Glossary

A 301 redirect sends visitors and machines from an old address to a new one permanently. The number is the code with which the server says: "this page has moved for good."

AEO (Answer Engine Optimization) is the practice of optimising content for AI answer engines such as ChatGPT, Perplexity or Google AI Overviews — so they find, understand and cite a website as a source. In practice it largely overlaps with GEO.

The AI Act is the European Union’s rulebook for artificial intelligence. It sorts applications by risk and attaches obligations to each tier — most visibly for businesses wherever content or answers come from an AI.

AI Overviews are AI-generated summaries that Google shows above the classic search results. They answer the query directly from several sources — often without anyone clicking through to a website. Visibility goes to whoever is cited as a source.

An AI agent is an AI system that does not just answer questions but completes tasks on a user's behalf: it plans steps, retrieves information and uses tools and interfaces — for instance to compare offers or prepare an enquiry.

An AI hallucination is a convincingly worded but false or fabricated statement by an AI system. Language models compute plausible phrasings — where a reliable source is missing, they guess. Risky for businesses: invented opening hours or services look deceptively real.

A website is AI-ready when AI systems such as ChatGPT, Gemini and Perplexity can read it without errors, classify it correctly and recommend it as a reliable source. It is not about a chatbot on your own site, but about being understandable to the AI systems out there.

Citability describes how easily AI systems like ChatGPT or AI Overviews cite a website as a source — the clearer, more current, and more unambiguous a piece of content is, the more often it gets picked as the answer.

Citation half-life describes how quickly unchanged content loses visibility in AI answers. A large 2026 study measured that, on average, content is cited less often after roughly 4.5 weeks — making freshness a measurable lever for AI visibility.

Cookieless describes a website that works without third-party tracking cookies — no foreign cookies, no cookie banner, full data sovereignty. Measurement is first-party, directly on your own domain.

A crawler is a program that fetches web pages automatically and follows their links — for search engines, archives or AI systems. Whatever it cannot fetch does not exist for the service behind it.

A digital knowledge and service interface is a website that makes its content readable for people and machine-queryable for AI agents at the same time — instead of merely displaying it as a designed page.

Dual-channel visibility means your website serves two audiences at once: people through classic search and AI systems through machine-readable content. Each channel has its own requirements — serve only one and you lose the other.

Earned media are mentions and recommendations you earn — in press, specialist articles or directories — instead of buying them or publishing them yourself. For AI visibility they count heavily: web mentions carry roughly three times the weight of classic backlinks.

First-party data is data a company collects directly and transparently from its own customers: enquiries, newsletter sign-ups, bookings or website statistics on its own infrastructure. It belongs to the business itself and works without third-party advertising cookies.

GEO (Generative Engine Optimization) is the practice of optimising content to be cited as a source in AI-generated answers — for example by ChatGPT or Perplexity. Statistics, sources and expert quotes increase the chance of being cited by up to 40% (ACM KDD 2024).

A Generative UI is an interface an AI assembles itself at the moment of use — tailored to the question, instead of pre-built. For businesses this means: content must exist as clear data, so the AI can place it into any interface.

A headless CMS separates content management from presentation: content is maintained once in a central place and delivered through an interface to any channel — website, app, shop or AI agent. No “head” (frontend) is hard-wired, hence the name.

hreflang is an HTML attribute that tells search engines which language and country version of a page is meant for which audience — so users and crawlers always land on the right language variant.

A knowledge graph links concepts, people and companies as entities with properties and relationships to one another — AI systems don't just read text this way, they understand what a business actually is and what it connects to.

A Large Language Model (LLM) is the language model behind AI assistants like ChatGPT or Gemini: trained on vast amounts of text, it generates answers in natural language. How it represents your business depends on how machine-readable your foundation is.

LLMO (Large Language Model Optimization) is the targeted optimisation of content for individual language models such as ChatGPT, Claude or Gemini — more technical and model-specific than GEO, which addresses generative search as a whole.

llms.txt is a small text file on your site that shows AI models, in a structured way, which content matters and how to understand it. It makes your foundation machine-readable — so assistants capture and recommend your business without errors.

MCP (Model Context Protocol) is an open standard that lets AI assistants access your website content in a structured way and with clear permissions. Your content becomes directly and correctly accessible to AI systems — instead of merely being crawled.

A media break is the switch of medium or system within a process — data has to be retyped or copied by hand. It costs time, creates errors and breaks automation. The more separate, isolated tools you run, the more expensive it gets.

NAP stands for Name, Address, Phone — your business's three core contact details. NAP consistency means they match everywhere. Contradictory NAP data leads AI assistants to recommend your business with wrong information.

Owned media are the channels you own and fully control — above all your own digital foundation. For fact questions it's the AI's preferred source: a large share of all AI citations comes directly from the company's own website.

RAG (Retrieval-Augmented Generation) is the technique AI answer engines work with: the language model combines its base knowledge with individual text passages retrieved live from the web before it answers. That's why the individual passage counts for visibility, not the whole page.

robots.txt is a text file in a website’s root directory. It tells automated clients which areas they may enter and which they may not. It is the oldest of these routes and by far the most widely served.

SEO is the work of making a website findable in search results: a comprehensible structure, sound technology, and content that genuinely answers a question. The newer disciplines aimed at AI answers build on this — they do not replace it.

Share of Model Voice (SOMV) measures how often, and in what context, a brand is mentioned in ChatGPT, Perplexity or Claude answers to typical industry questions — the modern equivalent of classic visibility and ranking metrics.

Single Source of Truth means: your important data lives in ONE central place instead of scattered across many programs. Everything draws from it — no contradictions, no outdated copies, one reliable picture of your business.

The Spatial Web distributes digital information in three-dimensional space instead of on a flat page — visible through augmented or virtual reality, anchored to real places and coordinates. The website becomes an object in space.

Structured data is machine-readable information embedded in a website's code, usually following the Schema.org standard in JSON-LD format. It tells search engines and AI systems unambiguously what content means — such as a business's opening hours, address, services or reviews.

A TDM opt-out is a machine-readable declaration that your content may not be used for text and data mining or the training of AI models. TDMRep is the format for it — it protects against training but does not prevent reading and citation by AI answer systems.

Upload sanitisation strips everything a browser could execute out of an uploaded file before it is stored. To the eye, the file is unchanged.

A vector database stores texts not as words but as numerical patterns of their meaning (embeddings). This lets an AI find the passage that fits by content — not the one with the same keywords. For AI visibility, clear meaning counts, not keyword density.

Vendor lock-in means switching to another provider becomes practically impossible or disproportionately expensive — because data, content or functionality is stuck inside a closed system. The antidote: open standards, exportable data and a domain you own.

Web accessibility means everyone can use a website — including people with impaired vision, hearing or motor skills. The legal basis in the EU is the European Accessibility Act (Directive 2019/882), applicable since 28 June 2025 and implemented nationally, in Germany as the BFSG.

Zero-UI describes interaction entirely without a visible interface: instead of typing or clicking, you ask by voice, gesture or context — and the environment answers. The interface disappears, the content remains.

Zero-click search means a search query ends without the user visiting a website — the answer comes directly from the search engine or AI. In early 2026, around 68% of US Google searches ended without a click; when an AI overview appears, the share rises above 80%.