SEO and GEO glossary: the key terms of search

This SEO and GEO glossary brings together the terms of organic search and those of AI answer engines, in alphabetical order. It is the vocabulary of an audit, a brief, a Search Console report or AI visibility tracking in ChatGPT. Here, GEO means Generative Engine Optimization, not geography. Each definition fits in one to three sentences and relies on the documentation of the vendor concerned when it exists. It links to the page of the site that explores the notion in depth.

0-9 A B C D E F G H I K L M N O P Q R S T U W X Z

0-9

301 redirect: permanent redirect from one URL to another. It passes on the signals of the old address and is a must during a migration or when merging pages: see the 301 redirect.

A

AEO (Answer Engine Optimization): a term used as a synonym of GEO, focused on answer engines. No standard definition distinguishes it from GEO.

AI crawlers: crawlers from AI vendors that visit websites. Each vendor distinguishes three uses: model training, the answer engine’s index and reading a page at a user’s request. Trade-offs and settings in our page on AI crawlers.

AI Mode: conversational tab of Google Search, which answers with a text and accepts follow-up questions. It uses the query fan-out technique: see our page on Google AI Mode.

AI Overviews: summaries generated by Google’s AI at the top of some results, with links to the source pages. Rolled out in France on July 22, 2026, they are covered in detail in the feature on AI Overviews.

AI referral traffic: visits arriving from an AI assistant: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai. They are isolated in a segment of the analytics tool.

AI SEO: expression with two meanings: using artificial intelligence to do SEO, or ranking in AI engines. Clarifying the meaning before discussing it avoids many misunderstandings; see what AI changes in SEO.

AI share of voice: the brand’s place among all the brands cited on a prompt panel, compared with its competitors. People also speak of share of voice or share of model.

AI visibility: a brand’s presence in the answers of AI engines, tracked through its mentions, its citations and its share of voice on a prompt panel. Tools and method in our page on measuring AI visibility.

Algorithm: set of rules and models that sort the pages of the index to answer a query. At Google, it is a combination of ranking systems, not a single formula: see Google’s algorithm.

Alt attribute: alternative text of an image, read by screen readers and by search engines. It describes the image when it carries meaning; it stays empty for a decorative image.

Anchor text: clickable text of a link. Google recommends a descriptive, concise anchor that announces the topic of the destination page; “click here” gives no information.

Answer engine: service that writes an answer to a question instead of showing a list of links, sometimes citing its sources. Also called AI search engine or generative engine: overview in our page on AI search engines.

Answer variability: an AI engine does not give the same answer twice to the same question. Reliable measurement therefore relies on repeated checks.

Average position: average rank of a page in the results, weighted by impressions, in Search Console. It is not a position observed at a given moment.

B

Backlink: link placed on another site and pointing to yours. Its value depends on the relevance and reliability of the source page, not on the raw number of links: see backlinks.

Backlink audit: analysis of a site’s inbound link profile: referring domains, anchors, target pages, suspicious links. See the backlink audit.

BERT: language understanding model used by Google to grasp how a combination of words changes the meaning of a query. It is listed in the guide to ranking systems.

Black hat: practices that try to manipulate rankings by breaking Google’s rules: hidden content, paid links, cloaking. They expose the site to a manual action or a loss of visibility.

Breadcrumb: navigation path that places the page in the site structure (Home > Section > Page). Marked up as BreadcrumbList, it can be shown in the results.

C

Cannibalisation: situation where several pages of the same site target the same intent and compete with each other in the results. It is fixed by merging, redirecting or differentiating the pages.

Canonical (URL): URL that Google keeps as the reference among several pages with identical or very similar content. The rel=canonical tag points Google towards it without imposing it: see the guide to the canonical URL.

CCBot: crawler of Common Crawl, an open corpus used by many models. Blocking it excludes your pages from the following collections, without erasing the earlier ones.

CDN layer: service placed between the site and its visitors, such as Cloudflare, which can block AI crawlers without any explicit robots.txt setting. Checking actual access in the server logs is the only reliable control.

ChatGPT Search: web search feature of ChatGPT, powered by OpenAI’s OAI-SearchBot crawler. Answers show links to the sources consulted; levers in our page on ranking in ChatGPT.

ChatGPT-User: agent that visits a page when a ChatGPT user asks it to. OpenAI states that robots.txt may not apply to these visits.

Chunking: the search engine splitting a text into passages. Google states that you do not need to split your pages into small pieces yourself to be understood.

Citable content: page whose passages answer a precise question on their own, with sourced and dated facts. It is the content-side goal of GEO optimisation.

Citation: link to a page of the site shown as a source of an AI answer. The site is said to be “cited as a source”.

Citation rate: share of the prompts in a panel for which a URL of the site is cited as a source.

ClaudeBot, Claude-SearchBot, Claude-User: Anthropic’s crawlers, respectively for training, for Claude’s search index and for visits requested by a user.

Cloaking: showing search engines content that differs from what visitors see. It is a violation of the spam policies.

CLS (Cumulative Layout Shift): visual stability metric: it measures unexpected layout shifts. A score of 0.1 or less is considered good.

Content Signals: statements added to robots.txt, proposed by Cloudflare, that express a usage preference (search, AI answer, training). They are declarations, with no technical blocking.

Core update: major change to Google’s ranking systems. It is rolled out several times a year and announced on the Search Status Dashboard. It reassesses all content, without targeting any particular site.

Core Web Vitals: three experience metrics measured on real visits, at the 75th percentile: LCP, INP and CLS (web.dev). Measurement method in the guide to Core Web Vitals.

Crawl budget: number of pages Googlebot can and wants to crawl on a site within a given time. According to Google, the topic mainly concerns very large sites and those where many URLs remain discovered but not indexed: see crawl budget.

Crawling: first stage of how Google works. Its crawlers discover URLs by following links and sitemaps, then download the pages (Google, how Search works).

CTR (click-through rate): ratio between the clicks and the impressions of a page in the results. Search Console provides it by page and by query.

D

Direct answer (answer-first): answer of two or three sentences placed at the top of a page or section, before the details.

Duplicate content: identical or very similar content available at several URLs, on the same site or across sites. Google does not penalise it as such, but chooses a single version to show: see duplicate content.

E

E-E-A-T: experience, expertise, authoritativeness, trustworthiness. This framework is used in Google’s search quality rater guidelines to judge a page; trust is its central member. The same framework serves as a benchmark for making a source credible to answer engines: see the feature on E-E-A-T.

Entity: uniquely identifiable thing: person, organisation, place, product, concept. Search and answer engines connect entities to understand what a page is about, and which brand, beyond the words.

F

Facets (faceted navigation): catalogue filters (size, colour, price) that often generate one URL per combination. Poorly managed, they create thousands of pages with no demand and waste crawling.

Featured snippet: answer highlighted above the classic results, taken from a page and followed by its link. Google alone chooses the page and the passage.

Freshness: ranking system that favours recent content when the query calls for it, for example for news. A changed date with no real change to the content brings nothing.

G

Gemini, Claude, Le Chat: conversational assistants from Google, Anthropic and Mistral AI, able to run a web search. None of them provides a visibility report to website publishers.

GEO (Generative Engine Optimization): all the actions that make a brand and its content visible and cited in the answers of generative AI engines. The term comes from a research paper posted online on November 16, 2023 and presented at the KDD 2024 conference (Aggarwal et al.). Full overview in our GEO guide. GEO extends SEO without replacing it. Google considers that optimising for its AI features is still SEO (Google Search Central, guide updated on July 10, 2026); comparison in GEO vs SEO.

Ghost citation: page used as a source without the brand being named in the answer. The site feeds the answer, the brand gets nothing out of it.

Google Business Profile: Google’s free business listing, formerly Google My Business, shown in Maps and in local results: opening hours, reviews, photos, categories. See Google Business Profile.

Google Search Console: free Google tool that shows how the search engine sees a site: queries, clicks, impressions, indexing, Core Web Vitals, manual actions. Getting started in the guide to Google Search Console.

Google-Extended: a robots.txt token, not a crawler in its own right. It lets you refuse the use of your content for the training and grounding of some Google AI systems. It has no effect on Search, AI Overviews or AI Mode, which are controlled by Googlebot (Google, “AI features and your website”).

Googlebot: generic name of Google’s crawlers. Its identity is checked by reverse DNS, because a user agent can be spoofed.

GPTBot: OpenAI crawler that collects content that may be used to train its models (OpenAI documentation).

Grounding queries: phrasings used by the AI to search for content before citing a page. Bing Webmaster Tools shows them in its AI Performance report, launched on February 10, 2026 (Bing Webmaster Blog).

H

Hallucination: false statement produced confidently by a model. For a brand, it justifies regularly checking what engines say about it.

Heading tags (Hn): headings from H1 to H6 that structure a page. A single H1 announces the topic, H2 and H3 headings divide the content without skipping a level: see heading structure.

Hreflang: attribute that tells Google about the language or regional versions of the same page. The declarations must be reciprocal between versions.

HTTPS: encrypted version of the HTTP protocol. It is the expected standard: 301 redirect from HTTP, internal links in HTTPS, no mixed content.

I

Inauthentic mention: brand mention obtained artificially, in bulk or through disguised payment. Google states that seeking it is less useful than it seems, as its systems filter out spam.

Indexing: stage where Google analyses a crawled page and decides whether to add it to its index. A page that is not indexed cannot appear in the results, nor serve as a source for AI Overviews: see the guide to Google indexing.

INP (Interaction to Next Paint): responsiveness metric: it measures the delay between an interaction and the page’s visual response. It replaced FID in March 2024; 200 ms or less is considered good.

Internal linking: all the links between the pages of the same site. It helps crawlers discover pages and signals their relative importance: method in the guide to internal linking.

K

Keyword: term or phrase that a user types into a search engine. In SEO, people also say query; a page targets a main query and its variants, chosen through keyword research.

Keyword density: share of a keyword’s occurrences in the text. Google sets no threshold; benchmarks of around 1% are agency safeguards against over-optimisation.

Knowledge cutoff date: date after which a model has no knowledge from its training. For more recent facts, it has to go through a web search.

Knowledge Graph: Google’s knowledge base that connects entities and their attributes. Other vendors build comparable databases.

L

LCP (Largest Contentful Paint): loading metric: it measures the time it takes to display the largest visible element. The “good” threshold is 2.5 seconds or less.

Link building: strategy for acquiring external links to strengthen a site’s popularity. Buying links intended to manipulate rankings breaks Google’s rules: see link building.

LLM (Large Language Model): language model trained on large amounts of text to predict and produce text. It is the writing engine of AI assistants.

LLMO (Large Language Model Optimization): another term used as a synonym of GEO, focused on the language models themselves and on how well a brand is known in their knowledge. People also speak of LLM SEO.

llms.txt: text file proposed in 2024 to guide models towards the useful content of a site. Google Search does not use it and states that no “AI” file is needed: details in our llms.txt page.

Local SEO: visibility work on searches with a geographic dimension, such as a service followed by a city. It covers the business listing, reviews, local pages and consistent contact details. See the guide to local SEO.

Log file analysis: reading the server logs, which record every request received, including those from Googlebot and AI crawlers. It is the only source that shows what crawlers actually explore: method in the guide to log file analysis.

Long tail: all the long, precise queries, each searched rarely but numerous. Taken together, they often weigh more than a few generic queries.

M

Manual action: penalty applied by a human reviewer at Google when a site breaks the spam policies. It is reported in Search Console, under “Security & Manual Actions”, and is lifted through a reconsideration request after the fix.

Mention: the brand’s name in the text of an AI answer, with or without a link. The brand is said to be “cited as a brand”.

Mention rate: share of the prompts in a panel for which the brand is named.

Meta description: HTML summary of a page, sometimes shown under the title in the results. Google builds its snippets mainly from the content; the tag does not directly influence rankings. See the meta description.

Microsoft Copilot: Microsoft’s assistant, backed by the Bing index. It is the only engine whose vendor provides a citation report to sites, in Bing Webmaster Tools.

Mobile-first (indexing): Google indexes and ranks pages based on their mobile version. Content missing on mobile is therefore invisible to it.

N

Nofollow, sponsored, ugc: values of the rel attribute that qualify a link: no endorsement, sponsored, or user-generated. Google treats them as hints.

Noindex: directive, in a robots meta tag or an HTTP header, that asks Google not to index a page. To be read, the page must remain crawlable.

Non-commodity content: Google’s expression (non-commodity content). It refers to content that brings its own point of view, experience or data, beyond what everyone already knows.

O

OAI-SearchBot: OpenAI crawler that powers ChatGPT search. A site that blocks it no longer appears in these answers, except possibly as a navigation link.

On-page / off-page: on-page SEO covers what is worked on the page itself: content, tags, internal linking. Off-page SEO covers what happens elsewhere: links, mentions, reviews. Levers in the guide to on-page SEO.

Organic traffic: visits coming from a search engine’s organic results, as opposed to paid traffic. It is best tracked excluding brand queries.

Orphan (page): page that receives no internal link. It is only discovered through the sitemap or an external link, and appears unimportant.

P

PAA (People Also Ask): block of related questions shown in Google’s results. Its questions serve as a list of sub-intents to cover.

PageRank: Google’s historical algorithm that estimates the importance of a page from the links pointing to it. It is still one of the ranking systems: see the feature on PageRank.

Pagination: splitting a long list into numbered pages. Google no longer uses rel=next and rel=prev: each paginated page keeps its own URL and sequential links.

Passage (or chunk): piece of a page that the engine extracts and assesses separately. A well-written passage can be understood on its own, without the previous paragraph.

Perplexity: answer engine that shows its sources throughout the answer. Its index is fed by PerplexityBot.

Perplexity-User: agent that reads a page at the request of a Perplexity user. It generally ignores robots.txt.

PerplexityBot: crawler that makes sites appear in Perplexity’s results. According to the vendor, it is not used to train models (Perplexity documentation).

Prompt: question or instruction given to a model. A prompt is often longer and more precise than a query typed into a classic search engine.

Prompt injection: text hidden in a page to give instructions to a model that reads it. A risky practice, to be banned, which can lead to a penalty.

Prompt panel: fixed list of questions representative of a market, checked at a fixed interval on several AI engines. Without a stable panel, no trend can be compared.

Q

Query fan-out: series of related queries run in parallel by the model to cover the sub-questions of a request. Google gives the example of a question about a lawn overrun by weeds, broken down into queries on weedkillers, chemical-free methods and prevention.

R

RAG, grounding: technique that consists of retrieving pages from a search index, then writing the answer from them. Google states that its AI features rely on this principle, backed by its usual ranking systems.

RankBrain: Google machine learning system that connects the words of a query to concepts, including for searches never seen before.

Ranking systems: all the automated systems Google uses to order results: language understanding, links, reliability, freshness, deduplication. Google publishes the list in its guide to ranking systems.

Re-ranking: second sorting of the retrieved passages, by relevance, before the answer is written.

Referring domain: distinct site that links to yours at least once. Ten links from the same domain count as a single referring domain.

Rendering (JavaScript): stage where Google runs a page’s JavaScript, with a recent version of Chrome, to see the final content. Content that is only visible after rendering is processed later: see JavaScript SEO.

Rich result: search result shown with additional elements (rating, price, image), made possible by structured data. Google adds and removes some over time.

Robots.txt: file placed at the root of the site that tells crawlers which areas not to crawl. It manages crawling, not indexing (Google, introduction to robots.txt); settings in the guide to robots.txt.

S

sameAs: property of the schema.org vocabulary that links a page to other profiles describing the same entity, for example in Organization markup.

Scaled content abuse: generating many pages with the main purpose of manipulating rankings, with no added value, whatever the tool used. This practice has been listed in Google’s spam policies since March 2024.

SEA (Search Engine Advertising): advertising on search engines, such as Google Ads: pay-per-click ads shown above or below the organic results. See the SEA guide and how SEO and SEA work together.

Search Console generative AI report: report that shows the impressions obtained in AI Overviews and AI Mode, by page, country and date. It gives neither clicks nor queries (Google Search Central Blog, June 2026). It has covered all sites since August 31, 2026.

Search intent: the real need behind a query: getting information, comparing, buying, finding a specific site. The type of pages already ranking in the SERP reveals it; method in the guide to search intent.

Semantic audit: analysis of a site’s content against the targeted queries: intents covered, missing pages, cannibalisation, lexical field. See the semantic SEO audit.

Semantic cocoon: method for organising pages and internal links by closeness of meaning, formalised by Laurent Bourrelly, a French SEO practitioner. It is a practitioner’s method, not documented by Google: see semantic cocoons.

SEO (Search Engine Optimization): organic search optimisation. These are the actions that help search engines crawl, understand and rank a site’s pages, to attract visitors without paying for the click. Overview in the guide to SEO.

SEO audit: diagnosis of a site on three axes, technical, content and popularity, leading to a list of actions ranked by priority. Steps and tools in our SEO audit method.

SEO consultant: specialist, freelance or in an agency, who audits a site, sets search priorities and supports the teams that implement them. See the role of the SEO consultant.

SEO content: all the pages designed to meet search intents: guides, articles, category pages, FAQs. Strategy and formats in our guide to SEO content.

SEO crawler: software that explores a site as a search engine crawler would, to collect HTTP status codes, tags, links and page depth. Screaming Frog SEO Spider is one example: see our Screaming Frog guide.

SEO migration: change of domain, URL structure, CMS or templates that affects indexed pages. A redirect plan and monitoring of Search Console and the logs limit the losses: see SEO migration.

SEO tools: software that measures and diagnoses a site’s visibility: Search Console, crawlers, rank tracking, link analysis. Comparison in our selection of SEO tools.

SEO writing: writing pages designed for both the reader and search engines: answering the intent, heading structure, tags, internal linking, sources. Method in the guide to SEO writing.

SERP (Search Engine Results Page): a search engine’s results page. It mixes organic results, ads, AI Overviews, related questions, maps and videos.

SGE (Search Generative Experience): former name of Google’s AI answers test, in 2023 and 2024. It gave way to AI Overviews and AI Mode: a source that still talks about SGE predates this change.

Silo: organisation of a site into sealed thematic branches, with internal links staying inside each branch. A practitioner’s method, not documented by Google.

Site reputation abuse: publishing, on a reputable site, third-party content unrelated to its business in order to benefit from its reputation. It is a violation of Google’s spam policies.

Site structure: hierarchical organisation of a site’s pages, from the homepage to deep pages, visible in the menu and the URLs. It guides visitors as well as crawlers: see the guide to website structure.

Sitemap index: file that lists several XML sitemaps, useful when a site exceeds the limits of a single file or separates its page types. Structure and rules in the guide to the sitemap index.

Soft 404: error or empty page that responds with a 200 code instead of a 404 code. Google treats it as an error, and it wastes crawling.

Structured data: markup, usually in JSON-LD format using the Schema.org vocabulary, that describes the content of a page to search engines (Google, introduction to structured data). It makes a page eligible for some rich results and must describe what is visible: see structured data.

T

Technical SEO: part of SEO that makes a site crawlable, indexable and fast: HTTP status codes, robots.txt, sitemaps, canonicals, rendering, performance. Levers and priorities in the guide to technical SEO.

Technical SEO audit: part of the audit devoted to crawling, indexing, rendering and page speed. It is carried out with a crawler, Search Console and, if possible, the logs. See the technical SEO audit.

Title tag: HTML title of a page, generally used as the clickable title in the results. Aim for 50 to 60 characters, with the main query first; Google may rewrite it. Details in the guide to the title tag.

Training data: texts used to train a language model. What it retained from them is frozen at the end of training.

U

URL: unique address of a page. A good URL is short, readable, in lowercase, with words separated by hyphens, and stays stable over time.

W

White hat: SEO practices that comply with search engine guidelines: useful content, clean technical setup, links earned on their merit.

Wikidata and QID: Wikidata is a free, structured knowledge base; each entity in it receives a stable identifier, the QID. It is used to designate a brand unambiguously.

X

X-Robots-Tag: HTTP header that carries the same directives as the robots meta tag, such as noindex. It is used for files without HTML, for example PDFs.

XML sitemap: file that lists the URLs to crawl, with their modification date. It helps discovery, without guaranteeing indexing, and should only contain canonical pages returning a 200 code: see the XML sitemap.

Z

Zero-click (search without a click): search that ends without a click to a website. The answer is shown directly in the SERP (snippet, AI Overview, business listing) or in an AI assistant.