Glossary
Every term is defined in one sentence, then points at the guide that develops it and cites its sources. The definitions deliberately avoid the trade's habit of explaining one word by another word of the trade: if a definition requires knowing the vocabulary, it has failed.
How this glossary is built
Three rules, and they explain both what is here and what is not.
- A definition fits in one sentence. If it needs three, the term was not a term but a subject — and it then belongs in a guide, not here.
- No word of the trade is explained by another word of the trade. That is the most widespread defect of glossaries in this field: a definition assuming the vocabulary is already known only serves those who did not need it.
- Every term points at the guide that develops it, and it only enters here once that guide exists. A term with no guide is a dead end: the reader who wants to go further has nowhere to go.
The consequence of the third rule is visible: this glossary is short. It grows when the site grows, never before. A glossary defining words the site does not handle is a copied glossary, and that shows in two clicks — the definitions are interchangeable from one site to the next.
The terms
- 503 status code — also called service unavailable
A response announcing a temporarily unavailable service. It slows crawling and preserves addresses in the index at first, dropping them if the outage lasts. The robots.txt file must never return it: that would block all crawling.
Developed in Pausing a website
- A/B test — also called multivariate test
A comparison of two versions of a page with real visitors. Nothing forbids it on the search side: what matters is the means used, the original address staying canonical, the redirect temporary, and the setup dismantled once concluded.
Developed in A/B testing and search
- Alternative text — also called alt attribute, alt
Description of an image intended for whoever cannot see it, human or machine. Presented as the most important element for describing an image, it is interpreted together with the content of the page around it.
Developed in Images and image search
- Anchor text — also called anchor
Visible text of a link. The documentation wants it descriptive, concise and relevant to the page you leave as well as the page you reach: it advises against generic phrases and against piling up keywords.
Developed in Internal linking
- Automatically generated content — also called AI-generated content
Text produced by a program rather than written. The published rules do not target the process but the intention: what is named is mass production aimed at occupying results, whatever the tool used. Machine-assisted text that has been reread does not fall under that description.
Developed in AI-generated content: what is known and what is not
- Average position
Average, over the impressions actually recorded, of the rank of a site's topmost result. It also improves when a site loses badly placed impressions: it is never read without its volume.
Developed in Reading Search Console honestly
- Breadcrumb trail — also called breadcrumbs
A sequence of links stating a page's position in the site hierarchy, climbed one level at a time. The engine uses its markup to categorise the page's information; one page may declare several trails when it is reachable by several routes.
Developed in Breadcrumbs and multiple paths
- Business profile — also called local listing
Record describing a shop or a service to an engine: name, activity, address, opening hours, contact details. It governs display in results tied to a place, and concerns only businesses that receive customers or travel to them.
Developed in Business profiles and local search
- Canonical address — also called canonical, rel canonical
Address an engine keeps to represent a group of pages judged equivalent. The site can signal it by several means, but the final decision is not the site's: these are signals, not instructions.
Developed in Canonicals, redirects and duplication
- Click-through rate — also called CTR
Share of a page's displays that lead to a click. It is always read at comparable position: the same page gains or loses clicks according to its rank, the query and what surrounds it in the results page, without its title having changed.
Developed in Title, meta description and the click
- Cloaking
Presenting different content to users and search engines with the intent to manipulate rankings and mislead visitors. Intent is part of the published definition: a subscription wall declared through the prescribed markup does not qualify.
Developed in Paywalled content and sampling
- Crawl budget — also called crawl budget
What an engine devotes to downloading a site's pages. It follows from the server's capacity and the engine's interest. The documentation reserves the subject for very large sites and for those that change very fast.
Developed in Crawl budget
- Crawling — also called crawl
Stage during which an engine downloads a page's content. It is distinct from indexing: a page can be crawled without being kept, and discovered without being crawled.
Developed in How a search engine works
- Customer review
Assessment published by a person about a business. The published rules allow soliciting them, and forbid buying them, writing them for yourself, or asking only satisfied customers. An inducement offered in exchange for a review is named among the forbidden practices.
Developed in Customer reviews: what the rules allow
- Direct answer — also called answer-first
Paragraph placed under a page's heading, answering the question asked before any list and any image. It has to stay understandable out of context, since a snippet may pick it up on its own.
Developed in Writing a page that answers
- Discover
A feed offering content according to a person's interests, with no query typed. An indexed page meeting the feed's policies is automatically eligible, with no special tag, and eligibility is not a guarantee of appearing.
Developed in Discover and traffic you do not choose
- Discovery — also called detection
Moment when an engine learns that an address exists, before any crawling. It happens through a link, a sitemap or a submission. A discovered address is neither crawled nor indexed for all that: these are three distinct stages, and nothing guarantees moving from one to the next.
Developed in Where to start on a new site
- Discussion markup — also called DiscussionForumPosting
Structured data describing a forum thread where people share first-hand perspectives. Replies nest the way the thread organises them; a page listing posts must name none of them as its main entity.
Developed in Marking up a forum and its discussions
- E-E-A-T — also called EEAT
Initials used in the search quality rater guidelines for four notions: experience, expertise, authoritativeness and trust. The document makes them an observational vocabulary aimed at human readers, and names trust the most important of the four. No scale or numeric value is described there.
Developed in The search quality rater guidelines
- Embedded metadata — also called IPTC
Information written into the image file itself — creator, licence, credit — that follows it from page to page. It is written once per image, but the page's structured data takes precedence over it in case of conflict.
Developed in Image licence metadata
- Featured snippet — also called position zero
A result with its order reversed, the descriptive snippet coming before the title. No markup lets you opt in, the engine deciding alone; two settings let you opt out. A click leads straight to the quoted passage inside the page.
Developed in Featured snippets
- Flexible sampling — also called metering, lead-in
How a paywalled site lets part of its content be seen: a monthly quota of articles, or the opening sentences before the wall. The engine's vendor publishes markers from experiments while stating that no single value works across businesses.
Developed in Paywalled content and sampling
- Google-Extended — also called AI token
A robots.txt token with no user agent of its own, deciding whether crawled content may train Gemini models and ground them at prompt time. It has no effect on presence in Search and is not a ranking signal.
Developed in AI crawlers and Google-Extended
- Hreflang — also called language annotation, localized versions
Annotation by which a page declares its versions in other languages or regions, itself included. The declarations of all versions of the same page must be identical to one another, failing which the arrangement is incomplete.
Developed in Multilingual sites and hreflang
- Impression
Event recorded when a site's page is offered in search results. It is the measure of the demand a page is attached to, independently of whether it was chosen.
Developed in Reading Search Console honestly
- Indexing
Stage during which an engine analyses a downloaded page and decides whether to keep it. It is never guaranteed: not every page processed enters the index.
Developed in Why a page is not indexed
- Infinite scroll — also called lazy loading
A display that loads new chunks as the reader moves down. It is reachable on one condition: each chunk carries its own persistent address with stable content, and those addresses link to each other. The engine does not scroll the page.
Developed in Infinite scroll and lazy loading
- Interstitial — also called pop-up, dialog
An element that covers the page before or after the content expected. Published recommendations target it only when it prevents access to the main content; legally mandatory interstitials are exempted, provided they overlay the page rather than replace it.
Developed in Ads that cover the main content
- Link profile — also called backlinks, inbound links
Set of links pointing at a site from other sites. No tool knows all of them: each reports what its own crawler saw, on the date it saw it. Two tools therefore give different inventories of the same site without either being wrong.
Developed in Analysing a link profile
- Link qualification attribute — also called nofollow, sponsored, ugc
Note added to a link to declare its nature: sponsored content, a third party's contribution, or a link you do not wish to endorse. It neither hides nor removes the link; it describes it, which is what the published rules ask for a link obtained in exchange for payment.
Developed in Earning links without buying them
- Loading speed — also called performance, load time
Time between a page being requested and the moment it becomes usable. It is measured by several distinct indicators, which do not say the same thing: the appearance of the first content, the stability of the display and the response to actions are three separate problems.
Developed in Speed and loading experience
- Main content — also called MC
The part of a page that directly serves its purpose, as opposed to supplementary content and advertising. The rater guidelines make it the centre of their grid: a page whose main content is deliberately obscured is judged untrustworthy, whatever its other qualities.
Developed in Ads that cover the main content
- Manual action — also called manual penalty
Decision taken after a human reviewer has found that a site does not comply with the published anti-spam rules. It is notified with its reason: absent that notice, there is no manual action.
Developed in Link buying, PBNs and penalties: what is documented
- Meta description — also called description
Short text declared in a page's code to offer a summary of it in a results page. It is a proposal and not a guaranteed display: an engine may prefer a snippet drawn from the body of the page, particularly when the query bears on a passage the description does not cover.
Developed in Title, meta description and the click
- Migration — also called address change, rebuild
Change of a site's addresses, with or without a change of content. Since what an engine has accumulated is attached to addresses, every old address must say where it went, permanently and lastingly.
Developed in Migrating a site without losing its addresses
- Nosnippet rule — also called data-nosnippet, max-snippet
A directive removing a page's text snippet from all forms of results, AI-based answers included, and also preventing the content from serving as a direct input to them. The page stays indexed: what is restricted is display and reuse, not presence.
Developed in Controlling what is shown of your page
- Profile page — also called author page
A page describing the person or organisation posting in a community: name, username, presences elsewhere, activity. Article and discussion markup point to it. It makes an existing identity legible; it manufactures no authority.
Developed in Profile pages and author identity
- Query — also called keyword
Phrasing actually typed by a person into an engine. To be distinguished from the keyword, which is a category built afterwards by whoever analyses. Two queries close in wording can answer opposite needs, and it is the query that counts.
Developed in Keyword research
- Rank tracking — also called position tracking
Repeated recording of a page's rank for a given query. The tool queries the engine from a place, a device and a moment it chooses: it measures its own probe, not what visitors see. It is the trend over time that is legible, never an isolated reading.
Developed in What a rank tracker actually measures
- Recrawl request — also called IndexNow
Flagging a new or changed address, through the inspection tool or a sitemap. It guarantees neither speed nor indexing, and re-requesting the same address does not get it crawled faster. A quota limits individual submissions.
Developed in Asking for a recrawl
- Removals tool — also called removal request
A console feature that takes an address out of the results within a day. The request lasts about six months: a reprieve meant to cover the time needed to put a permanent measure in place and have it crawled.
Developed in Removing a page from the results
- Rendering — also called render
Execution of a page's scripts by a headless browser, before the result is indexed. It is deferred: a page can wait in the queue a few seconds, sometimes a great deal longer.
Developed in JavaScript and engine-side rendering
- Republishing
Putting content back online at a new address or a new date without its substance having changed. It scatters the links and the history of the original page, where an update keeps them. Changing the displayed date without touching the text is the same gesture.
Developed in Updating rather than republishing
- Reverse DNS lookup — also called reverse DNS
The operation returning the name attached to an IP address. It is filled in by whoever holds the address, so it proves nothing on its own: the published method requires resolving that name back into an address and finding the original one.
Developed in Verifying that a crawler is Googlebot
- Review snippet — also called star rating
Stars shown beside a result, earned through review markup. It requires a rating visible on the page and attached to a specific item; an entity controlling the reviews about itself is not eligible, embedded third-party widget included.
Developed in Stars in the results
- robots.txt — also called robots file
Text file placed at the root of a site, telling crawlers which paths they are asked not to walk. It bears on crawling, never on indexing or on confidentiality: an address it blocks stays public, and can appear in an index without having been read.
Developed in How robots.txt is interpreted
- Scaled content abuse — also called mass-produced content
Generating many pages primarily to benefit the site rather than help its visitors, no matter how they are produced. The published definition targets not the tool used but the ratio between the number of pages and the value they bring.
Developed in The lowest quality pages
- Search intent — also called intent
Kind of answer a person expects when phrasing their query. The split into categories the profession uses is an analysis tool, not a classification published by any engine.
Developed in Search intent
- Search volume — also called volume
Number of searches attributed to a query by a tool. It is an estimate, produced by extrapolation from samples, and not a reading: two tools announce different values for the same query, and none publishes the measurement that would settle between them.
Developed in Keyword research
- Seasonality
Variation in demand tied to the calendar or a trend, which draws the same shape on a traffic curve as a technical problem. The documentation recommends displaying sixteen months before concluding, precisely so it can be recognised.
Developed in Analysing a traffic drop
- Site name
The label shown beside a result to name its source, distinct from each page's title. It is generated automatically from the home page and outside references; markup expresses a preference, never a command.
Developed in The site name in the results
- Sitemap — also called XML sitemap
File listing a site's addresses in order to point them out to an engine. It guarantees neither crawling nor indexing, and two of its optional fields, priority and change frequency, are declared ignored by the engine's vendor.
Developed in What sitemaps actually do
- Soft 404 — also called soft 404 error
An address that answers a success code while displaying an empty page or an error message. The engine sees the contradiction between code and content and reports it in Search Console. Redirecting a removed page with no equivalent to the homepage produces one.
Developed in Error pages and the soft 404
- Structured data — also called markup, JSON-LD, schema.org
Description of a page's content in a format a program can read. It may give rise to a rich display, but no markup earns one by right, even when correctly written.
Developed in Structured data and rich results
- User-generated content — also called UGC
Content published on a site by people outside its team: forum posts, answers, comments, reviews. The rater guidelines treat it as main content in its own right, and a page can reach the highest ratings while its authors are identified only by a username.
Developed in Forums and Q&A pages
- Watch page
A page whose main purpose is to show a single video. Eligibility for video features assumes one, and the documentation adds a rarely quoted condition: the page must already be performing well in Search before its video can be considered for indexing.
Developed in Video in search
What you will not find here
No software acronym, no algorithm update name, no notion this site would have nothing sourced to say about. Update names in particular are excluded everywhere: the timelines in circulation are copied from one another and regularly wrong on dates, which makes them the most recognisable marker of second-hand content.
A missing term is therefore not an oversight to fill in on principle. If one you need is missing and the site already handles it, on the other hand, saying so is useful: it probably means a guide exists whose vocabulary was never brought back here.
Content that pushes a visitor into a dangerous action, either by imitating a trusted entity or by obtaining what one would only do for such an entity. Embedded through an advertisement, it is a policy violation for the host page and triggers a browser warning.
Developed in Security issues and browser warnings