Home / Help center / Glossary
Plain English definitions for technical SEO and AI visibility
Use this reference while reading audits, search reports, backlink reviews, and citation tracking. Each term explains what it means and why the evidence matters.
How to use this glossary
SEO Terms rarely live alone. A URL can return 200, appear in a XML sitemap, and still remain outside the index since its canonical points point elsewhere. A ranking may improve while clicks decrease because an AI overview or featured snippet changed the outcomes page above it. The useful question is not solely what a term means, but also what it changes.
Each entry below gives you the clear meaning, the reason why the term is important, and the practical note you should consider. Use the search box for a report label, metric, or issue you saw. The filter searches the full declarations, so a search for “duplicate” also finds canonicals, URL parameters, and faceted navigation.
Crawling and discovery
Before a search engine can evaluate a page, it must find and retrieve it. These terms explain how crawlers move through a website, what they get, and where their time can be wasted.
Health Score
A health score is a summary of the technical issues found during a crawl, from 0 to 100. It groups many types of issues into a trend line by weighting their severity and the number of pages affected. It's useful for comparing over time, but it's not a Google metric and it's not a direct prediction of rankings.
In practice, you read the rating next to the crawl area and the number of problems. A score may increase because real problems were fixed or since fewer pages were crawled. Swiss Knife SEO compares the calculation to the crawled set of pages, making the number harder to shrink by reducing the sample.
Crawl
A crawl is a recorded passage through a site. The crawler begins with one or additional known URLs, fetches each page, follows allowed links, and stores the answers and page signals it finds. Each report belongs to a stale crawl, which is why two reports may not match after a site change.
In practice, treat the first site scan as a baseline. Scheduled crawls then show whether the issue count, page count, status codes and internal link patterns are moving in the right direction.
Crawlers
A crawler is software that automatically requests pages and follows links. Search engines use crawlers to find and revisit web content. They use audit tools to map a location and test technical signals. AI companies can run separate crawlers for training, search queries, or user-requested page access.
In practice, the crawler's user agent, rendering ability, behavior of robots.txt, and startup of URLs affect what it can review. A report should always tell you which crawler settings were used.
Crawl area
The crawl area defines which URLs a crawl is allowed to visit. It can include a hostname, all subdomains, a specific folder, or a provided URL list. Rules can exclude parameters, file types, external links, or areas that should not consume crawl quota.
In practice, you determine the scope before comparing crawls. A sudden drop from 8,000 pages to 3,000 could represent a serious detection problem or simply mean that a subdomain was excluded during the second crawl.
Crawl budget
Crawl budget is the practical amount of crawling attention a search engine devotes to a website over time. Large, frequently changing or inefficient websites feel the impact the most. Small websites rarely need to search for an exact budget number, but can still waste time crawling endless parameter combinations and redirect chains.
In practice, you combine crawl evidence with server logs. Logs display which URLs search bots actually requested, while audit shows the duplicate paths, traps, and errors these requests may be consuming.
Crawl frequency
Crawl frequency indicates how often a crawler revisits a URL or website. Search engines tend to return additional often to URLs that change frequently, receive strong internal links, or appear important. Slow servers, repeated errors, and a low-value URL extension can reduce meaningful revisits.
In practice, server logs deliver the clearest evidence. A sitemap date or scheduled scan may indicate that a URL should be checked, but only the logs show when a specific bot fetched it.
JavaScript rendering
JavaScript Rendering means loading a page into a browser engine and allowing its scripts to create or modify the visible content. A simple HTML crawler sees the first server response. A rendering site scanner can also review links, text, metadata, and structured data that appear after scripts are executed.
In practice, compare raw and rendered versions. Important content that is present after rendering may be slower or more difficult for some crawlers to process. Broken scripts can also cause a seemingly valid URL to remain almost empty.
Raw HTML
Raw HTML is the original document returned by the server before a browser runs page scripts. It contains the status response text, metadata written by the server, links and immediately available content. Search crawlers typically start with this version, even if they render the page later.
In practice, check whether titles, canonicals, body text and primary links already exist in the raw HTML. If the raw version is thin, the page is more dependent on successful rendering.
Rendered HTML
Rendered HTML is the document after a browser engine runs JavaScript and applies the resulting URL changes. It may contain content and links that were missing in the first answer. Elements can also be lost if a script replaces the page incorrectly.
In practice, the difference between raw and rendered images is often more useful than either version alone. It shows which scripts have been added, removed or rewritten, including canonical text and structured data that should not change unexpectedly.
Server-side rendering
Server-side rendering means that the server sends meaningful URL content in the first HTML response. JavaScript may add interaction later, but the body text and links do not depend on the visitor's browser being displayed.
In practice, it is easier for many crawlers to retrieve and diagnose server-rendered content. It also gives users something useful faster. The term describes delivery and not quality. Therefore, a server-rendered page may still have poor content or technical errors.
Crawler trap
A crawler trap is a URL pattern that builds a very large or virtually endless set of crawlable URLs. Calendar navigation, layered filters, internal search URLs, session identifiers, and repeated sorting parameters are common causes.
In practice, look out for rapidly growing URL counts, repeated templates with minor parameter changes, and paths that get deeper without adding useful content. Fixes may include better link controls, parameter rules, canonicals, and carefully tested robots.txt directives.
Log file analysis
Log file analysis interprets the web server's request records to review what crawlers and users actually requested. Unlike a simulated crawl, logs show real bot visits, response codes, request frequency, and wasted requests on URLs that may no longer be linked.
In practice, you use protocols to test assumptions from an audit. A URL may look crawlable but not receive search bot visits, while an old parameter pattern may still consume thousands of requests.
Robots.txt
Robots.txt is a text file in the root of a hostname that provides crawling instructions to participating bots. It can allow or disallow URL paths and point to XML sitemaps. It controls crawling, not support indexing, and different bots may interpret unsupported rules differently.
Field test accurate user agent groups and paths ahead of publishing a change. Blocking a URL in robots.txt can prevent a site scanner from seeing a noindex directive placed on the same page.
XML Sitemap
A XML sitemap is a machine-readable list of URLs that a website wants to inform search engines about. It supports detection, especially for new, deep or weakly linked URLs. A sitemap is a reference, not a command, to crawl or index all listed URL.
In practice they comprise a canonical, indexable 200 pages. Remove redirects, errors, duplicate variants and the noindex URLs. Accurate last modified data can help crawlers detect meaningful changes, while updated data with no content changes adds additional noise.
Indexing and URL signals
Discovery receives a URL in the verification process. Indexing signals tell a search engine whether URL should appear in the results, which version should represent the copy, and how language or parameter variants relate to one another.
Index
A search index is the stored collection of URLs and information that a search engine can reference when answering a query. Being crawled does not mean that a page has been added to the index, and being indexed does not mean that it ranks for a useful search.
In practice, in addition to the canonical selection, content quality, internal links and server responses, also check the index status. Search Console is the best source for Google's view on a specific URL.
Indexable page
An indexable page is technically eligible to appear in search. Typically it returns 200, is not locked for inspection, has no noindex directive, and does not declare another URL as canonical. Entitlement is not the same as inclusion.
In practice, you use indexability as the first filter in a check. If an important URL is not indexable, copy improvements cannot solve the underlying access or directive issue.
No index
Noindex is a directive that asks search engines not to keep a URL in their searchable index. It can appear in a meta-robots tag or a HTTP X-Robots-Tag header. The crawler generally needs to access URL to read the directive.
In practice, you use noindex for pages that might be useful to visitors but shouldn't appear in search, such as some internal results pages. Don't use it casually on canonical category, product, or article pages.
X-Robots-Tag
X-Robots-Tag is a HTTP response header that contains indexing directives such as noindex or nofollow. It's useful for non-HTML files and for server-level rules, but can be easily missed because it doesn't appear in the page source.
In practice, check both response headers and HTML. A page may look indexable in its markup, while a server or CDN adds a Noindex header to every response.
Canonical
A canonical tag names the preferred URL for the content of a page. It helps search engines combine signals of duplicates or near-duplicates, but it is more of a hint than an absolute command. Internal links, sitemaps, redirects and content similarities can support or contradict this.
In practice, a canonical should normally point to an indexable 200 URL representing the same content. Canonical chains, broken targets, and canonical pages to unrelated pages weaken the signal.
Duplicate content
Duplicate content appears when the same or very similar material is available on more than one URL. It can come from printed pages, tracking parameters, protocol or hostname variants, copied product text, or CMS routing.
In practice, duplicates are not automatically a penalty. The problem is uncertainty and shared signals. Select a preferred version via redirects or consistent canonicals, link to it internally, and keep sitemaps aligned.
URL Parameters
A URL parameter is the part following a question mark, often applied for tracking, sorting, filtering, pagination, or application status. Different parameter values can change the content, leave it unchanged, or create nearly identical variants.
In practice, you classify parameters by purpose ahead of controlling them. Removing each query string can break useful pages, while leaving tracking and sorting variants open can lead to duplicates multiplying and unnecessary crawling.
Pagination
Pagination divides a long set of elements across multiple URLs, for example page 2 of a product category or an article archive. Search engines need crawlable links between pages to find deeper elements and understand the order.
In practice, give each useful URL its own self-referencing canonical and standard links to neighboring pages. Avoid forcing crawlers to use only a JavaScript button without a crawlable URL.
Hreflang
Hreflang annotations connect alternative languages or regional versions of equivalent copy. They help a search engine show the correct version to a user instead of treating each page as an independent duplicate.
In practice, annotations should be reciprocal, run valid language and country codes, and reference a canonical 200 pages. Each sentence should contain the page itself. Broken return links are a common reason for an entire cluster to be ignored.
HTTP status code
The status code HTTP is the server's short outcome for a request. It explains a browser or crawler whether the request was successful, postponed, failed, or should be retried. The visible page can be misleading if the status says otherwise.
In practice, you check the code and the content together. A sophisticated error message that returns 200 is still a soft error, while a useful page that returns 500 may disappear from search if the error persists.
200 OK
200 means the request was successful and the server returned a normal response. Indexable pages usually require this status, but 200 alone does not prove that the content is useful, indexable, or even the right page.
In practice, look out for soft 404 errors, blank rendered pages, login screens, and error messages that incorrectly return 200. The code says that the request worked, not that the page served its purpose.
301 redirect
A 301 is a permanent redirect from one URL to another. It tells browsers and search engines that the destination should replace the old address. This is the normal choice for permanent moves, merged pages, and preferred hostname or protocol versions.
In practice, you will be redirected directly to the last relevant page. Update internal links and sitemaps so crawlers don't need the redirect forever, and avoid sending lots of unrelated old pages to the homepage.
302 redirect
A 302 is a temporary redirect. It states that the requested URL is still important and the move may not be permanent. Search engines may interpret long-lasting temporary redirects differently if the surrounding signals suggest a permanent change.
In practice, you use 302 for truly temporary routing, testing, or short maintenance flows. For a permanent URL replacement, a 301 gives clearer instruction.
404 Not found
A 404 means that the server could not locate the requested resource. A real 404 is normal for a URL that never existed or was removed without replacement. Problems arise when important linked pages return this value or when the error page itself returns the value 200.
In practice, restore the page, redirect to a nearby replacement, or remove links to it. Don't redirect all 404 responses to the homepage as this will obscure the actual cleanup work.
410 Gone
A 410 indicates that a resource was intentionally removed and is not expected to return. It's more explicit than a 404, although both can ultimately cause a search engine to drop a URL.
In practice, you use 410 when the removal is intentional and there is no sensible replacement. If a close successor exists, a permanent redirect usually serves users and preserves signals better.
5xx server error
A 5xx code means the server was unable to complete a valid request. Examples contain 500 for an internal error, 502 for a failed gateway, and 503 for temporary unavailability. Repeated errors can affect crawling and remove pages from results.
In practice, you separate short incidents from persistent patterns. Review application logs, server use, proxy or CDN behavior, and the exact templates affected. A planned outage should return 503 with sensible retry instructions instead of 200.
Links and site architecture
Links are the paths that members and crawlers follow. They also show which pages a website considers important and what those URLs are about. These terms explain the structure behind this movement.
Broken link
A broken link points to a destination that doesn't work, often a 404, 410, or persistent server error. Internal broken links break navigation and waste crawl requests. Broken external links divert visitors from the information they were expecting.
In practice, repair the source link instead of just counting the failed destination. Update it to the correct live file URL, remove it if there is no replacement, or restore the missing page if the link reveals an accidental deletion.
Redirect chain
A redirect chain occurs when one URL redirects to a second URL, which redirects again before reaching the last page. Each additional step increases latency, creates another source of error, and makes crawling less efficient.
In practice, you change internal links to point directly to the last 200 URL. Then shorten the old forwarding rules where it is safe to do so. Keep the oldest useful source redirects if people or external sites may still be using them.
Redirect loop
A redirect loop sends a request through redirects that eventually return to an earlier URL. The browser cannot reach the final page and usually displays the “Too many redirects” error. Conflicting protocol, hostname, locale, and trailing slash rules are common causes.
In practice, follow the full response path and review each rule in turn. CDN Behavior, application routing, and server configuration can all participate in the same loop.
Orphaned page
An orphan page is an indexable URL that has no internal links pointing to it in the crawled site graph. It may still appear in a sitemap, receive external links, or remain indexed, but normal navigation doesn't go there.
In practice, decide whether the URL deserves to exist. Valuable URLs need contextual internal links from relevant sections. Old, duplicate, or obsolete orphans may need to be consolidated, removed, or a noindex decision made.
Internal link
An internal link connects one URL to another page on the same website. It helps users move through the content, provides a discovery path for crawlers, and passes context and link equity to the target.
In practice, add links where the goal actually helps the reader. Site-wide navigation creates a broad structure, while a relevant text link can express a much more specific relationship between two pages.
Inlink
An inlink is a link that points to a page from another URL. Inlink counts help show how connected a URL is, but the sheer quantity does not reveal the strength, position, or relevance of those links.
In practice, check the source URLs and anchors. Ten contextual links from important related pages can be more important than hundreds of repeated footer links.
Outlink
An outlink is a link from the examined page to another URL. It may lead to another location on the same site or to an external domain. Outlinks show how a page connects readers and crawlers to supporting or next material.
In practice, check target status, relevance, anchor wording, and link attributes. A site without useful outlinks can become a dead end, even if many sites link to it.
Anchor text
Anchor text is the clickable text of a link. It tells the reader what to expect and provides search engines with context about the target. General phrases like “click here” contain less thematic information than a concise descriptive phrase.
In practice, first write anchors for people and vary them naturally. Repeating an exact phrase everywhere can seem forced, while using the same vague anchor for different targets creates ambiguity.
Link equity
Link equity is a useful model for authority spreading through links. A site that receives strong links can pass on some of that strength to other sites. The amount is not a public fixed unit and search engines consider many signals beyond the number of links.
In practice, redirect internal links from established URLs to important pages that need support. Swiss Knife SEO also takes link position into account, so body links are not treated as the same as repeated footer links.
Internal PageRank
Internal PageRank is a site-level simulation that shows how link strength flows through the internal graph. It orders pages relative to each other based on which pages link to them and how the linked pages are connected to each other.
In practice, you can use this to find important pages that don't receive enough internal support, unexpectedly high-performing pages with little value, and strong source URLs that can pass useful context to a priority target.
Click Depth
Click depth is the minimum number of internal link steps required to get to a page from a chosen starting point, usually the homepage. A depth of one means that the page is linked directly from that beginning. A depth of five means at least five clicks are required.
In practice, depth is a structural cue and not a universal ranking rule. Important pages should generally be easy to reach, while large websites naturally have more in-depth, long-tail content.
Nofollow
Nofollow is a link attribute that explains search engines not to treat a link as a normal recommendation. Modern search engines may run it as a hint rather than an absolute rule. The link can still direct visitors and help them achieve the goal.
In practice, you use nofollow where the relationship requires it, for example on untrustworthy or commercial links. Avoid including it in normal internal navigation as this may conflict with the site's link diagram.
Backlink
A backlink is a link from another site to yours. Backlinks can direct referral visitors, aid discovery, and contribute to search authority. Their value varies greatly depending on relevance, editorial context, source quality, placement and whether the link remains active.
In practice, consider new and lost links as events, not trophies. A relevant editorial link can be more useful than a large number of automated directory links.
Referring domain
A referring domain is a unique website that links to your website. A domain can create one backlink or thousands of site-wide backlinks. Counting domains therefore gives a different perspective than counting each individual link.
Check both in practice. An increase in backlinks without an increase in referring domains can be due to a repeated template link, while a new relevant domain can represent a truly new source of visibility.
Backlink lost
A backlink becomes lost when a link recorded during an earlier check can no longer be reached at the expected source or destination. The referring page may be gone, the target may have moved, or the publisher may have changed the link attributes.
In practice, confirm the loss before trading as temporary errors and crawler timing can lead to false positives. Reclaim a valuable link when there is a clear relationship and a suitable live target.
Content and page marking
Search engines read visible content and the labels surrounding it. These terms include the page elements that summarize, structure, describe and classify that content.
Title tag
The title tag is the HTML title of a URL. It often provides the main blue link in search outcomes and appears in browser tabs. Search engines can rewrite it if other page signals appear more useful for a particular search query.
In practice, make each important title specific, precise and unambiguous. Start with the topic of the page, add meaningful context, and avoid repeating a standard sentence so many times that every result looks the same.
Meta description
A meta description is a short HTML summary of a page. It is not a direct ranking factor, but can influence how the search result is displayed and whether a user selects it. Search engines often generate their own snippet from the page text.
In practice, write a useful promise that fits the page. Avoid empty slogans and don't rely on a precise character limit as display space changes depending on device and query.
Heading H1
H1 is the top-level HTML heading that names the main topic of the page. It helps users and assistive technologies understand the content structure and gives search engines a clear thematic indication.
In practice, you should use a clear main heading for most pages and keep the visible hierarchy underneath logical. Multiple H1 elements are valid HTML, but a single main heading is usually easier to understand and review.
Alt text
Alt text is the text alternative that is attached to an informative image. Screen readers use it, browsers may display it when an image fails, and search systems run it as image context. Decorative images should normally have empty alt text, so they are skipped.
In practice, describe the purpose of the image in the page context. Don't fill in keywords, don't repeat adjacent captions word for word, and don't start every description with "Image of."
Thin content
Thin content offers too little original value for the purpose of the page. The number of words alone does not define it. A two-sentence contact page can be complete, while a 1,000-word category URL can still be thin if it repeats generic text and doesn't answer a purchasing question.
In practice, evaluate the page based on user intent, uniqueness, evidence, and usefulness. Consolidate overlapping weak URLs or add the missing information that makes the page worth visiting.
Content expiration
Content decay is a sustained loss of organic performance of a page that previously received more clicks or visibility. Facts may become outdated, competitors may be able to answer the topic better, search intent may change, or stronger pages on the site may compete with it.
In practice, compare several identical time windows and take seasonality into account. Refreshing the page may help, but only after determining whether the loss is due to content, indexing, SERP changes, or cannibalization.
Cannibalization
Cannibalization is when multiple pages on the same website compete for the same search intent and search engines have difficulty choosing the best one. Normal keyword overlap is not enough to prove this. A website can easily rank multiple useful pages based on related meanings.
In practice, look out for unstable URL changes, shared links, similar content, and two pages that serve the same need. The right repair may be consolidation, clearer differentiation, or better internal linking.
Structured data
Structured data is machine-readable markers that identify the people, products, organizations, articles, events, and separate entities described on a URL. Search engines can run supported types to understand content and produce richer results features.
In practice, the markup must fit the visible content and meet the requirements of its type. Valid syntax does not support a rich result and markup cannot result in qualifying an inappropriate or misleading page.
JSON-LD
JSON-LD is a JavaScript object notation format often used to publish structured data within a script block. It separates the machine-readable description from the visible HTML, making it easier to manage than older inline markups.
In practice, check both the syntax and the meaning. A block may be valid JSON, but have the incorrect canonical name URL, omit required properties, or describe information that visitors cannot see.
Entity
An entity is a unique thing that a search system can identify, such as a person, organization, place, product, or concept. Entity understanding connects facts and relationships that go beyond the exact words used on a page.
In practice, keep names, descriptions, profile links, organization details and structured data consistent. Clear factual context helps search and AI systems understand who created the content and what the website represents.
Performance and search data
Rankings are only part of organic performance. These terms explain the page experience, results page features, Search Console measurements, and the business actions that take place after a visit.
Core Web Vitals
Core Web Vitals are Google's present user experience metrics for loading, visual stability, and responsiveness to interactions. The set contains Largest Contentful Paint, Cumulative Layout Shift and Interaction to Next Paint.
In practice, interpret them by URL type and evidence source. Field data reflects real Chrome users over time, while lab testing is done during a controlled visit and is better suited to diagnosing a specific problem.
Largest Contentful Paint, LCP
LCP measures how long it takes to render the largest visible content item in the first viewport. This element is often a hero image, a headline block, or a large banner. Google considers 2.5 seconds or less at the 75th percentile to be good.
In practice, identify the actual LCP element before changing the code. Slow server response, delayed image recognition, oversized media, files blocking rendering, and client-side rendering can delay the process.
Cumulative layout shift, CLS
CLS measures unexpected movements of visible page elements during the life of the page. Images with no dimensions, inserted banners, font changes, and late-loading advertisements can cause content to move after a user has started reading. A value of 0.1 or less is considered good.
In practice, reserve space for media and dynamic components. User-initiated movements may be acceptable, but moving content without input is the experience the metric is intended to capture.
Interaction with Next Paint, INP
INP measures how quickly a page provides visual feedback following user interactions such as clicks, taps, and keystrokes. It observes interactions throughout a visit, beyond the first. Google considers 200 milliseconds or less at the 75th percentile to be good.
In practice, long JavaScript tasks, heavy event handlers, and excessive main thread work are common causes. A page can load quickly and still feel slow if a menu or button responds too late.
Time to first byte, TTFB
TTFB measures the time between starting a request and receiving the first byte of the response. It reflects network travel, connection establishment, CDN behavior and server processing. It's not a Core Web Vital, but it does affect how quickly you can start rendering.
Practically compare cached and uncached responses, locations, and URL types. Slow TTFB can be caused by hosting, database work, application code, or a remote uncached origin.
SERP
SERP means search engine results page. A modern SERP can contain organic links, advertising, maps, videos, products, featured snippets, people also ask, and AI-generated answers. Therefore, the same ranking number can receive very different attention for two searches.
In practice, track the characteristics surrounding a ranking and its position. A page ranked fourth among several large features has less visible space than a clean page ranked fourth.
SERP feature
A SERP feature is any result format that goes beyond a standard organic listing. Examples contain local packs, featured snippets, images, video carousels, shopping outcomes, and AI overviews. Features can create a visibility opportunity or push normal results further down the line.
In practice, map each feature to the content format and intent it can use. A local package requires different work than a featured snippet or product result.
Featured snippet
A featured snippet is an extracted answer that appears significantly above or below standard organic results. It can appear as a paragraph, list, table, or other compact format with a link to the source page.
In practice, answer the question clearly near a relevant heading and support the short answer with broader context. Owning a snippet can increase visibility, but the impact on clicks varies depending on the query.
People Also Ask
People Also Ask is a series of expandable, related questions in Google results. Opening an answer can reveal additional questions and provide a practical overview of how searchers refine a topic.
In practice, you should use the questions as research, not as a writing template that you can blindly copy. Add a section if the question fits the purpose of the page and you can answer it with useful, accurate details.
Rank Tracking
Show rank tracking data where a domain or URL appears for selected queries over time. Useful monitoring includes country, language, device, result type and surrounding SERP features. A position without this context can be misleading.
In practice, observe groups and trends rather than reacting to each daily movement. Search outcomes vary by location, personalization, testing and data center, so an audit is an example rather than an absolute truth.
Impressions
An impression is recorded in Search Console when a outcome from your property is displayed or loaded in a way that complies with Google's reporting rules. It signals search visibility, not a visit, and the exact count may vary depending on the result type.
In practice, increasing impressions on flat clicks can lead to greater query coverage, lower average positions, or a weaker click-through rate. Segment by page, query, country and device before deciding why.
Click rate, CTR
Organic CTR is clicks divided by impressions. It describes how often an impression becomes a click. Position has a big influence on this, but titles, snippets, brand awareness, query intent and SERP features also change the result.
Compare similar requests and positions in practice. A low CTR doesn't always mean the title is bad, especially if the results page directly answers the query or contains large competing features.
Average position
The average position in Search Console is your property's average top position for recorded impressions. It combines many search queries, devices, locations and result presentations, so it is not a fixed rank of a URL.
In practice, narrower query and page segments are used for diagnosis. The site-wide average may improve because low-ranking impressions will disappear even if no important page has moved up.
Search query
A search query is the exact wording that a person types or speaks into a search system. It is different from a keyword, which is often the phrase that a marketer wants to track or target.
In practice, you group queries by intent and topic ahead of making page decisions. Several different phrases can express a need, while nearly identical words can have different commercial or informational intent.
Organic traffic
Organic traffic is the number of visits attributed to unpaid search results. Analytics platforms classify visits based on referrer and campaign rules, meaning the number of Search Console clicks may vary and change as attribution settings change.
In practice, you use Search Console for search appearances and queries, as well as on-site analytics for sessions and results. Expect reasonable differences rather than forcing both systems to match exactly.
Conversion
A conversion is a user action that the company has selected as a meaningful outcome, such as making a purchase, submitting a qualified form, starting a trial, making a phone call, or creating an account. The definition belongs to the measurement plan, beyond to SEO.
In practice, link organic landing pages to conversions and revenue whenever possible. More traffic isn't automatically better if it reaches the wrong audience or doesn't drive useful actions.
AI visibility in search
AI response systems can summarize sources, mention brands, and cite pages without displaying a familiar list of ten organic links. The field is still changing, so these terms separate measurable evidence from early conventions and assumptions.
AI overview
An AI summary is an answer generated by Google that may appear in search results for some searches. It can combine information from multiple sources and display supporting links. Layout, availability, and cited sources may change between searches.
In practice, note whether the overview appears, whether your domain is cited and which page deserves the citation. Do not treat a citation as a permanent position as the results generated are variable.
Citation
A citation is a visible source reference within or next to an AI-generated answer. It links a claim or part of the answer to a URL that the system used or displayed. A brand can be mentioned without being cited, and a page can be cited without the brand name appearing.
In practice, save the exact answer, cited URL, citation position, prompt, market and date. This evidence makes subsequent comparisons meaningful.
AI visibility
AI visibility describes how often and how prominently a brand or domain appears in responses from AI search and assistance systems for a defined set of prompts. This can include citations, brand mentions, source position, sentiment and share of voice.
In practice, define the prompts and markets ahead of reading the score. A percentage based on ten buyer questions means something very different than a percentage based on hundreds of common prompts.
AI crawler
An AI site scanner is an automated bot operated for an AI-related purpose. Various bots can collect training evidence, build a search index, retrieve a page for a user query, or review content for another service. Allowing one bot does not automatically allow all bots from the same company.
Review documented member agents, robots.txt groups, CDN rules, and server logs in practice. Decide access based on the bot's purpose and site policies.
Knowledge graph
A knowledge graph connects identifiable people, organizations, products, places, and the facts that relate them. Search systems and answer engines consult those connections when they resolve names, support factual responses, and decide which entity a query describes.
In practice, consistent organizational details, authoritative profiles, SameAs links, and accurate structured evidence can support clear entity relationships. A chart entry is not something that can launch a site with a single tag.
Answer machine
An answer engine responds to a question with a generated or extracted answer, beyond a ranked list of links. Search engines, chat assistants, and specialized research tools can all act as response engines.
In practice, evaluate the response, sources and variability together. The same prompt may result in different wording or citations on a different day, so repeated reviews are more useful than a single screenshot.
Prompt
A prompt is the question or instruction sent to an AI system. Small changes in wording, audience, location, limitations, or conversation flow can change the answer and its sources.
Real-world prompts that reflect real buyer questions and capture their market and language. Search Console queries can provide useful ideas, but conversational prompts often require more context than a traditional keyword.
llms.txt
Llms.txt is an emerging voluntary file format that represents a website's favorite pages and short descriptions in a format intended for language model tools. It is usually published to the root of the site. It is not a search ranking policy and support across AI systems is still under development.
In practice, treat it as a cost-effective reference layer and not as a replacement for crawlable URLs, strong internal links, sitemaps or structured evidence. Keep the list selective and current.
Agents.json
Agents.json is an early manifest convention for describing a site or service to software agents. Suggested uses include identity, capabilities, policies, and machine-readable interaction details. It is not an established web standard and acceptance remains limited.
In practice, only publish it if it accurately describes the site and validate the JSON structure. Don't portray it as a support path to AI citations or traffic.
AI readiness score
An AI readiness score is a summary of product-level signals that can impact whether AI crawlers can access and interpret a website. Inputs may contain crawler permissions, server rendering, entity clarity, structured evidence, content accessibility, and the quality of internal discovery paths.
In practice, read the component exams instead of just chasing the score. The number is not assigned by an answering machine and cannot support citation.
Brand mention
A brand mention occurs when an answer names a company, product, or organization, whether or not it references a source of that brand. Mentions can show awareness, but they do not prove that the system used the brand's own website.
In practice, note mentions separately from citations and check surrounding wording for accuracy and sentiment. The most useful analysis shows who was named, who was cited, and what source supported the response.
Apply the terms to a live website
Definitions are preserved if they describe your pages. Take a free audit and watch the vocabulary become a list of options.