A/B Test
An A/B test is the controlled comparison of two variants — typically of a webpage or ad — for statistically robust determination of which variant performs better.
Glossary
Defined terms from performance marketing, SEO, AI SEO, conversion optimization — each with a standard definition and a Calvarius practice block. Pillar entries link out to in-depth services and articles.
Inventory
An A/B test is the controlled comparison of two variants — typically of a webpage or ad — for statistically robust determination of which variant performs better.
Ad scheduling controls at which times of day and weekdays a campaign's ads are played out — operatively important for B2B setups with business hours and industries with clear activity patterns.
AI crawlers are the web crawlers of AI search and AI language model providers that collect web content for training data and search functions — with different control logics per provider.
AI Mode is Google's fully-featured conversational AI search experience as a separate tab in search results — structurally distinct from AI Overviews through multi-turn conversation and deeper answer structures.
AI Overviews is Google's AI-generated answer block that appears above the classical organic search results and presents content from multiple sources in summarized form.
An attribution model is the methodology by which a conversion is assigned to one or more touchpoints — decisive for the evaluation of individual marketing channels in complex customer journeys.
B2B (Business-to-Business) and B2C (Business-to-Consumer) refer to the two fundamental marketing logics by recipient type. Structural differences: B2B purchase processes last weeks to months, involve multiple decision-makers, require rational argumentation and detailed information; B2C purchase processes are shorter, more emotional, often single-person decisions. From this follow different marketing tools: B2B uses LinkedIn, webinars, whitepapers, lead generation; B2C uses Instagram, Performance Max, retargeting, faster conversion paths. Those who steer B2B campaigns with B2C methodology (or vice versa) systematically produce worse results.
The backlink profile encompasses all external links pointing to a website. Evaluation criteria include number, thematic relevance of linking sites, their domain authority, and source diversity. Historically the backlink profile was the dominant ranking factor in Google's algorithm. In the AI search era, its relative importance has receded somewhat — AI search systems increasingly evaluate content substance, semantic clarity, and first-hand observations as equally weighted criteria. Artificially constructed backlinks (link farms, paid links) remain a penalized risk, not a leveraged shortcut.
Bid inflation is the effect that average click prices in advertising auctions rise over time — amplified by the use of automated Smart Bidding strategies. When multiple competitors simultaneously rely on Target CPA or Maximize Conversions, their algorithms each optimize for conversion probability and outbid each other in lucrative micro-auctions. Economic consequence: CPCs rise industry-wide without conversion rates improving correspondingly. Strategic response: don't follow every competitive auction; hold your own efficiency thresholds — even if it costs volume.
Bounce rate is the percentage of sessions in which a user accesses only one page and then leaves the site. In Google Analytics Universal (GA3) bounce rate was a central metric. In GA4 it was structurally replaced by „engagement rate" — a session counts as engaged when it lasts at least 10 seconds, contains a conversion, or has at least two page views. Practice note: high bounce rate is not automatically bad — on information pages (glossary entry, blog post with concrete answer) many „bounces" are actually successful short visits. Assessment must be context-specific.
Budget pacing is the temporal distribution logic by which ad platforms like Google Ads spend the configured budget over a period — either evenly (standard pacing) or accelerated.
The canonical tag (`<link rel="canonical" href="...">`) signals to search engines, in cases of multiple identical or similar pages, which is the preferred version. Typical use cases: tracking-parameter URLs (`?utm_source=...`), filter and sort URLs in shops, print versions of a page, language versions with non-complete translation setup. With missing or incorrectly set canonical, duplicate-content problems arise — search engines then choose a URL themselves, often with disadvantageous results for visibility. With multilingual sites, observe the interplay with hreflang.
Claude-Web refers to Anthropic's crawler family for the web search functionality of Claude. Anthropic differentiates between multiple sub-bots: `ClaudeBot` for training data collection, `Claude-Web` for search functionality, `claude-user` for direct user requests (for instance, when Claude users explicitly request web content). Control via robots.txt with the corresponding User-Agent strings. Practice observation: Claude cites its sources in the search function significantly more prominently than ChatGPT — those who want visibility in the Anthropic world should not block Claude-Web.
Click-through rate (CTR) is the ratio of clicks to impressions, expressed as a percentage. Calculation: clicks divided by impressions, times 100. CTR is a central indicator for ad or snippet attractiveness — poor CTR signals that the presentation does not appear relevant enough. In Google Ads, CTR feeds into Quality Score. In organic SERPs, Google uses CTR as a ranking signal — when snippets are systematically clicked more at comparable positions, ranking rises. In the AI search era, CTR becomes structurally more complicated because many answers take place in the AI overview without a click.
The conversion funnel is the structured visualization of the steps a user typically takes from first brand contact to conversion. Classical stages: Awareness, Interest, Consideration, Conversion, occasionally extended by Retention and Advocacy. Practice use: by measuring drop-off rates between funnel stages, structural weaknesses can be identified. Example: if 100 users start a trial, 20 of them complete onboarding, and only 3 convert, the primary lever is between onboarding and conversion. Important: not all user journeys are linear — multi-touch attribution is often more realistic than pure funnel models.
Conversion rate is the percentage share of visitors or sessions that execute a defined conversion action — the central metric in performance marketing and conversion optimization.
Conversion Rate Optimization (CRO) is the discipline of systematic increase in the share of visitors who perform a desired action — purchase, lead sign-up, download. Methodologically, CRO works with three pillars: quantitative analysis (web analytics, funnel data), qualitative analysis (heatmaps, session recordings, user surveys), and iterative tests (A/B tests, multivariate tests). Economic relevance: often a 30 percent conversion rate improvement is cheaper to achieve than a 30 percent traffic increase — with identical volume effect. CRO is therefore usually the economically most rational optimization discipline for sites with sufficient traffic.
Cookieless tracking refers to tracking methods that work without third-party cookies — as a response to browser privacy updates (Safari ITP, Firefox ETP, Chrome third-party cookie phase-out) and regulatory requirements. Important methods: server-side tracking (data goes from server instead of browser to analytics tools), first-party tracking (own cookies on the own domain), hash-based user recognition (e.g., via email hash), context-based advertising strategies without person tracking. Practice implication: classical funnel attribution works less reliably; in return, aggregation models (modeled conversions in Google Ads, conversion modeling in GA4) gain importance.
Core Web Vitals are Google's standardized performance metrics for quantitatively evaluating user experience of a webpage — LCP, INP, and CLS — with direct influence on ranking.
CPA (Cost per Acquisition, Cost per Action) refers to the costs an advertiser pays on average for a conversion — whether a sale, a lead sign-up, a download, or another defined action. Calculation: advertising costs divided by conversions in the same period. Economic relevance: CPA is the central performance marketing metric. Smart Bidding strategies like Target CPA explicitly optimize for this metric. Important: CPA must be evaluated in relation to Customer Lifetime Value (CLV) — a low CPA with low CLV is economically worse than a higher CPA with high CLV.
CPC (Cost per Click) is the dominant billing model in search and display advertising: advertisers pay only when a user actually clicks on the ad. The actually paid CPC results from an auction logic with co-bidders' bids and Quality Score as modifier — rarely the maximum bid value. Strategic importance: CPC is an intermediate metric, not a target metric. Low CPC with high bounce rate is economically worse than higher CPC with good conversion rate. The economic target metric is CPA, not CPC.
CPM (Cost per Mille, Cost per Thousand Impressions) is the advertising billing model per thousand ad impressions — regardless of whether users click or convert. Standard for display advertising, YouTube branding campaigns, and programmatic buying. Sensible when the marketing goal is reach and visibility (branding, product launch), not direct performance. Economic complication: CPM campaigns deliver reach data, but no direct conversion signals — success measurement shifts to indirect KPIs like brand lift, direct traffic increase, or search volume rise.
The crawl budget is the number of URLs a search engine crawler — primarily Googlebot — accesses on a domain within a specific period. For small sites (under 1,000 pages) practically irrelevant, because Google crawls everything anyway. For large sites (tens of thousands to millions of URLs) economically critical: if crawl budget is consumed by unimportant URLs (filter combinations, tracking parameters, old content), important new content remains unindexed. Levers for optimization: robots.txt Disallow for unimportant paths, XML sitemap prioritization, canonical tags, server performance — crawlers reduce their frequency when the site responds slowly.
Day-parting refers to the strategy of delivering advertising ads by time of day, weekday, or other temporal patterns — for example, only during business hours for B2B mandates or in the evening for B2C impulse purchases. Day-parting is implemented via the ad scheduling of the respective advertising platform. Important: with the change to Google Ads budget pacing from June 2026, day-parting works differently than before — daily budget logic and ad scheduling logic are no longer proportionally coupled. Those who use narrow day-parting windows must adjust their budget strategy accordingly.
DefinedTerm is the Schema.org type for term definitions — specifically designed for glossary entries, dictionary definitions, and technical terminology. Important properties: `name` (the term itself), `description` (definition), `inDefinedTermSet` (reference to parent collection), `termCode` (technical identifier). AI search systems like ChatGPT, Claude, and Perplexity evaluate DefinedTerm markup more highly than generic Article or BlogPosting markup for „What is X" queries — because it semantically signals more precisely that the page is a term definition. Combined with DefinedTermSet on the glossary overview page, this is a strong AI SEO lever.
DefinedTermSet is the Schema.org type for a collection of DefinedTerm entries — the semantically appropriate markup for the overview page of a glossary. Important properties: `name` (name of the collection, e.g., „Calvarius Glossary"), `description`, `hasDefinedTerm` (list of contained DefinedTerm references via `@id`). Search engines recognize through DefinedTermSet the structural relationship between overview page and detail entries — which is relevant for site architecture assessment and indexing logic. Combined with DefinedTerm on individual detail pages, a consistent schema setup for glossaries of all sizes.
Domain Authority refers to the overarching trust and visibility strength of a domain in search engines — conceptually central, technically defined differently depending on the tool and model.
E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is Google's official evaluation framework for content quality — central in the Search Quality Rater Guidelines and bound into ranking via algorithms.
Featured snippets are highlighted answer boxes at the top of Google's search results — sometimes called „Position Zero" because they stand above the first organic result. Content formats: short definitions, enumerations, tables, video snippets. Featured snippets are the direct predecessor of AI Overviews — both reduce click-through rates to organic results below. Since May 2024 in the US, since March 2025 in the DACH market, many featured snippets are being displaced by AI Overviews. Strategically: content optimized for featured snippets often also works for AI answers.
First-party data is data a company captures directly from its own interactions with users — newsletter sign-ups, purchase behavior in its own shop, account data, contact inquiries, web analytics on its own domain. In contrast to third-party data, which comes from external providers (data brokers, third-party cookies). Strategic importance: with the phase-out of third-party cookies in Chrome (continuous process since 2024) and comparable browser restrictions, first-party data becomes the primary source for targeting, personalization, and advertising optimization. Economic consequence: investments in own data capture (newsletter, account systems, CRM) are strategically more valuable today than 5 years ago.
A funnel analysis examines drop-off rates between the steps of a conversion funnel — a diagnostic tool for identifying structural conversion weaknesses.
GDPR-compliant tracking refers to web analytics and advertising tracking observing the General Data Protection Regulation. Three core principles: consent (tracking cookies may only be set after active agreement), data minimization (only collect data needed for the defined purpose), transparency (privacy statement with concrete listing of tools). Operationally relevant: cookie consent banners (Consent Management Platforms like Cookiebot, Usercentrics, Borlabs), server-side tracking as a privacy-friendlier alternative, abandoning third-party trackers where possible. Economic consequence: tracking data after consent typically captures only 60-85 percent of sessions, making performance analytics methodologically more complicated.
Generative Engine Optimization is the discipline of systematic optimization of web content for AI search systems like ChatGPT, Claude, Perplexity, and Google's AI search products — the counterpart to SEO for the AI search era.
GPTBot is OpenAI's crawler that collects web content for training future ChatGPT models. Introduced in August 2023, blockable via robots.txt using `User-agent: GPTBot` + `Disallow: /`. Important: GPTBot is separate from OAI-SearchBot, the crawler for ChatGPT Search functionality — a site can block GPTBot but allow OAI-SearchBot (or vice versa). Strategic decision: those who deny training data exclude themselves from future ChatGPT visibility; those who allow contribute to model substance without direct citation claim. Most site operators allow GPTBot — the visibility disadvantage of blocking outweighs the data protection argument.
A heatmap is a color-coded visualization of user behavior on a webpage. Three types are common: click heatmaps show where users click; scroll heatmaps show how far they scroll; move heatmaps show mouse movement patterns. Tools: Hotjar, Microsoft Clarity (free), Mouseflow, Crazy Egg. Valuable for conversion optimization — heatmaps show whether CTAs are perceived, whether important content is reached, and whether there are click attempts on non-clickable elements. Important limitation: mouse movement does not correlate with attention on mobile; click heatmaps only show what happened, not why.
Hreflang is an HTML attribute that defines the relationship between different language and region versions of a page on multilingual or region-specific websites. Each page refers via hreflang tags to its counterparts (e.g., `de`, `en`, `x-default`) so that search engines deliver the appropriate locale version to users. Common practice errors: missing self-references, asymmetric counterpart references (page A refers to B, but B doesn't refer back), or conflicts between hreflang and canonical tag. Search Console reports hreflang issues as a dedicated report.
Hydration is the process by which server-side pre-rendered HTML is „brought to life" in the browser through JavaScript and made interactive — central mechanic of modern SSR web frameworks.
Inline linking is the practice of placing source links in AI-generated answers directly next to the matching text fragment, rather than exclusively displaying them in a source list at the end. Google rolled out inline linking more prominently in AI Overviews and AI Mode in May 2026. From the sites' perspective, content with fragment-specifically citable substance benefits primarily — clear definitions, concise step statements, precise enumerations. Narrative argumentation that only makes sense as a whole is inline-linked less often.
Internal linking refers to linking between pages within the same domain. Three functions are central: first, distribution of PageRank from authoritative hub pages to thematically related detail pages; second, steering crawler navigation and crawl depth; third, signaling thematic clusters and topical authority to search engines. Strategically important is the discipline of linking logic — indiscriminate many links create link inflation; selective links to strategic target pages distribute authority efficiently.
JSON-LD (JavaScript Object Notation for Linked Data) is a format for structured data in HTML pages, embedded in a `<script type="application/ld+json">` block. Google prefers JSON-LD over older alternatives like Microdata or RDFa — it's cleaner to maintain, separated from visible content, and easier to validate. Schema.org markup is today predominantly implemented as JSON-LD: DefinedTerm, Article, Product, Organization, BreadcrumbList, and so on. Important practice discipline: JSON-LD must content-wise match the visible content of the page — structured data claiming different information than the page shows is considered a spam attempt.
A knowledge panel is an information box that Google displays to the right of search results for queries about entities — people, companies, places, works. The data source is Google's Knowledge Graph, a structured database of entities and their properties. Sources for knowledge panel content: Wikipedia, Wikidata, official websites, structured data of the entity itself. For companies, this is relevant: a knowledge panel signals authority and occupies prominent SERP real estate. Building a knowledge panel requires consistency in entity data across multiple sources — Google Business Profile, Wikipedia entry, schema.org markup on the own site.
The central Core Web Vitals consist of four metrics: LCP (Largest Contentful Paint) measures the time until rendering of the largest visible content element — target under 2.5 seconds. FID (First Input Delay) measures the delay of the first user interaction — was replaced by INP in March 2024. CLS (Cumulative Layout Shift) measures visual stability, that is, how much layout shift occurs during loading — target below 0.1. INP (Interaction to Next Paint) measures responsiveness to all user interactions — target below 200 milliseconds. Poor values on one of these metrics can noticeably burden rankings.
Lead generation is the discipline of systematic acquisition of qualified prospects — persons or companies that have signaled demonstrable interest in the offering. In B2B context often the primary KPI of performance campaigns, because the sales process is longer and consultation-intensive and direct online conversions are rare. Typical lead conversion paths: whitepaper download, demo request, consultation appointment, contact form. Economic differentiation: Marketing-Qualified Lead (MQL) versus Sales-Qualified Lead (SQL) — not every lead has the same economic value, which complicates CPA steering in advertising campaigns.
Long-tail keywords are specific, often multi-word search queries with low search volume but high intent clarity. Example: „shoes" is a short-tail keyword; „waterproof hiking shoes for wide feet women" is long-tail. Economic relevance: long-tail queries have significantly higher conversion probability because search intention is clearer. In the AI search era, long-tail queries are increasingly answered in AI Overviews — those cited there as a source receive qualified traffic without competing with large competitors for short-tail keywords.
Maximize Conversions is a Smart Bidding strategy in Google Ads that fully spends the set budget and algorithmically optimizes for maximum conversion volume. Unlike Target CPA, there is no fixed cost-per-action target value — the strategy scales bids as far up as the budget allows. Practice relevance: with rising budgets without strategy adjustment, Maximize Conversions tends to lead to declining conversion quality, because increasingly expensive clicks with vaguer intent are purchased. Sensible primarily with clearly defined budget and sufficient conversion volume for algorithm learning (at least 30 conversions/month).
OAI-SearchBot is OpenAI's crawler that indexes web content for the ChatGPT Search function (SearchGPT, ChatGPT Search) — separated from GPTBot, the training crawler. This separation is strategically relevant: site operators can decide differently whether their content may be used for training data (GPTBot control) and whether they appear as a source in ChatGPT Search answers (OAI-SearchBot control). Those who block OAI-SearchBot via robots.txt forgo the possibility of being cited in ChatGPT Search — which is rarely sensible given growing user numbers.
Performance marketing is the marketing discipline in which every action produces measurable performance and is steered along quantitative KPIs (clicks, conversions, revenue, CPA, ROAS). In contrast to branding marketing, in which impact is longer-term and harder to measure. Typical performance marketing channels: Google Ads, Meta Ads, Microsoft Ads, LinkedIn Ads, affiliate marketing, SEO, email marketing. Economic logic: every invested euro should bring back more than one euro in value — efficiency before reach. Important: performance marketing isn't „better" than branding, it serves a different purpose. Mature brand strategies combine both disciplines.
Performance Max is Google's campaign type introduced in 2021 that automatically delivers a single campaign across all Google inventories — Search, Display, YouTube, Discover, Gmail, Maps.
PerplexityBot is the crawler of Perplexity AI, an AI search platform structurally oriented toward prominent source citation. While ChatGPT and Google AI Overviews often place sources at the edge or bottom, Perplexity shows 6-10 numbered source links directly under each answer. From an SEO perspective, Perplexity is thus the AI search platform with the highest referral traffic potential. Control via robots.txt with `User-agent: PerplexityBot`. Blocking is rarely strategically sensible — those visible in Perplexity gain direct qualified traffic.
Pillar content refers to particularly extensive, deeply researched articles that comprehensively cover a topic — a strategic format with significantly better visibility in classical search and AI search systems than short standard posts.
Pre-rendering is the technique by which HTML versions of a Single-Page Application are generated server-side in advance and held in a cache. When a crawler accesses the site, the pre-render service delivers the pre-rendered HTML version instead of the empty SPA shell. Real users continue to receive the normal SPA version. Tools: Prerender.io, Rendertron, Puppeteer-based custom solutions. Economics: cheaper than complete server-side rendering rebuild (typically €300-1,500 setup, €30-150/month), but structurally weaker — doesn't work reliably for AI search systems and creates additional complexity.
Quality Score is Google's evaluation of the quality of a Google Ads ad on a scale of 1 to 10. Three factors feed in: expected click-through rate (CTR), ad relevance to the search term, landing page user-friendliness. Economic relevance: higher Quality Score lowers the effective CPC and improves ad rank — a Quality Score 9 account often pays 50 percent less per click than a Quality Score 5 account in the same auction. Levers for improvement: precise keyword-ad-landing-page congruence, clear conversion paths, fast loading times. Not directly visible in Performance Max campaigns.
Robots.txt is a configuration file in the root directory of a domain (`/robots.txt`) that steers crawler access. Via `User-agent` and `Disallow` directives, individual bots or bot groups can be excluded from accessing specific paths. Following robots.txt is voluntary — serious crawlers like Googlebot, GPTBot, or PerplexityBot comply with it; some more aggressive or malicious bots ignore it. Important: robots.txt only prevents crawling, not indexing — a page can still be indexed despite Disallow if other sites link to it.
SaaS Trial-Gen is a conversion strategy for Software-as-a-Service providers in which the advertising goal is not direct sale, but sign-up to a free trial access. Characteristics: low conversion threshold (often no credit card required), longer subsequent customer activation process, high number of trials of which typically only 5-15 percent convert to paying customers. CPA steering is demanding: too tight trial CPAs lead to insufficient volume; too wide ones to numerous unqualified trials. Strategically sensible: combine trial CPA with downstream activation rate and optimize for „Cost per Activated Trial" instead of just „Cost per Trial".
Schema.org is the vocabulary for structured data on websites maintained by the search engine consortium — the standard framework for semantically marking up content for search engines and AI search systems.
Search Console is Google's free tool for website operators to observe and steer organic Google visibility. Core functions: performance report (impressions, clicks, CTR, average position per search query and page), indexing status, technical crawl issues, Core Web Vitals data, manual action warnings, sitemap submissions, URL Inspection tool for real-time indexing checks. Important: Search Console shows only organic search data, no paid ads. Data typically has 2-3 days delay. Limitation: anonymized search queries reduce the complete picture, which becomes noticeable in long-tail analysis.
Server-Side Rendering refers to the architecture in which HTML content of a web page is pre-rendered server-side and delivered to the browser — central for SEO visibility of modern web apps.
Session recordings are recordings of actual user sessions on a webpage — mouse movements, clicks, scroll behavior, inputs are captured and can be played back later. Tools: Hotjar, Microsoft Clarity, Mouseflow, FullStory, LogRocket. Value: qualitative insight into where users get stuck, have comprehension problems, or take unexpected paths — data that quantitative analytics doesn't deliver. Privacy note: session recordings require consent under GDPR (see GDPR-compliant tracking). Text inputs in sensitive fields (password, credit card, personal data) must be technically masked — most tools offer configuration options for this.
A Single-Page Application is a web architecture in which the entire application runs in a single HTML page and content is loaded and rendered client-side via JavaScript.
Smart Bidding refers to automated bidding strategies in Google Ads that machine-learning-based optimize bids per auction in real-time — in contrast to manual bidding strategies.
Snippet suitability refers to the structural property of content to be usable in the form of short fragment citations in search engine answers, featured snippets, and AI-generated answers.
Source bias describes the structural tendency of search and AI systems to disproportionately cite established sources — Wikipedia, large news publishers, authoritative domains. New or smaller sites typically need 6 to 12 months of continuous substantial content production before they regularly appear prominently in AI Overviews or classical search results. Source bias should not be confused with quality assessment — it means that established sources have a visibility advantage even when content substance is comparable to or weaker than a newer source.
Subscription Linking is a Google program that allows news publishers to link the subscription status of their readers with their Google accounts. When a logged-in user performs a search and holds a subscription with a participating publisher, that publisher's content is preferentially shown in Google search, in AI Overviews, and in AI Mode. Activation requires a registration form with Google plus technical markup adjustments (Reader Revenue Manager, schema.org/CreativeWork with isAccessibleForFree property). Practically relevant only for news publishers with a paywall model — not applicable to other site types.
TanStack Start is a modern full-stack React framework with server-side rendering, type-safe routing, and integrated data loading — the new Lovable standard architecture since the April 2026 update.
Target CPA is a Smart Bidding strategy in Google Ads that controls bids so that conversions on average occur at a defined cost-per-action target value.
Target ROAS is a Smart Bidding strategy in Google Ads that controls bids so that a defined Return on Ad Spend target value is achieved — relevant primarily for e-commerce setups with different conversion values.
UTM parameters (Urchin Tracking Module) are URL appendages added to a target URL to uniquely identify a visitor's source for web analytics tools. Standard parameters: `utm_source` (source, e.g., „newsletter"), `utm_medium` (medium, e.g., „email"), `utm_campaign` (campaign, e.g., „spring-launch"), optionally `utm_content` and `utm_term`. With Google Ads often replaced by Auto-Tagging (`gclid` parameter), which allows deeper data linkage. Practice discipline: define unified UTM naming conventions and consistently adhere to them — otherwise scattered source designations like „Newsletter", „newsletter", and „NEWSLETTER" emerge in GA4 as separate entries.
Vite is a build tool and dev server for modern web applications, developed by Vue.js creator Evan You, free and open source. Characteristic: extremely fast Hot Module Replacement (HMR) in development because Vite uses native ES Modules in the browser instead of generating bundles. In production build, it bundles via Rollup. Vite is framework-agnostic — works with React, Vue, Svelte, vanilla JS. Adoption has grown significantly since 2021, now a common standard choice in frontend projects. Important: Vite alone produces a Single-Page Application with the associated SEO limitations.
A Vite SPA is a Single-Page Application built with the Vite build tool. Characteristic: the browser initially loads an almost empty HTML page plus a JavaScript bundle, which then renders the content in the browser. Advantage: very fast subsequent interactions without page reloads. Disadvantage: on first indexing, crawlers see practically empty HTML — which leads to systematic SEO problems. Until April 2026, Vite SPA was the standard architecture of Lovable, which made many Lovable sites invisible in search. Since the TanStack update, TanStack Start with server-side rendering is used instead.
The zero-click mechanic refers to the structural phenomenon that search queries are answered directly in featured snippets or AI answers without the user clicking on a source site.
If you're looking for substance on a specific performance, SEO or stack question — let's talk. We answer with the same clarity we write with.