SEO Terms: The 90 SEO and GEO Terms to Know in 2026

An SEO glossary gathers the definitions of SEO terms: crawling, indexing, backlink, title tag, search intent, click-through rate. This one gives equal weight to the vocabulary of GEO, born with answer engines such as ChatGPT, Perplexity and Google's AI Overviews, and flags the terms whose meaning has changed.

Here are the 90 terms I really use on client work, each defined in one to three sentences and grouped by category. No exhaustive list copied from a dictionary: only what helps you make decisions. To find a specific word, the filterable index just below does the search for you.

Search for an SEO term, filter by category, or show only the concepts that appeared with answer engines. Your selection copies in one click.

Essential SEO terms: the basics of search

These fourteen concepts are enough to understand how a search engine works and to read a report. Everything else flows from them.

SEO. Search Engine Optimization: everything you do to make a page appear in the unpaid results. The term dates back to the late 1990s and has outlived every marketing rival.

SEA. Search Engine Advertising: the paid ads shown in the results, led by Google Ads. The traffic stops when the budget stops, unlike SEO.

SEM. Search Engine Marketing: the umbrella term for SEO and SEA, in other words all your visibility in search engines, paid or not. It is often wrongly used as a synonym for SEA.

SERP. The search engine results page. In 2026 it no longer looks like a list of links: it stacks a generated answer, a block of questions, sometimes a map, videos, and finally the classic blue links.

Position zero. The featured snippet: a passage from a page shown in a box above the classic results, with a link to its source. It answers the question without the need to click.

People Also Ask. The block of expandable questions on the results page. The best free source for finding the real questions your content should cover.

Crawl. A bot visiting your pages to read their content. It is the first step, and the most often neglected: a page no bot reaches has no chance at all.

Googlebot. Google's crawler. It comes in a smartphone version and a desktop version, and it is the smartphone version that now reads every site.

Index. The search engine's database. Being crawled doesn't guarantee being indexed, and a page that isn't indexed doesn't exist, however good it is.

Algorithm. The set of ranking rules. Plural in practice: dozens of combined systems, several of them based on machine learning, none of them ever published in detail.

Core update. A major change to Google's ranking systems, announced several times a year. A drop in rankings right after one is not a penalty, but a reassessment of your pages against the others.

Query. What the person types or speaks. Not to be confused with the keyword, which is your own way of grouping similar queries.

Search intent. What that person really wants: to understand, to know how to do something, to compare before choosing, or to take action. It has become the central unit of work, ahead of the keyword.

Organic traffic. Visits from the unpaid results, excluding ads and direct visits. It is the number SEO tries to grow, and the one generated answers reduce on part of the queries.

The 3 types of SEO and the 4 pillars of SEO

The question comes up in every SEO glossary, with answers that vary. The most common split separates three types of SEO.

Technical SEO. What lets bots crawl, understand and index the site: structure, speed, tags, rendering. Its vocabulary is the subject of the next category.

On-page SEO. Everything you work on within the page itself: content, title tag, subheadings, internal links, matching the search intent.

Off-page SEO. Everything that happens outside your site and shapes its reputation: links received, brand mentions, customer reviews.

As for pillars, I keep three areas of work: technical, content and authority. Four-pillar versions usually add user experience or local SEO. The number matters little; what counts is not neglecting any of them.

Technical SEO vocabulary

This category scares people, and it doesn't deserve to. Sixteen terms cover most of what you'll need to discuss with a developer.

Title tag. The page title in the code. The search engine's first relevance signal, and the link text shown in the results. Sixty characters in practice; beyond that it gets truncated.

Meta description. The summary shown under the title. No direct effect on ranking for years, a real effect on click-through rate. One hundred and fifty-five characters.

Heading tags. The page headings and subheadings, from H1 to H6. A single H1 announcing the topic, H2s and H3s structuring the answer: the hierarchy helps the busy reader as much as the search engine.

Alt attribute. The alternative text of an image, read by screen readers and search engines. It describes the image; it is not there to stack keywords.

Canonical. The tag that names the reference version when several addresses show the same content. Essential as soon as a site generates URL parameters.

301 redirect. A permanent redirect from an old address to a new one. It passes on most of the value built up, unlike a 302, which signals a temporary change.

404 error. The server response when a page doesn't exist. A few 404s are normal; an internal link or a backlink leading to one is lost value, to be recovered with a redirect.

Duplicate content. The same text available at several addresses, on your site or elsewhere. Rarely a penalty, almost always a dilution: the search engine picks one version, not necessarily the one you prefer.

Robots.txt. The file that allows or blocks bots. Classic trap: it blocks crawling, not indexing. A blocked page can stay listed without its content.

Sitemap. The list of your pages in XML format, handed to the search engine to help it miss nothing. Useful on a large site, almost useless on a twenty-page site with good internal linking.

Core Web Vitals. Three loading experience metrics: how long the main content takes to display, responsiveness to the first interactions, and visual stability during loading. A real ranking factor, but a minor one compared with relevance.

Mobile-first indexing. Google crawls and indexes the mobile version of pages. Since July 5, 2024, every site has been crawled by the smartphone bot, and a site that can't be accessed on mobile is no longer indexed.

HTTPS. The encrypted version of the web protocol, shown by the padlock in the browser. Google made it a light ranking signal in 2014; today it is a prerequisite.

Structured data. Standardized markup, most often using the schema.org vocabulary, that explicitly declares what the page contains: an article, a product, a local business, a frequently asked question. Its importance has grown with answer engines, which read it without ambiguity.

Crawl budget. The number of pages a bot agrees to read on your site over a given period. A real issue beyond a few thousand URLs, a non-issue below that.

JavaScript rendering. The step where the search engine runs the page's code to see its final content. Costly, so sometimes delayed by several days. Content that only exists after rendering starts with a handicap.

For this category, my guide to the technical SEO audit explains the order in which to fix problems.

Content SEO vocabulary

This is the category where jargon has aged the most. Several widely used terms describe practices that no longer work.

Keyword. The phrase you target. A handy unit of work, but not enough: two identical queries can hide two different intents depending on context.

Long tail. Rare, specific queries. Each one weighs little, together they make up most of the volume, and competition there is much weaker. It is the natural playing field of a young site.

Semantic field. The set of words and concepts associated with a topic. A text that covers its subject well uses them naturally; forcing them in from a list of terms has the opposite effect.

Internal linking. The links between your own pages. They spread the value you receive and tell the search engine what matters on your site. The cheapest lever, and the most often forgotten.

Semantic cocoon. A tree structure: a pillar page covers a whole topic, child pages cover its branches, and internal links follow that hierarchy. The term comes from French SEO (cocon sémantique) and is close to what English speakers call topic clusters. It is useful, and often sold for more than it's worth.

Keyword cannibalization. Two pages on your site targeting the same search intent and depriving each other of rankings. You can spot it in Search Console when two URLs alternate on the same query.

EEAT. Experience, expertise, authoritativeness, trustworthiness. An evaluation framework used by Google's human quality raters, not a computed score. It describes what the search engine tries to approximate; it can't be measured anywhere.

Helpful content. The criterion Google has pushed since 2022: content written for a person, bringing something other pages don't have. Vague wording, concrete consequence: pages produced to fill an editorial calendar have lost their return.

Content refresh. Updating existing content rather than producing new content. The best return in a mature content plan, and the most neglected move.

Content brief. The order given to the writer: target intent, angle, outline, sources, acceptance criteria. This document decides the quality of the text, well before the writing starts.

Keyword density. The percentage of occurrences of a term in a text. An obsolete term: no search engine has worked that way for fifteen years, and aiming for it produces unreadable text. I mention it because it still circulates.

The full method is in my guide to SEO writing.

Link building vocabulary

Nine terms, and a necessary reality check on the scores sold by tools.

Backlink. A link received from another site. The historical currency of authority, still relevant, but its relative weight has dropped compared with content relevance.

Referring domain. A distinct site that links to yours at least once. More telling than the number of backlinks: a hundred links from a single site are not worth ten different sites.

Link building. The work of earning those links (often called netlinking in French SEO). It ranges from local partnerships to content so useful it gets cited spontaneously, with some practices along the way that I don't recommend.

Anchor text. The clickable text of the link. It announces the topic of the destination page. Anchor text that is too repetitive and too optimized is a sign of manipulation.

Nofollow, sponsored, ugc. Three attributes that flag, respectively, a link that isn't endorsed, a paid link, and a link posted by a visitor. They inform the search engine without forbidding it to follow the link.

Domain authority. A score invented by tool vendors (Domain Authority at Moz, Domain Rating at Ahrefs, Trust Flow at Majestic), not by Google. Useful for comparing two sites, absurd as a goal in itself. No search engine knows this number.

Black hat. All the techniques that break the search engines' published rules. Sometimes fast returns, lasting risk, and incompatible with a business that relies on its reputation.

White hat. The opposite: practices that follow the search engines' published guidelines. Slower, and the only sustainable choice for a company that lives on its reputation.

Manual action. The penalty applied by a Google reviewer when a site breaks its spam policies, artificial links included. It shows up in Search Console, under Manual actions, which sets it apart from a simple drop in rankings.

The method for earning links without risk is detailed in my link building strategy guide.

Local SEO vocabulary

Five terms to know as soon as a business has an address or a service area.

Google Business Profile. Google's business listing, formerly Google My Business. The ID card of your business, and the centerpiece of local SEO.

Local pack. The block of three businesses with a map, shown above the classic results for location-based queries. It captures most of the leads.

NAP. Name, address, phone. Writing them exactly the same way everywhere on the web confirms to the search engine that the business exists and is located where it says.

Local citation. A mention of your business details on a third-party site, with or without a link. Its value lies in consistency, not volume.

Service area. The area your customers really come from. It drives your pages, your priorities and the area you declare that you serve. The subject is covered in full in my local SEO guide.

Answer engine vocabulary: what's really new

This category didn't exist in 2022 and now takes up half of my conversations with clients. Thirteen terms, only three of which are essential.

GEO. Generative Engine Optimization: everything that makes content reusable by engines that write the answer instead of listing links. The counterpart of SEO for these systems, and the subject of my GEO SEO guide.

AEO. Answer Engine Optimization. In everyday use, a marketing synonym for GEO. Some people distinguish the two; the distinction changes nothing about what you need to do. My AEO guide shows how to write for these answers.

Answer engine. A system that writes an answer instead of presenting a list: ChatGPT, Perplexity, Gemini, Claude, and Google's AI Overviews.

AI Overviews. Google's name for these summaries, shown above the results with a few cited sources. Rolled out from May 2024, they have been available in France since July 22, 2026.

AI Mode. Google's conversational search mode, launched in France on the same day as AI Overviews: the person talks with the search engine, which writes a detailed answer drawn from several searches. I explain how it works in my article on Google AI Mode.

Citability. How easily a passage can be reused as is in a generated answer. It rests on three qualities: a self-contained statement that makes sense out of context, a verifiable data point, and clear attribution. It is the central concept of GEO.

RAG. Retrieval Augmented Generation. The process by which a model fetches documents before answering, instead of relying on its memory alone. This is exactly where your pages come into play, or don't.

Grounding. Anchoring an answer in verifiable sources. An engine that uses grounding cites its sources, which gives you a chance to be among them.

Hallucination. A statement made up by a model, delivered with the same confidence as a verified fact. In practice: what these systems say about your business can be wrong, and you need to check it regularly.

LLM. Large Language Model. The statistical engine behind these systems. It predicts plausible text, it doesn't look up a database of facts, and that nuance explains most of its mistakes.

llms.txt. A file placed at the root of a site to tell language models which content to read, proposed by Jeremy Howard in September 2024. Adoption remains partial, and its effect on generated answers is still unproven.

Context window. The amount of text a model keeps in view during an exchange. It limits how much of your content it can read at once.

Citation share. The proportion of generated answers in which your brand appears, across a defined set of questions. It is the visibility metric specific to GEO, and it has to be tracked by hand for lack of a complete tool.

AI vocabulary, from a business point of view

Six terms have come up in every conversation for the past two years. Knowing them keeps you from being sold artificial intelligence where there is none.

Generative AI. The family of systems that produce text, images or code rather than classify or predict a value. It is this kind of artificial intelligence that has upended online search, not the kind that has been sorting your email for ten years.

Natural language processing. The discipline that teaches a machine to handle human language. All the processing a search engine does to understand a query falls under it, well before generative models arrived.

Token. The unit a model splits text into, often a fragment of a word rather than a whole word. API pricing is counted in tokens, and so is the bill.

Prompt. The instruction given to the model. An overrated term: what changes the result is almost always the context provided, not the wording of the instruction.

Agent. A model allowed to chain actions in an environment, call tools and loop until it reaches a result. The word is wrongly used for any slightly clever automation.

Fine-tuning. Retraining a model on your own data. Expensive, rarely necessary: in most cases, handling the context well is enough.

User experience, page speed and navigation

Four terms that describe what the person experiences on your page, and that weigh indirectly on rankings.

User experience. The quality of the journey from the user's point of view: find, understand, act. Google doesn't measure it directly, but several of its signals depend on it.

Page speed. The time before the main content is visible. On mobile, every extra second costs a measurable share of visitors. Speed is a minor ranking factor and a major conversion factor.

Navigation. The structure that lets a user reach any page in three clicks. It serves the visitor as much as the bot, which follows the same paths.

Mobile usability. Real adaptation to small screens, not just a layout that shrinks. Tap targets big enough, text readable without zooming, short forms.

Social signals: what the term really covers

Social signals. Shares, mentions and interactions on social networks. They are not a direct ranking factor, and it bears repeating given how widely the idea circulates.

Social media optimization. Optimizing your posts on these platforms. Useful work in itself, separate from SEO, with its own rules: content that works on Instagram looks nothing like a page that ranks.

What connects the two worlds is indirect but real: content that spreads on social media generates searches for your brand, mentions and sometimes links. That is how, and the only way, social media helps your visibility in search engines. An Instagram story and a service page are not worked the same way, and trying to apply the optimization of one to the other gets you nowhere.

How to measure SEO performance: the vocabulary

Seven terms, and two interpretation traps I see every week.

Impressions. The number of times your page appeared in the results. The denominator of everything else, and the first number to check when traffic drops.

Click-through rate. Clicks divided by impressions. The best signal for judging a title and a meta description, since it isolates what happens on the results page.

Average position. First trap. As soon as a page shows up for dozens of queries, the average mixes unrelated situations. A worsening average position can mean you appear for more queries, so everything is fine.

Bounce rate. Second trap. It is the share of visits with no second action. A visitor who finds the answer right away and leaves satisfied produces exactly the same number as a disappointed one.

Google Search Console. Google's free tool: your queries, your clicks, your impressions, your indexing problems. Non-negotiable, and enough for the vast majority of sites.

Google Analytics. Measures behavior once people are on the site: page views, journeys, conversions. It complements Search Console, which stops at the click.

Conversion. The action you expect: a form, a call, a purchase. The only metric a business really cares about, and the first one forgotten in front of traffic curves.

Which SEO tools should you use in 2026?

Tool vocabulary deserves a reality check, because it is often mistaken for search engine vocabulary.

Two free tools cover the essentials: Search Console for what happens in the results, and Google Analytics or a privacy-friendly equivalent for what happens on the site. A site with fewer than a hundred pages needs nothing else to get started.

The paid suites on the market add three things: search volume estimates, a view of the links your competitors receive, and automated rank tracking. Their numbers are estimates, never measurements. A search volume announced at a thousand a month can be worth three hundred or three thousand.

So-called semantic tools, which I use, compare your text with pages that already rank and flag missing vocabulary. Useful to make sure you forget nothing about a topic, dangerous if the text is written for the tool rather than for the reader.

Finally, tools that track visibility in answer engines are the newest. They ask ChatGPT or Perplexity questions and record who gets cited. The approach is sound, the results vary a lot from one run to the next: treat them as orders of magnitude.

What are the SEO trends for 2026?

Three shifts, each with its own vocabulary.

Zero-click search. A growing share of queries find their answer directly on the results page. Consequence: organic traffic can drop while visibility rises, which makes tracking impressions more important than ever.

SEO and GEO converging. The signals that get a page cited by an answer engine largely overlap with those that make it rank. There aren't two jobs, there is one job with an extra requirement: being reusable. I developed this point in my GEO vs SEO comparison.

AI-assisted production. Writing with a language model has become common, and the vocabulary follows: prompt, context, agent, processing pipeline. What separates useful text from generic text is not the tool but the material you feed it.

Terms to stop using

An honest glossary also says what is dead. Five expressions to drop from your vocabulary, and the mistake each one leads to.

These vocabulary mistakes are not trivial: every obsolete term carries a reasoning error, and that error ends up producing a bad decision. That is why I always start an audit by clarifying the words being used.

Keyword density. See above. No search engine works that way, and aiming for a percentage is the most visible mistake in a text: you can hear it when you read it aloud.

PageRank. Google's historical algorithm still exists internally, but the public score disappeared in 2016. Anyone quoting "your PageRank" is talking about a tool score.

First page of Google. A fuzzy notion in 2026: the results page varies by device, location and personalization, and it often starts with a generated answer. Talk about specific queries, not pages.

Search engine submission. Useless for twenty years. A link from a known page is enough to be discovered.

SEO is dead. Not a term, a perennial headline. It resurfaces with every interface change since 2005, and it has always been wrong. What dies are techniques, never the need to be found.

How to use this SEO glossary

Three uses, depending on your situation.

You're just starting. Read the "basics" and "content" categories and ignore the rest for now. The basic terms are enough to have a useful conversation with a provider.

You're managing a provider. Start by asking for an SEO audit, technical and editorial. The "measurement" and "links" categories will be the most useful to you: they are the ones where vocabulary is most often used to impress. A consultant who talks about domain authority as a goal is aiming at the wrong target.

You already work on the subject. The answer engine category is the one that moves, and the only one where the vocabulary hasn't settled yet. Expect some of these terms to change names within two years, without the reality they describe changing much.

How to keep this vocabulary up to date

A glossary goes stale. Here is how I keep mine current without spending my days on it.

Follow primary sources. Google's documentation, announcements from model providers, official help pages. That is where terms are born, before they get distorted. One hour a month is enough, and the information there is more reliable than in any news roundup. Subscribe to two or three serious sources, not fifteen.

Be wary of agency buzzwords. Every year brings its acronyms invented to sell a service. The test takes a quick check: if the term appears in no official documentation and no independent analysis uses it, it describes nothing new. It is only trying to claim a slice of your budget.

Check that your own usage is consistent. Using three different words for the same concept in one document is the most common mistake. Consistent vocabulary helps your readers as much as the automated processing of your pages: a search engine that has to guess that two terms mean the same thing does a poorer job.

Get trained once a year. A short training session resets your ideas more effectively than twelve months of passive monitoring. It is a good part of my work: I run sessions inside companies so that teams become autonomous, rather than relying on me for everything. Well-timed training saves more time than one more audit.

One last piece of advice, from one user to another: write down the terms you don't understand when you come across them, instead of letting them slide. That list quickly becomes your own glossary, and it is the one that will serve you best. It is also how mine started, between two audit projects and a subscriber who asked me the question by email.

This SEO glossary is reviewed every quarter. The definitions that change are the ones at the bottom of the page, never the ones at the top: the basics of search engine optimization haven't changed in fifteen years, and there is no reason they will change tomorrow.

Written by

Anthony Courtin

SEO, GEO & AI automation consultant

I help SMEs and independent professionals with their visibility (Google, ChatGPT, Perplexity) and with the real use of AI at work. I run sessions inside companies to train teams and rethink processes. Everything I write here, I use on client work.

Follow me on LinkedIn My YouTube channel (in French)