SEO: 12 Content-Structure Practices That Still Rank
Classic SEO, structured for 2026: twelve on-page and site-architecture practices that make a page easy for people to read, for Google to crawl and rank, and for AI search to reuse.
Structure
Clear hierarchy
One H1, planned H2s
- When it helps
- The page grew paragraph by paragraph, so three sections overlap and the main topic is nowhere in a heading.
- What to do
- Plan it in a content brief before writing: one main topic, broken into distinct subtopics in a logical order, each with its own heading and the proper tag level (H1, H2, H3), so each subtopic covers one aspect.
- Why it works
- A clear hierarchy tells a search engine, and a model, what a page is about before it reads a word of body copy; it also makes the page easier to update without orphaning sections.
- Example
- H1 "AI agent development services" → H2 "What is an AI agent?", H2 "How the build works" with H3 "Discovery", H3 "Build", H3 "Launch", then H2 "Pricing" and H2 "FAQ". One subtopic per heading, decided in the brief.
Structure
Headings that carry the point
Summarise, then explain
- When it helps
- Headings say "Overview", "Details" and "More", so neither a scanning reader nor a crawler learns anything from them.
- What to do
- Summarise the key point of each section in its heading, use real heading tags for them (never styled text), and place the relevant term in the heading where it fits naturally.
- Why it works
- Headings help scanning readers find information, give assistive technology a page outline, and signal relevance to search engines; for models they act as semantic anchors for each section.
- Example
- Before: H2 "Overview" After: H2 "Onboarding takes two working days" The heading states the section's point, so a reader and a crawler both get it without the body.
Structure
Paragraphs and bullet points
Short blocks, whitespace
- When it helps
- A 1,200-word page is four walls of text, and on a phone nobody reads past the first.
- What to do
- One idea per paragraph in three to four sentences, bullet points of one or two sentences for lists and summaries, whitespace between blocks, and consistent typography across devices.
- Why it works
- Readable, scannable pages increase engagement and time on page, reduce friction on mobile, and give models clean blocks to extract.
- Example
- A 900-word service page: eleven paragraphs of three to four sentences, two bullet lists ("what is included", "what is not") with one sentence per point, and a line of space between every block.
On-page
Keywords placed with intent
Title, headings, body
- When it helps
- The page never says the phrase people search for, or says it twenty times in a row.
- What to do
- Place the target term in the title, the H1, an early paragraph and the headings where it fits; use natural variations to cover more queries; align the wording with what the audience actually searches; never stuff.
- Why it works
- Keyword prominence (headings, early placement) signals relevance to specific queries; variations add depth; stuffing reads as spam and works against the page.
- Example
- Target "AI agent developer in Barcelona": in the title tag, the H1, the first paragraph and one H2; variations ("build an AI agent", "LLM agent development") in the body; the exact phrase never more than a handful of times.
On-page
Prepared images and media
Named, compressed, described
- When it helps
- Screenshots are 4 MB files called IMG_4812.jpg with no alt text, and the page takes six seconds to load.
- What to do
- High-quality images, compressed without visible loss, sized for the display, with descriptive file names and alt text, captions where they add context, and an image sitemap so search engines find them.
- Why it works
- Search engines value a fast, accessible page; alt text and file names tell them what the image shows and support visually impaired users.
- Example
- File: ai-trading-assistant-dashboard.jpg (not IMG_4812.jpg), 1600 px wide, under 200 KB, alt "AI Trading Assistant dashboard with the trade-rules checklist", listed in the image sitemap.
On-page
Readable URLs and breadcrumbs
The path says the topic
- When it helps
- Pages live at /p?id=482, and visitors who land deep in the site cannot tell where they are.
- What to do
- Readable URLs that contain the target term, a breadcrumb trail on every inner page, and schema or structured data (BreadcrumbList, Article) that matches what is visible, to aid crawling.
- Why it works
- A hierarchical architecture with readable URLs, breadcrumbs and structured data gives clearer indicators about what each page means and how pages connect, which improves crawlability and visibility.
- Example
- URL: /portfolio/ai-trading-assistant/ (not /p?id=482). Breadcrumb: Home › Portfolio › AI Trading Assistant, marked up as BreadcrumbList so the path shows in results.
Architecture
Internal links with real anchors
Authority to key pages
- When it helps
- Every link says "read more", and the most important page gets fewer internal links than the privacy policy.
- What to do
- Link related pages within the site with descriptive anchor text, point more internal links at the pages that matter most, and organise content into categories and subcategories (content silos) so topical relationships are explicit.
- Why it works
- Internal links help search engines understand the structure and the importance of each page, distribute authority, and make it easy for users to move between relevant pages.
- Example
- From the blog post on RAG: "see the crypto regulation RAG chatbot case study", linking to the project page; from every case study, one link back to the service it belongs to. The service pages end up with the most internal links.
Architecture
External links with intent
Cite, nofollow, maintain
- When it helps
- Claims have no sources, a vendor link passes authority you did not mean to give, and a third of the links are dead.
- What to do
- Link to reputable sources where they add value, apply rel="nofollow" to external links you do not want to pass authority to, and check regularly for broken or outdated links, fixing or removing them promptly.
- Why it works
- External links signal well-researched content connected to reliable references; broken links harm user experience and SEO.
- Example
- Cite the KIME guide and the Princeton GEO study with normal links; put rel="nofollow" on a link to a tool you mention but do not endorse; run a quarterly broken-link check and fix or remove whatever returns 404.
Architecture
Site architecture and sitemap
Silos, sitemap, audits
- When it helps
- Pages are added wherever is convenient, the sitemap is a year old, and nobody knows which pages are orphans.
- What to do
- Organise pages hierarchically into categories and subcategories, keep an up-to-date XML sitemap so search engines can crawl and index everything, and run regular site audits for broken links and pages with poor crawlability.
- Why it works
- A well-structured site is easier for bots to crawl and index and for users to navigate; grouping related pages into silos makes topical relationships explicit for crawlers and models alike.
- Example
- Silo: /materials/ → /materials/aeo/, /materials/geo/, /materials/kpis/, /materials/seo/, all linked from the hub and from each other; the sitemap lists every page with its language alternates; a quarterly crawl finds orphans and broken links.
Technical
Mobile-friendly, fast pages
Responsive, readable, light
- When it helps
- The desktop page looks fine, but on a phone the text is tiny, the images are heavy and the layout needs endless scrolling.
- What to do
- A responsive layout for every screen size, a clearly readable font size, media prepared for mobile, a layout that works without excessive scrolling, fast loading, and a test on real phones and with Lighthouse before each release.
- Why it works
- Mobile usage keeps growing and mobile users expect quick access; long, unstructured blocks and slow pages lose them, and search engines reward the better experience.
- Example
- This site: a responsive layout, 16 px base text, images optimised to several sizes, technique cards that stack in two columns on a phone, a 170 KB JavaScript budget on the home page, and a Lighthouse run on the release checklist.
Technical
Periodic audit with SEO tools
Crawl, rank, links, rivals
- When it helps
- The site was optimised once, two years ago, and nobody has looked at rankings, backlinks or broken links since.
- What to do
- Audit on a schedule with the right tool per job: keyword research (Semrush, Ahrefs, Google Keyword Planner), rank tracking (Moz, Serpstat), backlinks (Majestic, Ahrefs), site crawls (Screaming Frog, Sitebulb), competitors (SpyFu, SimilarWeb); tools surface the problems, judgement fixes them.
- Why it works
- A periodic content audit is the simplest way to keep pace with the changing needs of users, search engines and LLMs; it is how you find the broken links, duplicate titles and lost rankings before they cost traffic.
- Example
- Quarterly: a Screaming Frog crawl for broken links and duplicate titles, Ahrefs for lost backlinks, Search Console for queries that gained or lost clicks; the pages that lost ground go first into the content audit.
On-page
Search intent before length
Match the query, not a word count
- When it helps
- A 3,000-word page targets a query that wants a price range, and a 200-word page targets one that wants a full comparison.
- What to do
- Work out the goal behind each search term first, then match the page type to it: a definition, a comparison, a price range, a how-to; write genuinely useful content for that intent and stop when it is answered.
- Why it works
- Search engines prioritise relevance and intent; clarity matters more than length, and well-structured content often outperforms longer, poorly organised pages in both search and AI answers.
- Example
- "AI agent cost" is a research query: a page that explains the cost drivers and typical ranges ranks; a page that only says "contact us for a quote" does not. Decide the page type from the intent before writing a word.
Pick a technique on the map to see when it helps, what to do and why it works.
Source: Victoria Olsina, "Content Structure for SEO and LLMs (2026): How to Get Cited" (first published Aug 2023, updated Oct 2026): the four SEO outcomes, the best-practice list, the SEO vs AI search comparison and the tools list. This map claims no figures; the AI-search side of the same post is in the AEO and GEO maps.
Content structure for SEO vs AI search
Traditional SEO and AI search overlap but reward different outcomes: SEO asks whether a page can be found, AI search asks whether a section can be lifted out and trusted on its own. One page has to serve both.
| Priority | Traditional SEO | AI search |
|---|---|---|
| Main signal | Crawlability and relevance | Extractability and clarity |
| Rewards | Depth and coverage | Precision and self-contained blocks |
| User behaviour | Visits the page | Often gets the answer without a click |
| Structure job | Hierarchy and keyword clarity | Quotable, standalone sections |
Source: Victoria Olsina, "Content Structure for SEO and LLMs (2026): How to Get Cited" (updated Oct 2026). Modern structure must support both: readable and well organised, and explicit enough to stand alone when reused by AI.