Planning a Website Structure: Requirements, Process and Checklist
Contents
The guide as a PDF: process, special cases and checklist
18 pages in a handy format: every section of this guide, the checklist with all checkpoints and the special cases for listing pages and location pages. More practical than printing the page.
A website structure works when visitors reach their goal by the shortest path and when search engines and AI systems can assign every document unambiguously. The technical term is demand-oriented information architecture. This guide covers the requirements, the process and the checkpoints for the design.
Website structure starts with user intent, competition and business goals
Three inputs come before any structural decision. They determine which pages have to exist and how they relate to each other.
- User intent. Which questions prospects ask, at which stage of the decision and with which expectation of the answer. The source material comes from search queries, from prompts in AI systems, from internal site search and from sales.
- Competitive offering. Which page types better-ranking providers run, which topics they cover and which parameters they offer in their listings.
- Business goals. Which services are meant to grow, which audiences count, which locations and markets are served.
A structure designed without these three inputs rearranges existing content. A structure that includes them produces the pages that are actually in demand and places them where they get found.
Click path: the target depth is defined and maintained
The shortest path from the home page to a document is a decision. The more clicks a piece of content sits away, the fewer visitors reach it and the weaker the internal signal to search engines.
The target is set per page type and monitored afterwards:
- Pages with commercial value sit two to three clicks from the home page.
- Every main category is reachable directly from the main navigation.
- No document meant to generate entries sits deeper than four clicks.
- Orphan pages without an incoming internal link do not exist.
Click depth is measured in the crawl, per URL and per page type. If it rises after an expansion, the team corrects course before rankings give way. This check belongs in every cycle, outside a relaunch as well.
Principles that keep a structure viable for years
Eight principles apply regardless of industry and size. They decide whether a structure can absorb growth.
- Every document is a self-contained unit with a purpose, an audience and its own attributes. Anything without its own task belongs in another document.
- Every level offers a manageable choice. A few clearly distinguished options lead to the goal faster than a complete list.
- What matters is visible immediately. Whatever visitors look for most often appears on screen without interaction.
- Abstract terms come with examples. Category names and entry points carry concrete evidence so the choice works without guessing.
- Every page is a possible entry point. Most visitors arrive from search, directories or AI answers directly on a subpage. Each page therefore answers for itself where the visitor has landed and where to go next.
- Several paths lead to one piece of content. Hierarchy, taxonomy, search and contextual links serve different search strategies.
- Every document has exactly one focus. One topic, one page, one goal. Two pages on the same topic weaken both.
- The structure grows without rebuilding. New content finds its place in existing page types.
Website structure: page types, hierarchy, taxonomies, navigation and URL structure
Five components form the structure. Each one is designed separately, and in the end all five have to fit together.
Page types: recurring forms with a fixed build
A page type is a template with a defined task: service page, location page, guide, case study, product detail, listing, glossary entry. Each type has mandatory elements, a fixed order and its own metric.
Large websites typically run ten to thirty page types. Fewer forces different content into the same shape; more becomes impossible to maintain editorially.
Hierarchy: the sitemap as a tree with clear levels
The hierarchy orders page types from general to specific. Every level answers a question of its own and leads to the next.
Levels that only forward belong removed. They cost a click and carry no statement of their own.
Taxonomies: attributes across the hierarchy
Categories order hierarchically, and every piece of content has exactly one place. Taxonomies order by attributes such as industry, use case, region or audience, and one piece of content can carry several attributes.
Large websites need both: the tree for orientation, the attributes for retrieval.
URL structure: short, descriptive, stable
The URL structure reflects the hierarchy, stays short and descriptive and contains no years. Every URL is set up to last, because each change means redirects and loss of signal.
Implementation in the CMS: page types, taxonomies and menus
The planned structure is rebuilt inside the CMS, otherwise it stays a document. In WordPress, page types map to post types and templates, taxonomies to categories and tags, the hierarchy to page assignment and the permalink path. In TYPO3, page types, categories and the page tree hierarchy do the same job. What counts is that an editor can create a new page without thinking about where it belongs.
The menus follow the hierarchy. The main menu in the header carries the levels that hold demand. The footer menu takes in mandatory pages and secondary topics that nobody searches for, yet everybody finds when they need them. Both menus are maintained in the CMS and not hard-wired into the template, so a structural decision stays implementable without a developer.
Before going live, check whether search engines and AI systems see the same structure as the visitor: XML sitemap from the CMS, internal linking from the templates, breadcrumb from the hierarchy. Where menu, sitemap and linking drift apart, Google evaluates a structure other than the planned one.
User research: personas, mental models and customer journey
A structure reflects how the target audience orders a subject area. That order regularly differs from the internal view.
Three forms of research supply the basis:
- Personas. Who decides, who researches, who implements. Vocabulary, depth of questions and decision criteria differ by role.
- Mental models. Which order users expect and which terms they work with. The source material comes from interviews, internal site search and support requests.
- Customer journey. Which paths lead to the goal, through which channels visitors enter and at which point they need comparative information.
The result is a list of tasks the website has to fulfil. That list later becomes the basis for user testing.
Inventory: content inventory and content audit
A complete record of the existing material comes before the design. The inventory lists every object: pages, files, images, videos, documents, with URL, page type, language and ownership.
The audit rates every entry against metrics: rankings, clicks, age, currency, internal links, duplicates, conversion contribution. Each piece of content then carries a decision.
- Keep – the content does its job.
- Revise – the demand exists, the execution falls short.
- Merge – several pages serve the same intent.
- Remove and redirect – the content has neither demand nor commercial value left.
Gaps become visible in the same step: questions from demand research with no document behind them.
The category structure reflects your own range, and it adapts to demand
In shops, portals and dealer websites the category tree decides the largest share of organic visibility. It grows over years and at some point follows the database.
Six questions decide on every category:
- Does the category match a combination that is genuinely in demand, evidenced by search volume and prompts?
- Does it carry enough items so the page looks neither empty nor unmanageable?
- Does it compete with another category for the same intent?
- Does its name come from the language of the audience or from internal product vocabulary?
- Do better-ranking competitors run parameters that are missing here?
- Is the category still accurate after a change in the product range?
The result is a decision per category: confirm, merge, rename, create or remove from the index. On top of that comes reporting at keyword and URL level that keeps the effect of each decision visible.
Filter combinations with demand of their own deserve an indexable listing of their own. Combinations without demand stay out of the index, because they consume crawl budget and create duplicates.
Naming: the terms the audience actually uses
Every label in navigation, headline and breadcrumb is a decision about findability. Internal product names, department labels and invented words cost visitors, because nobody searches for them.
A naming system defines which terms apply, which synonyms exist and how variants are written. It governs navigation, headings, anchor texts, filter names and structured data alike.
Card sorting puts it to the test: users group content by their own understanding and name the groups. The result names the expected order and the terms the audience actually uses.
Navigation types: main, breadcrumb, footer, contextual, faceted and search
Navigation makes the structure operable. Six forms serve different tasks and complement each other.
- Main navigation. Leads to the main areas, stays consistent across pages and covers the most important entry points.
- Breadcrumb. Indicates the position in the hierarchy and offers the way back. It is additionally marked up as BreadcrumbList.
- Footer navigation. Carries service, legal and orientation pages plus entry points with no room in the main navigation.
- Contextual navigation. Points from the text to deeper documents and related services, where the question arises.
- Faceted navigation. Filters large inventories by attribute. Demand decides which combinations are indexable.
- Internal search. Catches whatever the structure fails to cover. Its queries are the most honest source for missing content and wrong labels.
On small screens every level costs space. Progressive disclosure works there: the essentials visible immediately, details on demand, with no content lost.
Internal linking: one anchor text leads to exactly one target
Internal links carry meaning. Every link tells the machine what the target page is about and the visitor why the click is worth it.
Four rules decide the effect:
- An anchor text belongs to exactly one target. The same anchor text pointing at two different pages devalues both.
- The anchor text names the focus topic of the target document or a synonym of its intent. Generic phrasing carries no meaning.
- The sentence around the link means precisely the term in the anchor, and the target document answers precisely that question.
- Links sit where they contribute something in substance.
How many internal links a page receives is a decision. Pages with commercial value get more, secondary pages fewer. A linking matrix records which anchor text leads to which target and remains the basis for editorial work for years.
Structured data: the meaning layer above the structure
The hierarchy orders documents. Structured data based on Schema.org explains what those documents are and how they connect.
Per page type the team defines which type applies and which properties are maintained:
- Organization with name, address, profiles and areas of expertise, defined once and referenced across all pages.
- Service for service pages, with provider, service area, audience and offer catalogue.
- LocalBusiness for location pages, with address, geo coordinates, opening hours and catchment area.
- Article for guides, with author, date and topical reference.
- BreadcrumbList for the position in the hierarchy.
- FAQPage for question-and-answer blocks.
The connection is what counts: every entity gets its own stable identifier, and all pages reference the same one. Individual statements turn into a coherent model of the company.
Chunking and citability: building documents for passage retrieval
Generative systems process passages. They split a document into sections, rate each one on its own and take the section that answers the question asked.
That puts requirements on every single document:
- The answer comes first, the explanation after it.
- Every heading names the topic of the section and its core statement.
- Every section remains understandable when quoted alone. A section that needs its predecessor does not get picked up.
- Lists, tables and question-and-answer blocks raise the probability of being selected as a passage.
- Facts appear with figure, period and source in the same section.
- Abbreviations are spelled out at first occurrence in each document.
The same requirements improve readability for people. Short main clauses, familiar terms, and technical vocabulary only where it adds precision.
Entities: unambiguous naming beyond your own domain
AI systems answer questions from several sources and weight what independent third parties say. A brand described identically everywhere is recognised as one entity. Diverging names, services and data break apart into several unclear records.
The structure therefore extends beyond the website: the same terms for the same services, the same facts about company and locations, the same relations between both, across directories, profiles, trade media and comparison platforms.
A facts page with the core data and an llms.txt in the root directory name the authoritative content on top of that. Both take little effort and make assignment easier.
Special cases: listing pages in retail and location pages
Two constellations follow rules of their own, because they carry content in large numbers and in constant change. Both decide the largest share of organic visibility in their industries.
Listing pages: the listing carries the visibility
In e-commerce and vehicle retail a search for available items lands on a listing. The individual item is sold tomorrow; the listing remains.
Four points are defined per listing:
- Selection. Which combination of category, brand, model, condition or region the listing covers, evidenced by demand.
- Inventory size. How many items it has to contain so the listing looks neither empty nor unmanageable, and what happens below that threshold.
- Item preview. Which details each card carries: price, availability, year, mileage, specification or delivery time.
- Competitor parameters. Which filters better-ranking providers run as listings of their own and are missing from the current inventory.
Filter combinations with demand of their own get an indexable listing with its own title, description and introduction. Combinations without demand stay out of the index.
Lifecycle management: sold items redirect to the most valuable listing
A sold item has rankings, incoming links and visitors. A 404 erases all of it at once.
The redirect leads to the most valuable listing the item belonged to. A sold estate car of a particular model leads to the listing of available estate cars of that model. The visitor lands on a genuine alternative, and the listing inherits the signals of the individual document.
Which listing is the most valuable follows from demand and rankings. The mapping is defined once and automated afterwards, because it applies again with every sale.
Location pages: complete location data across the entire structure
A location page carries the complete location data, the offering of that location and the catchment area it actually serves. Swapped place names on identical text produce no visibility.
In structured data that means: LocalBusiness with address, geo coordinates, opening hours and phone number, areaServed for the catchment area, hasOfferCatalog for the services of that location and Offer for offers that apply there. Every location page references its Google Business Profile and the directory entries of that location.
Consistency of location data across the entire structure is what counts. Name, address and phone number appear identically in the on-page text, in the structured data, in the business profile, in directories and on social profiles. Deviations split one location into several unclear records, and with many locations the damage multiplies. Maintenance therefore runs through one record per location that feeds every channel.
Process: audit, benchmark, demand, design, implementation
Audit: record the existing material completely
A crawl records every URL with status code, page type, click depth, internal links, titles and headings. Metrics from Search Console and analytics per URL come on top.
The result is a table in which every page carries a decision.
Benchmark: measure the competition
Three to five competitors are recorded against the same criteria: page types, category depth, volume per page type, internal linking, structured data, visibility in search engines and in AI answers.
That shows which gaps exist and which of them structure can close.
Demand: collect keywords, questions and prompts
Demand is collected per topic and decision stage, from search volume, from questions in AI systems, from internal site search and from sales.
Every intent gets exactly one document. Two pages for the same intent compete with each other.
Design: draw the structure and test it with users
The design consists of a tree diagram, a page-type table, a content outline per type, a linking matrix and a schema plan.
It gets tested with users before anything is built.
Implementation: build in priority order and measure
Implementation follows the order of expected effect on enquiries and revenue. Every URL that disappears gets a redirect to the closest new address in substance.
After the rebuild the team checks whether rankings, clicks and enquiries arrive at the new addresses.
Testing with users: card sorting, tree testing and usability testing
Three methods examine different questions and are used in this order.
- Card sorting clarifies before the design which order users expect and which terms they use.
- Tree testing examines the designed structure without design: users receive a task and navigate the plain text structure. The measurement covers whether they arrive, how directly and where they turn off.
- Usability testing examines the built page with realistic tasks. It records success rate and duration alongside the participants’ reasoning.
Two measures are kept apart: findability describes whether users reach content they were looking for. Discoverability describes whether they come across content they were not looking for but that helps them.
Metrics: from click depth to mentions in AI answers
A structure is judged against four groups of metrics. The comparison runs with the state before the rebuild.
- Structure. Click depth per page type, number of orphan pages, number of pages without entries, distribution of internal links.
- Usage. Success rate on defined tasks, time to goal, use of internal search, drop-offs on the most important paths.
- Search. Rankings and clicks per page type, number of pages with entries, index coverage, development after redirects.
- AI answers. Mentions and citations per topic, share of own sources in the answers, consistency of statements across platforms.
Whatever stands out goes into the backlog of the next round.
Checklist for the finished website structure
Every point gets checked before go-live. The list follows the components of the structure.
Basics
- User intent per topic and decision stage is collected and documented.
- Three to five competitors are recorded against the same criteria.
- Business goals per service, audience and location are stated.
- Personas and mental models are available.
- The task list for later user testing is ready.
Inventory
- The content inventory records all pages, files, images, videos and documents.
- Every piece of content carries a decision: keep, revise, merge, remove.
- Content gaps are named per intent.
- Duplicates and competing pages are merged.
- Every category carries a decision: keep, merge, rename, create or remove from the index.
Structure and hierarchy
- Target depth per page type is defined and maintained.
- Pages with commercial value sit two to three clicks from the home page.
- There are no orphan pages without an incoming internal link.
- Every level answers a question of its own; pure forwarding levels are removed.
- Every page type has a purpose, mandatory elements, an audience and a metric.
- Every intent has exactly one document.
- The structure absorbs new content without requiring exceptions.
Naming and navigation
- All labels come from the language of the audience.
- The naming system records terms, synonyms and spelling variants.
- Card sorting has confirmed order and terms.
- The main navigation covers the most important entry points and stays identical everywhere.
- Breadcrumbs exist on all subpages and are marked up as BreadcrumbList.
- Footer navigation carries service, legal and orientation pages.
- Contextual links sit where the question arises.
- Indexable filter combinations are defined on the basis of demand.
- Internal search is evaluated and feeds back into the structure.
- Every piece of content stays reachable on small screens.
URLs and redirects
- The URL structure reflects the hierarchy, stays short and contains no years.
- Every URL that disappears has a redirect to the closest new address.
- Redirect chains are resolved.
- The list of old URLs with rankings exists before the rebuild.
Internal linking
- Every anchor text leads to exactly one target.
- Every anchor text names the focus topic of the target document.
- The sentence around the link means precisely the term in the anchor.
- Pages with commercial value receive more internal links than secondary pages.
- The linking matrix is documented and maintained.
Technology and structured data
- The schema type per page type is defined and implemented.
- Every entity has a stable identifier that all pages reference.
- The statements in the schema match the visible content.
- Every image carries a descriptive file name, alternative text and title.
- Videos carry a transcript and markup.
- The loading order delivers what matters first.
- Every piece of content is reachable without JavaScript.
Individual document
- The answer comes first, the explanation follows.
- Every heading names the topic and core statement of the section.
- Every section remains understandable when quoted alone.
- Facts appear with figure, period and source in the same section.
- Abbreviations are spelled out at first occurrence.
- Every document has exactly one focus and one goal.
Listing pages and location pages
- Selection, minimum inventory, preview details and filter parameters are defined per listing.
- Indexable filter combinations are evidenced by demand.
- Every indexable listing carries its own title, description and introduction.
- The target listing per category is defined and automated for sold items.
- Every location page carries substance of its own.
- LocalBusiness, areaServed, hasOfferCatalog and Offer are maintained per location.
- Name, address and phone number match across all channels.
- Every location page references its business profile and directory entries.
Beyond your own domain
- Name, address and phone number match on-page, in the schema, in the business profile and in directories.
- The same terms apply to the same services, in advertising and trade articles as well.
- A facts page summarises the core data.
- The llms.txt names the authoritative content.
Testing and measurement
- Tree testing has confirmed the designed structure.
- A usability test with realistic tasks has been carried out.
- Findability and discoverability are measured separately.
- Click depth, success rate, rankings and mentions in AI answers are recorded as baseline values.
- The measurement date after the rebuild is fixed.
Mistakes that hold a website structure back
- The structure follows the org chart. Departmental boundaries are of no interest to visitors.
- Important content sits too deep. Every additional click costs visitors and internal signal.
- Several pages serve the same intent. They compete for the same keyword and weaken each other.
- Naming comes from internal usage. Nobody searches for it.
- The same anchor text leads to different targets. Both targets lose clarity.
- Filter combinations without demand are indexable. They create duplicates and consume crawl budget.
- Structured data comes last. The connections that make the difference are then missing.
- The structure is tested without users. Internal plausibility is no substitute for a test.
- Nothing is measured after the rebuild. Without baseline values every effect remains an assertion.
Deliverables: structure document, content outline, linking matrix
At the end there are handover documents that editorial and development teams can work with.
- Structure document with tree diagram, page-type table, taxonomies and target depth per type.
- Content outline per page type with mandatory elements, heading logic and answer sentences.
- Linking matrix with anchor texts, sources and targets.
- Schema plan with types, properties and identifiers per page type.
- Redirect plan for every URL that disappears.
- Prioritised backlog with expected effect and metric per measure.
Website structure and AI answers: the outline decides citability
Generative systems select passages, resolve entities and weight what independent sources say. A clear structure works at all three points.
The hierarchy makes it recognisable which document is responsible for which topic. The markup connects the documents into a model. The document build supplies the passages that can be cited.
Structure, markup and document build therefore belong in the same design.
Frequently asked questions about website structure
How many levels should a website have?
As few as possible without losing orientation. What counts is whether every level answers a question of its own. Levels that only point to the next one belong removed.
How many clicks may an important document sit from the home page?
Two to three clicks have proven effective for pages with commercial value, and four at most for every page meant to generate entries. What counts is that the target is defined and monitored in the crawl.
What is the difference between website structure and navigation?
The structure is the model: page types, hierarchy, taxonomies and rules. Navigation makes parts of that model visible and operable. A website can have several navigations and still carries only one structure.
What is card sorting and when is it used?
In card sorting users group content by their own understanding and name the groups. It is used before the design and supplies the order and the terms the audience expects.
What is tree testing and what does it measure?
Tree testing examines the designed structure without design. Users receive a task and navigate the plain text structure. The measurement covers whether they arrive, how directly and at which point they turn off.
How can a website structure be presented?
As a tree diagram with page types and levels, supplemented by wireframes for the most important types and user flows for the most important paths. Those three representations together form a concept document for editorial, design and development.
How are categories checked in a shop?
For every category, demand, item count, internal competition, naming, competitor parameters and the current product range are gathered. Each category then carries a decision: confirm, merge, rename, create or remove from the index.
What role does structured data play for the structure?
It forms the meaning layer above the hierarchy. Per page type the team defines which schema type applies and which properties are maintained. A stable identifier per entity connects the documents into a model of the company.
How does the structure affect visibility in AI answers?
Generative systems split documents into passages, resolve entities and weight independent sources. Clear responsibilities per document, unambiguous naming, structured data and consistent statements beyond the website raise the probability of being cited as a source.
What happens to sold items in shops and vehicle retail?
They redirect to the most valuable listing they belonged to. Rankings, incoming links and visitors are preserved, and the visitor lands on a genuine alternative. The mapping is defined once and automated afterwards.
How are many locations represented in the structure?
Through a page type of its own with a consistent build and location-specific content. Every location page carries complete location data, the offering of that location and the catchment area, and it matches the entries in directories and map services.
When is a structural rebuild worth it?
When content exists several times over, when important pages generate no entries, when navigation has been extended with exceptions over the years, or when a relaunch is due. In all other cases incremental improvement is enough.
Can the structure be improved without a relaunch?
Yes. Merging, redirecting, relinking and catching up on markup can all be done within the existing site without putting rankings at risk.
Which metrics prove that the structure works?
Click depth per page type, number of pages with entries, success rate and duration on defined tasks, rankings and clicks per page type, plus mentions in AI answers. The comparison runs with the state before the rebuild.
