40 clear, sourced definitions written for humans and AIs. Missing a term? Email us — we'll add it.
Experience, Expertise, Authoritativeness, Trustworthiness — Google's quality criteria.
Quality Rater Guidelines criteria used to assess a site, especially in YMYL niches. Implementation: author pages, sources, Person/Article schemas, press mentions.
Pages that may impact health, finance or safety — stricter requirements.
Google applies stricter quality standards to YMYL content. LLMs likewise demand stronger sources before citing.
Acquisition of authoritative backlinks pointing to the site.
Still a major factor. OMNIRK prioritizes digital PR, editorial guest posts and entity links (Wikipedia, Crunchbase).
External link pointing to a page.
Evaluated by Google on source domain authority, topical relevance and anchor text. One Wikipedia backlink beats 100 directory links.
Clickable text of a link — semantic relevance signal.
Ideal profile: varied (brand, exact, partial, naked URL). Over-optimization = Penguin penalty.
Click rate on a search result.
Levers: catchy title < 60 chars, meta description < 160, schemas (stars, FAQ, breadcrumb) enriching the snippet.
Secondary links displayed under the main Google result.
Auto-generated by Google based on internal architecture, anchors and SiteNavigation schema. Can double or triple branded CTR.
Network of links between pages of the same site.
Distributes PageRank, structures semantics, speeds up crawler discovery. Underrated lever: good internal linking = significant traffic gain with no new content (internal OMNIRK estimate).
Goal behind a query: informational, navigational, transactional, commercial.
Align content format with the dominant SERP intent. Common mistake: writing a guide for a transactional query.
Search Engine Results Page.
Now includes AI Overviews, People Also Ask, Featured Snippets, Knowledge Panel, videos, images. Pure organic share shrinks every year.
Box at the top of the SERP that directly answers the query.
Paragraph, list or table format. Won with a concise (40-60 word) answer right after the H2 restating the question. A Position 0 page also feeds AI Overviews and LLM citations.
Block of related questions displayed in Google's SERP.
Clicking a question expands more, infinitely. Optimization: question-shaped H2 + 40-50 word synthetic answer + FAQPage schema. Underused long-tail traffic source.
Google update targeting content written for SERPs rather than humans.
Launched in 2022, integrated into the core algo in 2024. Penalizes sites with unedited AI content, scraping, over-optimization. Inverse criteria: real expertise, unique value, post-click user satisfaction.
Mass generation of landing pages from a template + dataset.
Examples: Zapier (20k+ 'X integrates with Y' pages), Wise (currency pairs), Tripadvisor. Extremely powerful strategy in 2026 if each page delivers unique value (data, comparison). Helpful Content risk if templates are empty.
Optimization for geo-located queries and the Google Maps pack.
Combines Google Business Profile (ex-GMB), consistent NAP citations, reviews, LocalBusiness schema, 'service + city' pages. Critical for law firms, restaurants, agencies, retail. Over 46% of Google searches have local intent.
Google business listing (formerly Google My Business), the pivot of local SEO.
Optimization: precise categories, weekly photos, posts, Q&A, proactive review management (>4.5 stars + replies), attributes, products/services. The profile appears in the Local Pack, Maps and increasingly in AI answers for 'near me' queries.
SEO applied to product pages, categories and merchant architecture.
Specifics: facets / filters (controlled canonical + noindex), Product + Offer + AggregateRating schema, variant handling, unique category copy, breadcrumbs, out-of-stock handling. Heavy weighting of Merchant Center and shopping AI Overviews.
Major Google algorithm update rolled out several times per year.
Recalibrates how quality, authority and relevance are perceived web-wide. No technical 'fix': a dropping site must rework E-E-A-T, intent, depth. Recovery cycles of 3-6 months minimum.
Search whose answer is consumed in the SERP with no click to a site.
Over 60% of Google searches are now zero-click (SparkToro). On ChatGPT/Perplexity, the ratio is close to 100%. Strategy: aim for citation/brand rather than the click, measure AI share-of-voice, monetize awareness captured in the SERP.
Number of pages Googlebot will crawl on a site in a given time window.
Critical beyond 10k URLs. Optimizations: robots.txt, removing low-value pages, parameter handling, segmented sitemap, IndexNow, server speed. GEO side: LLM bots (GPTBot, ClaudeBot, PerplexityBot) have their own budget, to monitor.
Rendering and indexing of JS sites (React, Vue, Angular) by Google and LLMs.
Google renders JS in a second wave with delays. LLM bots barely render JS: SSR or SSG is mandatory for GEO. Tests: URL Inspection, view-source, JS disabled, fetch as GPTBot.
Multi-country / multi-language strategy aligning hreflang, ccTLD, localized content.
Architecture choice: ccTLD (.fr/.de), subdomain (de.site.com) or subfolder (site.com/de). Full and symmetric hreflang, truly localized content (not just translation), Search Console per property, currency and tax per market.
Digital press relations generating authoritative backlinks + brand mentions.
Data studies, press releases, expert quotes (HARO / Qwoted), op-eds. Double benefit: DA 70+ backlinks for SEO, co-occurrence mentions for GEO. Favorite format of LLMs that readily cite reference media sources.
Server log study to understand real bot behavior.
The only 100% reliable crawl source: who (Googlebot, GPTBot, ClaudeBot, PerplexityBot), what, when, HTTP status. Detects crawl-budget waste, forgotten pages, 4xx/5xx errors. Essential beyond 50k URLs and to steer GEO.
Image optimization for Google Images, Lens and multimodal understanding.
Modern format (WebP/AVIF), descriptive filename, accurate alt text, lazy-loading, explicit dimensions, ImageObject schema. Critical for e-commerce, real estate, travel. Gemini and GPT-4o read images: rich alt boosts AI citation.
Google results with enriched display (stars, price, FAQ, recipes, events).
Distinct from Featured Snippet: a Rich Result visually enhances a classic result via schema (Product, Recipe, Event, FAQPage, HowTo, Review). Boosts CTR by +20 to +35%. Test via Rich Results Test (search.google.com/test/rich-results).
Google Search Console tool to neutralize toxic backlinks.
A disavow.txt file submitted to Google declaring links to ignore (PBN, spam, negative SEO). Use sparingly since 2022: Google generally ignores low-quality links by itself. Still useful for manual penalties or proven negative SEO attacks.
Third-party scores (Moz DA, Ahrefs DR, Majestic TF/CF) estimating domain authority.
Scores out of 100 NOT used by Google but industry standards to qualify link-building opportunities. Target: DA/DR 50+ for a worthwhile backlink. Complement with Majestic TF (Trust Flow) and CF (Citation Flow) to assess real editorial quality.
Spoken queries via Siri, Alexa, Google Assistant — often conversational.
Longer, natural-language queries with strong local intent ('Italian restaurant open near me'). Optimizations: explicit FAQ, conversational content, Speakable schema, Position 0, up-to-date Google Business profile. Strong convergence with GEO optimizations.
Optimizing a site so AI answer engines cite it as a source.
GEO is the discipline of maximizing the chance that content is retrieved, understood and cited by ChatGPT, Perplexity, Gemini, Claude and Copilot. It combines llms.txt, advanced schema.org, entity anchoring (Wikidata), semantic structuring and topical authority.
Optimization for answer engines: Google AI Overviews, Bing Copilot, You.com.
AEO focuses on direct-answer extraction by hybrid engines (search + LLM). Key levers: Q&A structure, FAQ and HowTo schemas, self-contained content chunks.
Root-level file guiding LLMs through a site's content map.
Open standard (proposed by Jeremy Howard, 2024) acting as a robots.txt + sitemap equivalent for LLMs. Lists priority pages, descriptions and citation permissions.
OpenAI manifest declaring a site's capabilities to AI plugins/agents.
A .well-known file exposing an API or structured content to AI assistants. Originally part of the ChatGPT plugins standard; still relevant as an LLM-friendly signal.
Establishing a brand as an entity in Google's Knowledge Graph.
Combines a Wikidata entry, consistent mentions (NAP), Organization/Person schemas, sameAs links to official profiles. Required to appear in Knowledge Panels and to be disambiguated by LLMs.
Free knowledge base feeding Google, Wikipedia and LLMs.
Owning a Wikidata Q-id with verifiable properties is one of the strongest authority signals for AI models. OMNIRK creates and maintains client entries.
Google's graph of entities, attributes and relations.
Fueled by Wikidata, Wikipedia, structured data and web signals. An entity in the KG gets a Knowledge Panel and stronger semantic comprehension by LLMs.
AI-generated answers at the top of Google results.
Rolled out in France in 2025. To appear as a source: factual structured content, schemas, topical authority. OMNIRK tracks per-query appearances.
AI answer engine focused on sourced citations.
Always cites its sources. The most accessible engine for new GEO entrants. OMNIRK monitors appearances on 100+ queries per client.
OpenAI's search mode powered by GPT + Bing index.
Available to ChatGPT users since late 2024. Picks 3-5 sources per answer. Optimizations: llms.txt, authority, extractable content.
Google's model family integrated into Search, Workspace and AI Overviews.
Gemini 2.5 (2026) powers AI Overviews. Favors sources strongly tied to Knowledge Graph entities.
Anthropic's AI assistant, with citations since 2024.
Claude 3.5/4 Sonnet favors structured (schema) sources, long-form content and explicit author expertise.
Microsoft's assistant powered by GPT and the Bing index.
Massive distribution (Windows, Edge, M365). Bing Webmaster Tools + IndexNow speed up indexation.
LLM architecture combining document retrieval and generation.
All modern answer engines (Perplexity, ChatGPT Search, Gemini, AI Overviews) are RAG systems. Understanding RAG = understanding how to be cited.
Vector representations used by LLMs to compare text.
Content is retrieved for citation if its embedding is close to the query's. Lever: cover an entire lexical field around the target entity.
Identity card of an entity displayed on the right of Google results.
Triggered by Wikidata presence + sameAs consistency + press coverage. Major authority signal, leveraged by LLMs.
Google Search conversational mode powered by Gemini, rolled out in 2025.
AI Mode is gradually replacing the classic SERP for complex queries. It uses query fan-out (decomposing one question into parallel sub-queries) to synthesize an answer from multiple sources. Optimizations: chunkable content, schemas, entity authority, broad semantic coverage.
Decomposition of a query into multiple parallel sub-queries by AI engines.
Central mechanism of Google AI Mode and most modern RAG systems. To be cited, a site must rank not on the initial query but on the dozen sub-queries generated. Lever: exhaustive topical coverage, semantic hubs, extended FAQs.
Visibility on Reddit, heavily weighted by Google and LLMs since 2024.
Google-Reddit deal in 2024 + massive appearances in AI Overviews. ChatGPT and Perplexity cite Reddit as an 'authentic voice' source. Underused lever: non-spammy expert presence in key subreddits, posts answering long-tail intent.
Brand citation in text without a hyperlink — authority signal for LLMs.
LLMs reason by entity co-occurrence, not PageRank. A brand often mentioned alongside its competitors and expertise topics is learned as legitimate, even without backlinks. Lever: mention-focused digital PR, podcasts, panels.
Mechanism by which an LLM grounds its answer in verifiable web sources.
Official Google Gemini (Grounding with Google Search) and Anthropic term. A grounded answer cites its sources and reduces hallucinations. To be a grounding source: SSR, schemas, freshness (lastmod), entity authority.
Understanding meaning and intent rather than keyword matching.
Backbone of modern engines (BERT, MUM, Gemini) and all RAG. Optimizing is no longer 'place the keyword' but 'cover the semantic field': related entities, synonyms, related questions, context. Tools: embeddings, NLP, coverage audits.
Embedding-similarity search, technical foundation of RAG and LLMs.
Content is encoded into vectors (OpenAI, Cohere, Voyage) and compared by cosine similarity. To be 'retrievable', a passage must be semantically dense, self-contained and well chunked (200-400 tokens). Direct GEO lever.
Brand citation share inside AI engine answers vs competitors.
Central GEO KPI. Measured via panels of standardized prompts probed monthly on ChatGPT, Perplexity, Gemini, Claude, Copilot. Tools: Profound, Athena HQ, Otterly.ai — or in-house barometer. Gradually replaces Google rank tracking.
AI engine crawlers that scrape the web for training and real-time citation.
GPTBot (OpenAI), ClaudeBot/anthropic-ai (Anthropic), PerplexityBot, Google-Extended, Bytespider, CCBot, Applebot-Extended. Must be explicitly allowed in robots.txt to be citable. Blocking = disappearing from AI answers. Monitor their hits via server logs.
Separate Google user-agent controlling content use for Gemini and Vertex AI.
Lets you allow Googlebot (classic SEO) while separately allowing or blocking Gemini training. For GEO: NEVER block it, or you'll be excluded from AI Overviews and AI Mode. Configured via robots.txt, independent of Googlebot.
Open Anthropic standard letting LLMs connect to external sources.
Launched late 2024 by Anthropic, adopted by OpenAI in 2025. Successor to ChatGPT plugins. A brand exposing an MCP server becomes citable and actionable directly from Claude, ChatGPT, Cursor. Strategic GEO pivot 2026: expose structured data via MCP.
Three successive stages of AI in Google Search, often confused.
SGE (Search Generative Experience, 2023, Labs only) → AI Overviews (2024, AI box atop SERPs, rolled out) → AI Mode (2025, dedicated conversational experience with query fan-out). All three coexist. Shared optimizations: schemas, chunkable content, entity authority, freshness.
Knowledge Graph successor, probabilistic fact base feeding Gemini.
Google project automatically extracting facts from the entire web with a confidence score, where the Knowledge Graph relied on curated sources (Wikidata, Wikipedia). To appear in it: atomic facts, exhaustive schemas, cross-source consistency, Wikidata anchoring.
Semantic markup vocabulary recognized by Google and LLMs.
Schema.org describes entities, organizations, products, FAQs, articles. Served as JSON-LD in <head>, it is the lingua franca of the semantic web and a key signal for AI understanding.
Google's perceived-performance metrics: LCP, INP, CLS.
LCP (Largest Contentful Paint) < 2.5 s, INP (Interaction to Next Paint) < 200 ms, CLS (Cumulative Layout Shift) < 0.1. Page Experience criteria and quality signal.
Largest Contentful Paint — render time of the largest element.
Target: < 2.5 s. Levers: optimize LCP image (preload, AVIF/WebP), reduce TTFB, prioritize critical CSS.
Interaction to Next Paint — global responsiveness to interactions.
Replaced FID in March 2024. Target: < 200 ms. Levers: avoid long JS tasks, prioritize the main thread.
Cumulative Layout Shift — visual stability.
Target: < 0.1. Levers: reserve space for images/iframes, avoid late-injected elements.
HTML rendered server-side, delivered complete to crawlers and LLMs.
Critical for GEO: AI crawlers don't all execute JS. SSR ensures content is immediately readable and indexable.
Push-indexation protocol for Bing, Yandex, Naver.
Instantly notifies engines of new content. Implemented in 5 minutes via API key. Recommended for GEO (Bing powers Copilot and ChatGPT).
File controlling crawler access, including GPTBot, ClaudeBot, PerplexityBot.
Blocking AI bots = exiting LLM results. OMNIRK explicitly allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended.
Structured list of a site's URLs for crawlers.
Essential. Must include hreflang for multilingual sites, current lastmod, logical priorities.
Attribute indicating the language/region of an alternate page.
Declared in <head> and in the sitemap. Include x-default as a fallback. Common mistake: missing self-reference.
Reference URL for content when duplicates exist.
Prevents link equity dilution. Must be absolute, self-referencing by default, consistent with hreflang.
HTML tags that describe their function (article, section, nav).
Improves accessibility, SEO and LLM understanding. One <h1> per page, logical <h2>/<h3> hierarchy.
Google page performance score (0-100).
Measured by Lighthouse (lab) and CrUX (field). OMNIRK target: 95+ desktop, 90+ mobile.
Google indexes and ranks using the mobile version of the site.
Active since 2024 on 100% of the web. Mobile content MUST match desktop, otherwise rankings drop.
A site's authority on an entire topic, not just a single keyword.
Built through internal linking, semantic hubs and exhaustive coverage. The #1 signal for 2026 SEO ranking and LLM source selection.
Splitting content into self-contained, LLM-citable blocks.
Each section must read on its own. Include definitions, numbers, attributed quotes. Maximizes citation likelihood.
Pyramidal content architecture (Laurent Bourrelly method).
French method popularized by Laurent Bourrelly: parent → child → grand-child pages with strict internal linking funneling authority to the commercial target. Still effective in 2026, even more relevant for GEO (covers query fan-out).
Broad pillar page + deep satellite articles tied by internal linking.
HubSpot model, English-speaking cousin of the cocon sémantique. The pillar captures the broad query, clusters capture long-tail and feed authority back to the pillar. Now critical for GEO: covers the full query fan-out.
Video optimization for Google Video, YouTube and multimodal AI citations.
VideoObject schema, full HTML transcript, chapters, thumbnail, YouTube hosting + embed. YouTube is Perplexity's 2nd citation source and heavily used by multimodal Gemini. The transcript remains the primary LLM fuel.
Verifiable reputation and expertise of an author, critical for E-E-A-T and GEO.
Dedicated author page + Person schema + sameAs to LinkedIn, X, Wikidata, ORCID. LLMs propagate expertise: an article signed by an identified author cited elsewhere is more likely to be cited as a source. Pivot post-Helpful Content.
Gradual traffic erosion of aging content, fixed via refresh.
Any article loses 20-50% of traffic per year without upkeep. Refresh process: semi-annual GSC audit, data update, addition of AI-friendly sections (FAQ, TL;DR), republish with updated date, IndexNow ping. Often multiplies traffic by 2-3×.