warehouse leak describes a hostAge field used to sandbox fresh spam at serving time, which is a narrower mechanism than the myth claims.
Google-Extended↗: Google's user agent for AI training data collection, separate from Googlebot. Can be blocked in robots.txt without affecting search indexing.
Googlebot↗: Google's web crawler. Operates in two-pass indexing: initial HTML processing, then JavaScript rendering.
Grounding↗: Constraining a language model's answer to source documents retrieved at query time rather than to what the model memorised in training. Grounded answers cite the pages they drew on; ungrounded ones are where hallucination appears. Also known as RAG.
GSC (Google Search Console)↗: Google's free webmaster tool providing data on indexing status, search performance, crawl errors, and manual actions. The primary diagnostic interface between site owners and Google.
GTM (Google Tag Manager)↗: A deployment layer for tags and scripts, not a measurement tool in itself. Useful in SEO for recovering data that consent rejection removes, capturing engagement signals GA4 misses, and moving measurement to a first-party path. Crawler directives belong in the HTML, not in a container.
Guest posting↗: Writing an article for another site in your field, usually with a link back to your own. Legitimate when the primary purpose is the reader; it crosses into a link scheme when the link is the point, as with paid placements passing PageRank.

H

Hallucination: A statement made by a language model that is not supported by the sources it was supposed to ground in. A reliability risk for AI answer engines; grounded answers link back to sources to prevent it.
Head Term: A broad, high-volume keyword with generic intent: 'shoes', 'laptop', 'seo'. Opposite of long-tail; more competitive but often targets larger search volumes.
Heading tags↗: HTML elements (H1 through H6) marking the structural hierarchy of page content. The H1 signals the primary topic; lower levels organise sub-sections. A modest on-page ranking signal.
Hreflang↗: An HTML attribute (or sitemap element) declaring the language and regional targeting of a page. Manages multilingual and multi-regional sites.
HSTS (HTTP Strict Transport Security)↗: An HTTP header that forces browsers to use HTTPS for the domain. Strong commitment; cannot be quickly reversed.
HTTP status code↗: The three-digit code a server returns with every response, telling crawlers how to treat the URL: 200 OK, 301 permanent redirect, 302 temporary redirect, 404 not found, 410 gone, 5xx server error.

I

Impressions↗: The number of times a URL appeared in search results for a given query, regardless of whether it received a click. Reported in Google Search Console; used alongside CTR and average position to assess SERP visibility.
Index bloat↗: An excess of low-value, thin, or duplicate pages in a search engine's index. Wastes crawl budget and dilutes quality signals. Addressed through noindex directives, canonicalisation, and content consolidation.
Indexing↗: A search engine storing and organising a crawled page so it becomes eligible to appear in results. A separate decision from crawling, and an editorial one: thin content, near-duplicates, and pages adding no distinct value are routinely excluded.
IndexNow↗: A push-based protocol for notifying participating search engines the moment a page is added, updated, or deleted, instead of waiting for a crawler to rediscover it. Supported by Bing, Yandex, Naver, Seznam, Yep, the Internet Archive, and Amazon's crawler. Google has never adopted it.
INP (Interaction to Next Paint)↗: A Core Web Vitals metric measuring overall page responsiveness to user interactions. Good: under 200ms.
Internal link↗: A hyperlink from one page on a site to another page on the same site. Distributes PageRank and signals topical structure.
International SEO↗: Optimising a site for audiences in more than one country or language: URL structure, hreflang, geotargeting signals, and market-specific keyword research.

J

JavaScript SEO↗: The discipline of ensuring search engines can crawl, render, and index content delivered via JavaScript.
JSON-LD (JavaScript Object Notation for Linked Data)↗: The recommended format for embedding structured data in a webpage as an inline script block, preferred by Google over Microdata and RDFa.

K

KD (Keyword Difficulty)↗: A score (0-100) estimating how hard it is to rank for a query, typically based on the link profiles of currently ranking pages.
Keyword cannibalisation↗: When two or more pages on the same site target the same keyword or query, causing them to compete against each other in search results. Splits ranking signals and can suppress both pages. Resolved through consolidation, canonical tags, or differentiated targeting.
Keyword mapping↗: The process of assigning specific target keywords to individual pages across a site. Prevents cannibalisation and ensures each page has a clear primary search purpose.
Keyword research↗: Finding the queries an audience actually searches, then prioritising them by volume, difficulty, and intent. Feeds keyword mapping, content planning, and the structure of a topic cluster.
Keyword search volume↗: See MSV (Monthly Search Volume).
Keyword stuffing↗: Repetitive, unnatural use of target keywords in content. Damages readability; flagged by quality systems.
Keyword universe↗: The complete set of search queries relevant to a business, assembled from multiple discovery sources before any filtering or prioritisation. Built from seed terms, keyword tools, autocomplete, People Also Ask, GSC queries, competitor rankings, and customer language. Treated as a persistent asset that feeds downstream keyword research activities.
Knowledge graph↗: A database of entities and their relationships used by Google (and AI systems) to answer entity-based queries.
Knowledge panel↗: An information box about an entity (person, organisation, place) shown alongside search results, drawn from Google's knowledge graph.

L

LCP (Largest Contentful Paint)↗: A Core Web Vitals metric measuring perceived load speed via the largest visible element. Good: under 2.5 seconds.
Liam Hayward↗: The creator of this handbook: a UK-based SEO specialist working across traditional and AI search, with over seven years of in-house and agency experience. Also a small demonstration of entity SEO: a personal name is ambiguous by default, and search engines only work out which one you mean once enough of the open web agrees.
Link attributes↗: The `rel` values that tell Google how to treat a link: `sponsored` for paid and affiliate links, `ugc` for user-generated content, and `nofollow` where neither applies. Google treats all three as hints rather than directives.
Link equity (PageRank)↗: The authority a page passes through its outbound links. The basis of Google's original ranking algorithm; still influential.
Link reclamation↗: Recovering links you have already earned: inbound links now hitting a 404 (fixed with a redirect, no outreach needed), your own broken links, and unlinked brand mentions. Distinct from broken link building, which chases dead links on other people's sites. Lower effort than new link building because the endorsement already exists.
Link velocity: The rate at which a site acquires new backlinks over time. Sudden spikes can appear unnatural and attract algorithmic or manual scrutiny.
LLM (Large Language Model): A neural network trained on large amounts of text data to predict and generate text. Powers AI answer engines like ChatGPT, Claude, and Perplexity. Operates on learned patterns rather than external knowledge retrieval alone.
LLMO (Large Language Model Optimisation)↗: A term for the practice of optimising content to be cited by large language models. Emphasises influencing what models know rather than what they retrieve in real time; in practice the content tactics are nearly identical to GEO.
llms.txt↗: A proposed Markdown file at the root of a domain providing an AI-readable index of a site's most important content. Not yet a formal standard.
Local landing page↗: A page serving location-specific queries for a business operating across several areas. Pages that rank carry genuinely location-specific content: local team information, area-specific testimonials, and services relevant to that market. Templated pages with only the city name swapped are routinely outranked.
Local pack↗: The map and three business listings shown at the top of search results for queries with local intent.
Local SEO↗: Optimising for queries with local intent: Google Business Profile, local pack visibility, citations and NAP consistency, reviews, and location-specific pages.
LocalBusiness schema↗: Schema.org structured data declaring a business's local entity properties: name, address, phone, hours, services.
Log file analysis↗: The process of parsing server access logs to understand exactly which URLs search engine bots crawled, at what frequency, and which were skipped. More reliable than crawl simulation tools.
Long-tail keyword↗: A specific, often longer search query with low individual volume but typically clearer intent and lower competition.
LSI Keywords: Latent Semantic Indexing keywords: terms semantically related to a primary keyword that signal topic comprehensiveness. 'Espresso', 'cappuccino', and 'latte' are LSI keywords for 'coffee'; using them can strengthen relevance signals.

M

Manual action↗: A penalty issued by Google's spam team after human review. Reported in Search Console; requires reconsideration request to lift.
MCP (Model Context Protocol)↗: An open standard developed by Anthropic for connecting AI models to external tools and data sources. Adopted across the AI industry including by Google and Microsoft. WebMCP is the browser-native adaptation that lets websites expose callable tools to in-browser AI agents.
Merchant Center↗: Google's product data platform. Accepts a feed or reads structured data from product pages, and supplies the listings behind Shopping surfaces and the Shopping Graph. Markup read by the crawl route must be in the served HTML, not generated by JavaScript.
Meta description↗: An HTML attribute summarising a page for search engines and link previews. Not a direct ranking factor; influences CTR.
Meta robots tag↗: An HTML tag (`<meta name="robots">`) instructing search engines how to index and display a page: noindex, nofollow, noarchive, max-snippet. Governs indexing rather than crawl access, so the page must stay crawlable for the directive to be read.
Meta title↗: See Title tag.
Microsoft Clarity↗: A free Microsoft behaviour analytics tool that now carries three SEO-relevant dashboards: Citation (observed citations and grounding queries from AI answers), Topic Insights (your own prompt sets run against a grounded model), and Bot Activity (CDN logs showing which AI crawlers hit your site).
Mobile-first indexing↗: Google's policy of using the mobile version of a site as the primary version for indexing and ranking. Standard since 2019.
Mojibake: Text corrupted by character encoding mismatches, typically appearing as garbled symbols (e.g. â€" instead of an em dash).
MSV (Monthly Search Volume)↗: The estimated average number of times a keyword is searched per month in a given market. Sourced from Google Keyword Planner data and processed by third-party tools, so figures are estimates, often rounded to buckets, and least reliable at low volumes. Also called search volume or keyword search volume.
Multimodal search↗: Search that combines more than one input type in a single query: image and text, video and voice, or camera and spoken question. Google Lens and AI Mode both support multimodal queries; the visual input narrows the query beyond what text alone can express. Multimodal queries are more likely to trigger AI Overviews than equivalent text-only queries.

N

NAP (Name, Address, Phone)↗: The core business information that must stay consistent across local citations for local SEO.
NavBoost↗: A Google system, named in the antitrust disclosures and the 2024 Content Warehouse leak, that re-ranks results using aggregated click behaviour over a rolling thirteen-month window, segmented by location and device. Described in sworn testimony as a lookup table of observed clicks rather than a machine-learning model.
Negative SEO↗: A deliberate attempt to harm a competitor's search rankings. The classic method, pointing toxic backlinks at a target, is largely defused: SpamBrain neutralises most spam links by ignoring them rather than penalising the site they point at.
NLWeb (Natural Language Web)↗: An open-source Microsoft protocol that turns a site's existing structured data into a natural-language queryable interface, exposing an /ask endpoint for REST queries and an /mcp endpoint for agent-to-agent communication so AI agents can query a site directly rather than crawling it page by page.
Nofollow↗: A link attribute (`rel="nofollow"`) telling Google not to associate your site with the linked page. Since March 2020 it is a hint rather than a directive, so a nofollowed URL can still be crawled and indexed if Google reaches it another way.
Noindex↗: A meta tag (or HTTP header) instructing search engines not to index a page. Pages must be crawlable for the directive to be seen.
Non-branded search↗: Queries that do not include a company or product name. The demand SEO earned rather than demand the brand already had. Newer sites should expect the large majority of their organic traffic to be non-branded.

O

Off-page SEO↗: Optimisation activities that happen outside a site: link building, digital PR, brand mentions, citations.
On-page SEO↗: Optimisation of elements on individual pages: title tags, headings, content, internal linking, schema.
Open Graph↗: Meta tags (og:title, og:description, og:image) controlling how a URL appears when shared on social media and messaging platforms.
Organic search↗: Search results not paid for by advertising. Organic traffic is the audience SEO is built around.
Orphan page↗: A page with no internal links pointing to it. Difficult for crawlers to discover and receives no internal link equity.
Outbound link↗: A hyperlink from a page on your site to a page on a different domain. Passes link equity if dofollow, and signals topical associations to search engines. Paid and affiliate outbound links must be qualified with `rel="sponsored"` or `rel="nofollow"`.

P

PAA (People Also Ask)↗: An expandable question-and-answer accordion in Google's results, present on around half of all SERPs. Google generates and expands PAA boxes in real time as users click answers. The same content signals that earn featured snippets earn PAA appearances.
Page speed↗: How quickly a page loads and becomes usable. Measured through Core Web Vitals in the field and tools such as Lighthouse in the lab. A ranking signal, but a modest one next to relevance.
Page title↗: See Title tag.
PageRank↗: Google's original algorithm for valuing pages based on the quantity and quality of inbound links. Still influential, though heavily evolved.
Pagination↗: Splitting long content or lists across multiple pages. Affects crawl, indexing, and how authority flows through a site.
Parasite SEO↗: Publishing content on a high-authority third-party domain to borrow its ranking signals rather than earning them on your own site. Google named this site reputation abuse as a spam policy in March 2024 and began manual actions that May.
Passage indexing↗: Google's ability to index and rank individual passages within a page independently of the page's overall topic, allowing relevant sections to surface for specific queries.
PBN (Private Blog Network)↗: A network of sites created specifically to manipulate link equity. Violates Google's guidelines.
Personalisation↗: The adjustment of search results based on a user's location, search history, device type, and other signals. Location remains the largest source of variation on classic results, but Google now promotes sites a user returns to often and connects Gmail and Photos to AI Mode. Means the same query can produce different results for different users, complicating rank tracking.
Pillar page↗: A comprehensive overview page on a topic, linking to deeper cluster pages on specific sub-topics.
Preferred Sources↗: A Google Search setting letting users nominate sites they want to see more of. Launched for Top Stories in 2025 and extended to AI Overviews and AI Mode in May 2026. Google reports users are roughly twice as likely to click through to a Preferred Source. Any site publishing fresh content is eligible, making it one of the few personalisation signals a publisher can actively ask readers to set.
Product schema↗: Structured data describing a product: name, image, and an Offer carrying price and currency are all Google requires for merchant listing eligibility. Return policy, shipping details and ratings are recommended, and are what produce the richer listing display.
Programmatic SEO↗: Generating pages at scale from structured data and templates, targeting many related queries at once rather than writing each page individually. Works when every generated page answers a genuinely distinct query; fails when pages are near-duplicates with one variable swapped.
Prompt Engineering: The practice of designing input prompts to guide language models toward specific, accurate outputs. Includes techniques like chain-of-thought reasoning and few-shot examples.

Q

QDF (Query Deserves Freshness)↗: Google temporarily raising the weight given to recency for queries that show signs of being time-sensitive, such as breaking news or trending topics. Triggered by a spike in search interest and fresh publishing on a topic; stable, evergreen queries do not qualify. The term comes from a 2007 interview with Google's then head of search, and now maps to what Google's documentation calls the Freshness Systems.
Query: The words or phrase a user types into a search engine. Distinct from 'keyword' in that a query is what the user typed; a keyword is what the SEO targets.
Query fan-out↗: A technique used by Google AI Mode in which a single user query is expanded into multiple parallel sub-queries, each targeting a different aspect of the question, with results synthesised into a single answer. Enables AI Mode to retrieve broader and deeper information than a conventional single-pass search. In multimodal queries, AI Mode fans out across both the scene as a whole and individual objects within an image.

R

RAG (Retrieval-Augmented Generation)↗: A technique combining document retrieval with language model generation: the system retrieves relevant passages, then feeds them to an LLM for synthesis. Allows grounding responses in current information rather than parametric knowledge alone.
Rank tracking↗: Tools that simulate a Google query for a defined keyword, location, and device, then record where a URL appears. Differs from GSC average position, which aggregates across all real users, query variants, devices, and locations.
RankBrain: Google's machine learning ranking component (introduced 2015) that learns from user behaviour and search patterns. Treats similar queries as equivalent and predicts relevance for never-before-seen queries.
Re-ranking↗: The second stage of Google's ranking process, in which an initial scored list of candidate results is adjusted, by boosts, demotions, and diversity rules, before it reaches the user. Performed by functions Google internally calls twiddlers.
Redirect (301)↗: An HTTP status code indicating a permanent URL change. Transfers ranking signals to the destination URL.
Redirect (302)↗: An HTTP status code indicating a temporary URL change. Does not transfer ranking signals.
Redirect chain↗: A sequence of redirects (URL A → URL B → URL C). Reduces signal transfer and adds latency. Best avoided.
Referring Domain: A unique domain that links to your site. A site earning backlinks from 10 different domains is stronger than one with 100 links from the same domain; domain diversity matters more than link count.
Render budget↗: An SEO-community term (not a Google label) for the finite computational resources Google's Web Rendering Service spends executing JavaScript on already-fetched pages. Extends the crawl budget idea to the second indexing pass; JavaScript-dependent content waits longer in the render queue, especially on large or lower-authority sites.
Rendering↗: The process of executing JavaScript and producing the final DOM. Googlebot renders pages in a separate pass after initial HTML processing.
Review gating↗: Soliciting reviews only from customers you expect to be positive, or routing dissatisfied ones to a private feedback form instead. Prohibited by Google's review policy, and treated by the CMA as cherry-picking under the DMCC Act.
Rich result↗: An enhanced search listing produced by schema markup: star ratings, recipe cards, event dates, product prices, video carousels. Google documents around 25 eligible schema types. FAQ rich results were removed in May 2026 and HowTo in September 2023.
Robots.txt↗: A plain text file at the root of a domain declaring which crawlers can access which paths. A request, not enforcement.

S

SafeSearch↗: A user-controlled filter that removes explicit content from Google's results at display time. Not a penalty: filtered pages stay fully indexed and rank normally for users with SafeSearch off.
Scaled content abuse↗: Generating large numbers of pages primarily to manipulate rankings rather than help users. A named Google spam policy since March 2024. The method is irrelevant: AI-generated, scraped, stitched, or spun content all qualify when produced at scale without adding value. Using AI is not itself a violation; manufacturing volume is.
Schema markup↗: Structured data using the schema.org vocabulary, declaring the meaning of page content explicitly to search engines and AI systems.
Scraping↗: Taking content from other sites, often by automated means, and republishing it to manipulate rankings. A named Google spam policy. The deliberate, manipulative side of duplicate content: copying competitors' pages, lightly rewording stolen content, or reproducing feeds wholesale without adding value. Distinct from the technical duplicate content Google resolves through canonicalisation.
Search everywhere optimisation↗: Building presence across every platform where discovery happens, not only Google: YouTube, Reddit, TikTok, LinkedIn, and AI answer engines. About one in three US consumers now start a search on a social platform.
Search intent↗: The underlying goal behind a query: informational, navigational, commercial investigation, or transactional.
Search Intent Drift: A shift in the dominant intent of a keyword over time, often from informational to commercial or vice versa. Requires content format and angle adjustments; a page optimised for the old intent may lose rankings as user needs evolve.
Search volume↗: See MSV (Monthly Search Volume).
Seed Keyword: A broad head term or topic that starts keyword research. 'Keyword research', 'SEO', 'content marketing' are seed keywords; cluster research expands them into related long-tail terms.
Semantic HTML↗: Using native HTML elements that communicate meaning rather than appearance: headings for topic hierarchy, article and section to bound content regions, nav and main as page landmarks. Not the same as structured data. Matters for AI retrieval because heading-based chunking is how AI platforms split a page into retrievable passages.
Semantic search↗: Google understanding the meaning behind a query rather than matching its exact keywords. Began with Hummingbird in 2013 and accelerated through RankBrain, BERT, and MUM. Makes keyword density thinking obsolete.
SEO (Search Engine Optimisation)↗: The practice of improving a site's visibility in unpaid search results.
SEO A/B testing (split testing)↗: Measuring the causal impact of an SEO change by splitting a set of near-identical pages into a control group and a variant group, applying the change only to the variant, and comparing organic traffic between them. Splits pages, not users, so crawlers and visitors see the same HTML. Requires a large set of templated pages to reach significance, which makes it a large-site technique.
SEO audit↗: A structured assessment of a website's technical health, on-page quality, and off-page authority, used to identify issues and prioritise improvements. Can cover the full site or focus on a specific area such as technical SEO or content.
SEO recovery↗: The process of diagnosing why a site's rankings or traffic fell and applying the right fix. A rankings drop and a traffic drop are different problems: a core update requires quality improvement and the next update cycle, a spam classification requires removing the violating content, and AI Overviews displacement is addressed by becoming the cited source rather than by improving position.
SEO ROI (Return on Investment)↗: The return earned on SEO spend. Harder to attribute than paid search, because organic rankings persist after investment stops and contribute to conversions without taking last-click credit. The CPC equivalency method gives a defensible monthly traffic value.
SEO title↗: See Title tag.
SERP (Search Engine Results Page): The page returned in response to a search query, including organic results, ads, and SERP features.
SERP analysis↗: Examining the top-ranking results for a query to reverse-engineer what Google believes best satisfies it. The content type, format, and angle of the current results reveal the intent model Google is applying.
SERP features↗: Non-standard result formats displayed alongside or in place of regular organic listings: featured snippets, AI Overviews, local packs, knowledge panels, image carousels, and more. Triggered by query type and structured data.
Share of voice↗: A site's proportion of total available organic visibility across a defined keyword set, relative to everyone else ranking for those keywords. Used as a market-level KPI rather than a tactical one. Since 2025 the term also refers to AI share of voice, a separate metric measuring share of brand mentions across AI-generated answers; the two do not translate into each other.
Shopping Graph↗: Google's index of product information, assembled from Merchant Center feeds, product page structured data and the wider web. It supplies Google Shopping, free product listings, and the product recommendations in AI Mode and AI Overviews.
Site architecture↗: How pages on a site are organised and linked together. A flat architecture keeps important pages within a few clicks of the homepage, concentrating crawl budget on high-value URLs and distributing PageRank efficiently. Deep hierarchies bury pages and reduce crawl frequency.
Site migration↗: A significant structural change to a website: moving to a new domain, switching platforms, changing URL structures, or consolidating sites. Carries high risk of ranking loss if redirects, canonical signals, and crawl directives are not handled correctly.
Sitelinks: Additional links to sub-pages shown beneath a brand's main search result. Algorithmically generated; not directly controllable but influenced by site structure and internal linking.
Sitemap↗: An XML file listing the URLs of a site, providing search engines with a catalogue for crawling.
Sneaky redirect↗: A redirect that sends users to a different destination than the one shown to search engines, or somewhere they did not expect. A named Google spam policy. Related to cloaking: both show one thing to crawlers and another to users, but a sneaky redirect moves the user elsewhere rather than swapping content in place.
Social search↗: Searching within a social platform rather than a search engine: TikTok, Instagram, Reddit, YouTube, Pinterest, LinkedIn. Each ranks by its own signals, and the results increasingly feed AI answer engines as citation sources.
Soft 404: A page returning a 200 OK status code but containing content that signals it should be a 404. Confuses indexing.
Sponsored (rel="sponsored")↗: The link attribute marking advertisements, paid placements and affiliate links. Leaving a paid link qualified by neither `sponsored` nor `nofollow` is link spam under Google's policies and can attract a manual action.
SSR (Server-Side Rendering): Rendering HTML on a web server before sending it to the browser. Full content is delivered to crawlers immediately; contrasts with client-side rendering, where crawlers may see incomplete pages.
Structured data↗: Machine-readable metadata embedded in a webpage. Most commonly delivered as JSON-LD using schema.org vocabulary.
SynthID↗: Google DeepMind's watermarking system, covering images, video, audio and text. For text it modulates the probability scores the model uses to pick each word, so a pattern is embedded without adding or removing anything. The public detector portal is in limited testing.

T

Technical SEO↗: Optimisation of the infrastructure that lets search engines crawl, render, and index a site: robots.txt, sitemaps, site architecture, canonical tags, page speed, and structured data.
Thin content↗: Pages with little or no unique value: doorway pages, scraped content, auto-generated text, or pages with minimal original material. A quality signal targeted since the Panda era.
Title tag↗: The HTML element (`<title>`) defining a page's title. Appears in browser tabs, SERPs, and link previews. One of the most influential on-page ranking factors. Also referred to as a meta title, page title, or SEO title.
Token: The basic unit of text processed by language models: typically a word or subword. Context window limits are measured in tokens, not characters or words. Longer responses consume more tokens.
Top Stories↗: A SERP carousel surfacing timely news content for news-oriented queries. Any publisher meeting Google's quality and freshness signals can appear; there is no application or approval process. Google News is a separate product with somewhat different requirements.
Topic cluster↗: A group of related pages organised around a pillar page, used to build topical authority.
Topical authority↗: A measure of how comprehensively and credibly a site covers a subject area, built through depth of content and internal linking structure. Influences rankings across a topic regardless of individual page signals.
Twiddler↗: A re-ranking function in Google's search architecture that adjusts an already-scored list of results, promoting, demoting, or filtering them before the page is shown. NavBoost is one example. The framework was named in 2023 antitrust testimony and the 2024 Content Warehouse API leak; Google confirmed the documents are genuine but cautioned they lack the context to infer exact weightings.

U

UCP (Universal Commerce Protocol)↗: An open standard for agentic commerce announced by Google at NRF in January 2026, co-developed with Shopify, Etsy, Wayfair, Target and Walmart. Defines a common language for discovery, checkout and post-purchase between agents, merchants and payment providers. A UCP-enabled site advertises itself at /.well-known/ucp.
UGC (User-Generated Content)↗: Content created by users rather than the site owner: comments, forum posts, reviews. Links within it should carry `rel="ugc"`, introduced alongside rel=sponsored in 2019 as a refinement of nofollow.
Unlinked Mention↗: A reference to a brand, product, or topic without a hyperlink. Unlinked mentions signal brand authority to search engines; link reclamation converts them into actual backlinks.
URL (Uniform Resource Locator)↗: The address of a webpage. Clean, descriptive URLs reinforce topical relevance and improve usability.
User Agent↗: A string of text sent by a browser or crawler with each request, identifying itself and its capabilities. Google operates multiple user agents (Googlebot, Google-Extended); different agents receive different treatment.

V

Viewport meta tag: An HTML meta tag (`<meta name="viewport">`) controlling how a page scales on mobile devices. Essential for passing mobile-friendliness signals.
Voice search↗: Searches performed by speaking to a device rather than typing. Queries are longer and more conversational than typed searches, with disproportionate local intent. Spoken results are sourced predominantly from the featured snippet: only one answer is read aloud, making it a winner-takes-all surface.

W

WebMCP (Web Model Context Protocol)↗: A proposed web standard letting websites declare callable tools for in-browser AI agents, co-developed by Google and Microsoft through the W3C. The action layer of agentic search: structured data tells AI what a page contains, WebMCP declares what it can do.
WebP↗: An image format with better compression than JPEG and PNG. Recommended for most web image use cases.
White hat SEO↗: SEO tactics that comply with search engine guidelines. Sustainable; carries no penalty risk.

X

XML sitemap↗: An XML-formatted sitemap providing search engines with a list of canonical URLs to crawl. See also: Sitemap.

Y

YMYL (Your Money or Your Life)↗: Content categories Google scrutinises with stricter E-E-A-T standards: health or safety, financial security, government/civics/society, and other.

Z

Zero-click search↗: A search where the user finds the answer in the SERP itself (via featured snippet, knowledge panel, or AI Overview) without clicking through.