AI search terms, in plain English

AI search

Finding information by asking an AI assistant a question in plain language and getting a written answer, instead of scrolling a list of links. It covers ChatGPT, Claude, Gemini, Perplexity and Google’s AI features.

Why it matters: More customers now get a single AI answer that names one or two businesses — if you’re not in it, you may never be seen.

Google AI Overviews

The AI-written summary box Google shows at the top of some results pages. It answers the question in a few sentences with a handful of cited links, then the normal results follow below.

Why it matters: Being one of the cited links puts you at the very top of the page, above the traditional results.

Large Language Model (LLM)

The kind of AI that powers ChatGPT, Claude and Gemini. It’s trained on huge amounts of text and predicts likely wording to answer a question — which is why it sounds human but can occasionally be wrong.

Why it matters: LLMs are now the engines deciding which businesses get named when someone asks for a recommendation.

Prompt

The question or instruction someone types into an AI assistant — for example, “best emergency plumber in Leeds”. It’s the AI-era version of a search keyword.

Why it matters: Knowing the exact questions your customers ask tells you which answers you need to show up in.

Retrieval-Augmented Generation (RAG)

A method that lets an AI look things up before it answers, fetching relevant, up-to-date pages instead of relying only on what it memorised in training.

Why it matters: RAG is how assistants pull in live web pages — so clear, well-structured pages on your site can become the source an assistant quotes.

Grounding

Tying an AI’s answer to real, verifiable sources (often live search results) rather than letting it answer from memory alone. Grounded answers usually come with citations.

Why it matters: When an assistant is grounded in the live web, your published pages can be the evidence it cites about your business.

AI crawlers (and robots.txt)

The automated programs AI companies use to read web pages. Some collect content to train models (GPTBot, ClaudeBot); others fetch live pages to answer a question now (OAI-SearchBot, Claude-SearchBot, Google-Extended, PerplexityBot). You control access with a small file called robots.txt at your site’s root.

Why it matters: Blocking the wrong bot can quietly make you invisible to AI — the safe move is to let the search/answer bots in, so you can be cited.

llms.txt

A proposed plain-text file at your site’s root that gives AI a tidy map of your key pages. It’s a community idea from 2024, not an official standard.

Why it matters: It’s low-cost to add, but as of 2026 there’s little evidence it changes how often you’re cited — treat it as tidy housekeeping, not a magic switch.

Structured data (Schema.org)

Hidden, standardised labels added to a page that spell out its facts to machines — your opening hours, address, prices or reviews. Schema.org is the shared vocabulary; JSON-LD is the usual format.

Why it matters: It hands AI clean, unambiguous facts about your business, making it easier for an assistant to quote the right details correctly. (Helpful, not required.)

Knowledge graph

A giant database of real-world “things” — people, places, businesses — and the verified links between them. Google’s runs on the idea of “things, not strings”: understanding meaning, not just matching text.

Why it matters: Being a recognised “thing” in these graphs gives AI a trusted, structured record to draw on when deciding whether to recommend you.

Entity (and entity SEO)

An entity is a specific real-world thing — your business — that AI treats as a distinct, identifiable object rather than just a word. Entity SEO means describing your business clearly and consistently everywhere so machines recognise it as one trusted thing.

Why it matters: If AI can confidently identify your business, it’s far more likely to recommend you with confidence.

Semantic search

Searching by meaning rather than exact keywords, so a system can match a customer’s question to your content even when the words are different.

Why it matters: It’s how AI connects a customer’s plain-language question to what you actually offer — so writing the way customers speak matters more than stuffing keywords.

Citation signals

The characteristics that make an AI more likely to name you as a source: a clear, consistent business identity, depth of coverage, machine-readable structure, third-party mentions (reviews, directories), and specific, checkable facts.

Why it matters: Citations are the AI-era equivalent of ranking first — being named is how customers discover and trust you. (How AI picks sources is still emerging, so treat any “do X and you’ll be cited” claim as an informed bet, not a promise.)

Generative Engine Optimisation (GEO)

Structuring your content and online presence so AI systems cite and recommend you in the answers they generate. Where SEO chased ranking positions, GEO is about influencing what the AI says.

Why it matters: It’s the core discipline for getting a small business named inside AI answers rather than lost below them.

Answer Engine Optimisation (AEO)

Optimising content to be the direct answer in AI-powered features — AI Overviews, Perplexity, voice assistants — by making it easy to lift and present as a clean response. In practice, many people use “AEO” and “GEO” to mean the same thing.

Why it matters: It focuses on winning the short, direct answers customers get without clicking — increasingly the first (and sometimes only) impression.

AI visibility (share of voice)

How often — and how prominently — your business shows up in AI answers compared with competitors, across a set of customer questions. Visibility is whether you appear at all; share of voice is how big your slice is versus rivals.

Why it matters: It’s the scoreboard for AI search. (There’s no industry-standard way to measure it yet, so numbers vary by tool — read them as a trend, not an exact figure.)

E-E-A-T

Google’s framework for judging content quality: real first-hand Experience, genuine Expertise, recognised Authority, and Trustworthiness. Trust is the most important of the four.

Why it matters: The same trust signals that satisfy Google — real credentials, reviews, accurate information — are what make AI confident enough to recommend you.

Hallucination

When an AI states something false but presents it confidently as fact — inventing a phone number, a service you don’t offer, or a competitor that doesn’t exist. It happens because the model predicts plausible wording.

Why it matters: An assistant can invent wrong details about you — giving AI clear, correct, easy-to-find facts reduces the chance it makes something up.

Zero-click search

When the AI answer fully satisfies the searcher, so they never click through to any website.

Why it matters: It’s why being cited in the answer — not just ranking somewhere below it — now matters so much.

A note on honesty: AI search is moving fast. The definitions above are settled, but the tactics people attach to them — llms.txt, specific schema, “citation signals” — are still being tested, and much of the advice out there is a correlation, not a proven cause. We’ll update this page as the evidence firms up.