GLOSSARY

What the terms you keep hearing mean

Thirty-four terms from SEO, GEO, speed and security. Each defined exactly as far as a decision needs — with the caveat where one is real.

SEO

Indexing: Adding a page to a search engine's database. Crawling and indexing are not the same: Google can read a page and still leave it out if it decides the page adds nothing new.
Crawl budget: How many pages a search bot fetches per visit. The limit only bites on large sites — below 2 000 pages it barely exists, so it is usually not your problem.
Canonical URL: A `rel="canonical"` link naming which version of duplicate content is the main one. It is a recommendation, not an order: Google may pick a different one when signals disagree.
robots.txt: A file at the site root telling bots what not to fetch. Important: `Disallow` means "do not read", not "do not index" — a blocked page can still appear in results without a description.
sitemap.xml: A list of URLs with the date each last changed. It guarantees no indexing — it only states the pages exist. Identical `lastmod` dates destroy trust in the whole file.
Title: The headline Google shows in the results list. The real limit is pixels (around 580 on desktop), which is roughly 50–60 characters. Too short and Google often rewrites it.
Meta description: The text under the headline in results. Not a ranking factor, but it drives a share of clicks. Around 120–160 characters; without one Google lifts a suitable sentence from the page.
Internal link: A link between your own pages. The one ranking signal you control completely. A page with no internal links pointing at it often stays unindexed even when it sits in the sitemap.
Cannibalisation: When two of your own pages compete for the same question. Google picks one, often not the one you want — and both rank lower than a single strong page would.
301 redirect: A permanent redirect that passes signals to the new address. The difference from 302 matters: 302 means "temporary", an explicit instruction not to consolidate signals.
hreflang: A declaration that a page has a version in another language. Google accepts only RECIPROCAL pairs: A must point to B and B back to A. A one-way declaration is ignored silently.
E-E-A-T: Experience, expertise, authoritativeness and trust — criteria from Google's quality rater guidelines. Not a ranking factor; no E-E-A-T score exists, because it is an evaluation frame rather than a signal.

SEO →

GEO and AI

GEO: Generative Engine Optimization — adapting a site so AI knows, understands and mentions it in an answer. It differs from SEO in what it measures: citation rather than position in a list.
AEO: Answer Engine Optimization — shaping a page so a direct answer can be lifted from it: a Google snippet or a paragraph in an AI answer. Structure decides it, not length.
AI SEO: The umbrella name for the whole field, covering both GEO and AEO. Technically it has no separate task list — it is the same work measured by a different question.
llms.txt: A file at the site root describing to a model what the site is and where its key pages are. Not a ranking factor and not an official standard, but it reduces the work for a reading agent.
AI Overviews: A Google-generated summary above the results citing a few sources. The cited list does not match the top positions: a tenth-place page can be quoted while the first is not.
Passage: A chunk of text a model takes separately from the whole document. Headings set the boundaries, so a block without them is one passage no precise answer can come from.
Citation: When an AI answer draws on your page and names it as a source. Different from a click: a citation gives visibility even when nobody opens the link.
Structured data: A JSON-LD block repeating page content in a form machines read. Not a ranking factor, but it unlocks rich results and turns an inference into a declared fact.
GPTBot: OpenAI's crawler collecting content for model training. It differs from `OAI-SearchBot` in purpose: the latter feeds ChatGPT search, so blocking it closes the citation route.
Google-Extended: A `robots.txt` flag controlling use of your content for Gemini training. It affects neither ordinary indexing nor AI Overviews — blocking it to disappear from AI Overviews does not work.
Markdown negotiation: The same URL serves HTML to a browser and plain markdown to an agent sending `Accept: text/markdown`. The agent spends several times fewer tokens for identical content.
Zero-click: A search where the answer arrives without any link being clicked. In Search Console it looks like steady impressions with falling clicks.

GEO and AI →

Speed

Core Web Vitals: Three Google metrics measuring page experience: LCP, CLS and INP. Judged separately — one combined score never says which of the three is actually failing. The thresholds: LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1.
LCP: Largest Contentful Paint — how long the largest visible element takes to appear, usually the first image or the heading. The closest number to what a person calls "slow". The threshold is 2.5 seconds for 75% of visits.
CLS: Cumulative Layout Shift — how much content moves while the page is still loading. The most common cause is one thing: an image with no `width` and `height` shoving text down.
INP: Interaction to Next Paint — the response to a click. It replaced the old FID and is stricter: it measures not the first interaction but the worst one of the visit. The threshold is 200 milliseconds.
PageSpeed Insights: Google's tool measuring speed in lab conditions and showing real visitor data where enough exists. The score varies between runs, so watch the trend rather than one number.
Rendering: Running JavaScript, after which content appears on the page. Several bots run no scripts, so a page that looks complete in a browser can be empty to them.

Speed →

Security

HTTPS: An encrypted connection between browser and server. A confirmed Google ranking signal, but the simpler consequence matters more: an expired certificate closes the site to everyone at once.
HSTS: A header telling browsers to speak to this domain only over HTTPS. It covers the first unencrypted request that an ordinary `http → https` redirect still allows through.
CSP: Content Security Policy — a list of where the page may load scripts, styles and frames from. The main defence against XSS once foreign code has already reached the HTML.
X-Content-Type-Options: A header with one value, `nosniff`. It stops the browser guessing a file's type from its contents — without it an uploaded image can be executed as a script.
Mixed content: When an HTTPS page loads some resources over HTTP. Browsers block or warn about them, so the practical result is usually a broken layout rather than just a warning.
Referrer-Policy: Controls how much of your URL is sent when leaving for another site. Without it the full address with its parameters travels to a foreign server, including parts nobody there needs.