The complete guide to website + SEO in 2026

The website + SEO conversation has changed more in the last 24 months than in the previous decade. Google's AI Overviews have absorbed roughly 35-45% of click-throughs on informational queries. AI answer engines — ChatGPT Search, Perplexity, Claude — are now meaningful traffic sources in their own right, with citation behavior that doesn't map cleanly onto the ranking factors we optimized for in 2022. The fastest sites in 2024 are the slow sites in 2026, because Core Web Vitals thresholds tightened and the median competitor's LCP dropped under 1.5 seconds. Headless commerce went from novel to expected for any serious DTC brand.

Most SEO advice you'll read in 2026 is still optimizing for the 2018 reality of "keyword research → write 1,200 words → wait for Google." That playbook works for low-competition long-tail and basically nothing else. The agencies and freelancers selling it aren't lying — they're just running a strategy that worked five years ago against an algorithm that's moved on. This guide is what we actually build for retainer clients in 2026: the technical foundation, the content architecture, the AI-search optimization layer, and the measurement framework that ties it all to revenue.

If you walk away with one thing: website + SEO in 2026 is a structural engineering problem with a topical-authority overlay, not a keyword-stuffing problem. The technical floor matters more than ever. The content moat is harder to build than ever. The agencies still pitching "we'll write 30 blog posts a month" are selling you content that won't rank because the authority underneath isn't there.

The four things that actually matter (and the dozen that don't anymore)

A serious 2026 SEO strategy concentrates on four areas. Everything else is either downstream of these four or doesn't matter the way it used to.

1. Technical foundation. Sub-1.5s mobile LCP. Server-side rendering or static generation, never client-side-only for content pages. Proper canonical handling. Clean URL structure. Schema markup on every meaningful page. Sitemap split by content type. Internal linking architecture that distributes authority intentionally. This is the floor — get this wrong and nothing else compounds.

2. Topical authority through depth. Pillar pages anchored by clusters of supporting articles, all interlinked. One pillar plus 8-12 supporting articles on a specific topic outperforms 50 disconnected blog posts on random topics. This is how new sites compete with established domains: by going deeper than the incumbents bother to, on topics that match their service exactly.

3. AI search optimization (GEO). Getting cited inside AI answers — ChatGPT, Claude, Perplexity, Google AI Overviews — is the new ranking layer. The signals overlap with traditional SEO but aren't identical. Declarative opening claims, structured FAQ schema, llms.txt, citation-friendly content patterns. We'll cover this in depth.

4. E-E-A-T signals (Experience, Expertise, Authoritativeness, Trustworthiness). Named authors with credentials. Real bios with verifiable sameAs links to public profiles. First-person experience in writing rather than research-compiled-from-other-sources. Schema-marked Organization data with founder identification. Google's Search Quality Rater Guidelines lean heavily on E-E-A-T; AI engines lean even harder on it.

The dozen things that don't matter the way they used to:

• Meta keywords tag. Dead since 2009 but somehow still on agency checklists.
• Keyword density. Hasn't been a ranking factor since 2013. Still gets analyzed in every "audit."
• Backlink quantity. Quality dominates. Ten links from category-relevant high-authority domains outperform 500 from PBNs.
• Exact-match keyword anchors. Over-optimization signal. Natural varied anchors win.
• "Long-tail keywords" as a primary strategy. Long-tail still works — but standalone, without topical authority underneath, it's diminishing returns.
• Hashtags on the site. Not a thing. Never was.
• Word count thresholds. "Blog posts must be 1,500+ words." False. The right length is the length the topic actually needs.
• Submission to directories. Largely irrelevant outside of category-specific ones (Yelp for restaurants, Avvo for lawyers, GMB for everyone).
• Anchor-text-only PageRank thinking. Modern Google distributes authority across many signals, not just anchor text.
• Domain age. Has minor weight; new domains can outrank old ones with better signals across every other dimension.
• Updating "last modified" dates without changing content. Search Console catches this. It's a negative signal now.
• "SEO content" as a separate genre. Content written for search engines reads like content written for search engines. Both users and the algorithm can tell.

Audit any agency proposal against this list. If they're pitching the dead column as their playbook, they're running 2018 in 2026.

The technical floor (where most sites fail)

Most "marketing sites" in 2026 are technically broken in ways that cap their SEO ceiling regardless of how good the content is. The five most common issues we see when we audit prospective clients:

LCP above 2.5 seconds on mobile. Google's Core Web Vitals thresholds tightened in 2024 — 2.5s is now the pass threshold, with serious ranking penalty for sites above it. The median competitor we audit in 2026 has an LCP of 1.3-1.8s. If your site is at 3+ seconds, you're losing ranking position to faster competitors even if your content is better.

Server-side rendering missing on content pages. Many sites built on React-without-SSR (CRA, Vite-only) ship a near-empty HTML shell that depends on JS to render content. Googlebot does eventually render these pages, but with significant delay and resource cost — and AI answer engines crawl less aggressively, often only seeing the empty shell. Static generation (SSG) or server-side rendering (SSR) is non-negotiable for content pages.

No schema markup or wrong schema markup. A surprising number of professional marketing sites have zero schema.org JSON-LD beyond what their CMS template ships with. Organization, LocalBusiness, Service, FAQPage, Article, BreadcrumbList — these are the basics. Without them, you're invisible to rich results, knowledge panels, and AI answer citations.

Broken canonical handling. When canonical tags conflict with URL structure or hreflang, Google sees inconsistent signal and discounts the page. We see this most often on sites with multiple URL formats for the same content (trailing slash issues, capitalization variants, URL parameter pollution).

Sitemap not split or not submitted. At 5,000+ URLs, a single sitemap.xml causes crawl-priority issues. Splitting by content type (pages, blog, products, etc.) and using a sitemap index gives the crawler much better signal about what to prioritize.

We audit these on every new engagement. Fixing them costs nothing in content effort and typically lifts organic traffic 15-30% in the 90-day window after the fixes ship.

The stack: Next.js vs Astro vs Webflow vs WordPress

The stack decision matters because it determines your floor on every technical dimension above. The serious options in 2026:

Next.js. Our default for marketing sites that need real performance + flexibility. App Router, React 19, Turbopack-fast builds, SSG for content + SSR for dynamic pages. Pros: a very high performance ceiling, excellent SEO defaults, large talent pool, deploy-anywhere (Vercel default, Cloudflare or self-hosted also fine). Cons: requires real engineering — not a no-code platform. Right for: serious marketing sites, content-heavy sites, agency clients with growth ambition.

Astro. Excellent alternative to Next.js for content-only sites. Ships almost no client-side JavaScript by default, which means LCP and INP are extraordinary. Pros: best performance ceiling of any framework. Cons: thinner ecosystem, harder to layer interactive components on later. Right for: content-only marketing sites where you're confident there won't be heavy app-like features later.

Webflow. No-code platform that's good enough for most small business marketing sites. Visual editor, decent performance, reasonable SEO defaults. Pros: editable by non-engineers, faster initial build, no engineering team needed for content updates. Cons: performance ceiling lower than Next.js or Astro, hits scaling limits at 1,000+ pages, lock-in to Webflow hosting. Right for: 5-30 page marketing sites without complex requirements.

WordPress. Still the largest market share globally, but its dominance is declining. Pros: largest plugin ecosystem, familiar to many content teams, vast pool of WordPress developers. Cons: performance is hard to make excellent (most WP sites have LCP > 3s without aggressive optimization), security maintenance burden, plugin sprawl introduces vulnerabilities and bloat. Right for: businesses with existing WP investment and content teams trained on it; not our default recommendation for new builds.

Our standing recommendation for most marketing sites is Next.js on Vercel because the performance ceiling is highest, the SEO defaults are excellent, and the platform handles deployment, edge functions, image optimization, and CDN without configuration overhead. We build mediapromotions.net on this stack — including this article you're reading.

For e-commerce, the stack decision is different:

• Shopify (off-the-shelf theme): $3-8k builds, suitable for under-$1M GMV brands. Use a fast theme (Dawn, Sense), avoid plugin bloat.
• Shopify (custom theme): $8-25k. Right for $1-5M GMV brands with brand-specific design needs.
• Shopify Hydrogen + headless: $40-100k+. For brands above $5M GMV who want full design control and performance ceiling.
• WooCommerce: rarely the right answer in 2026 outside specific cases.

Schema markup that actually moves visibility

Schema markup is one of the highest-ROI engineering investments you can make on a content site. It's also one of the most commonly skipped. Here's the priority order:

Organization + WebSite + SearchAction — on the root layout. These get you into knowledge panels, sitelinks search box, and AI answer citations. Non-negotiable.

LocalBusiness (and specific subtypes) — on local-presence pages. ProfessionalService for agencies, MedicalBusiness for healthcare, LegalService for law firms, FoodEstablishment for restaurants. Use the specific subtype, not generic LocalBusiness, when applicable.

Service — on service-offering pages. Includes provider, areaServed, hasOfferCatalog. We mark every service page on this site with the appropriate Service schema.

FAQPage — on any page with a Q&A section. This is one of the highest-leverage schema types because it gets pulled into AI answers and into the "People also ask" rich results.

BreadcrumbList — on every non-home page. Tiny effort, real signal to both Google and AI engines about page hierarchy.

Article (with author Person reference) — on every content article. The author Person reference, with sameAs links to public author profiles, is critical E-E-A-T signal.

Product — for e-commerce.

Review and AggregateRating — for products and services with real reviews. Only with actually verified reviews; fake review markup is a Manual Action waiting to happen.

Event — for any time-bound thing on the site.

What doesn't matter much: HowTo schema (deprecated for most categories), Recipe schema (only relevant if you're a recipe site), exotic types Google doesn't render rich results for.

We audit schema implementation on every prospect. The most common pattern is "Organization schema only, ten years out of date, with broken sameAs URLs." Fixing schema across an existing site typically yields a 10-20% organic visibility lift in the 60-day window, before any new content ships.

Generative Engine Optimization (GEO) — the new SEO layer

Here's the structural shift most agencies haven't internalized. AI answer engines — Google's AI Overviews, ChatGPT Search, Claude, Perplexity, Gemini — are becoming substantial traffic sources for any business that gets cited in their answers. Citation behavior doesn't map cleanly onto traditional SEO ranking factors. There's a new optimization discipline emerging, usually called Generative Engine Optimization (GEO).

The signals that drive AI engine citations:

Definitive opening claims. AI engines extract sentences from your content that read as authoritative answers to questions. "Customer acquisition cost in 2026 averages X" is more citable than "Many businesses are seeing rising customer acquisition costs." Open paragraphs with claims, not setup.

Structured FAQ format. AI engines preferentially cite content structured as Q&A — both explicit FAQ sections with FAQPage schema and prose-format Q&A. The first article on this site to break into ChatGPT citation in 2026 was a service-page FAQ, not a long-form blog post.

llms.txt and llms-full.txt files. A new convention (proposed by Anthropic, gaining adoption industry-wide) where you publish a plain-text index of your most important URLs with descriptions, at /llms.txt and /llms-full.txt. AI crawlers preferentially read these to find authoritative content on your domain. We ship both files on this site — easy implementation, real visibility lift.

Robots.txt allow for AI crawlers. GPTBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, Bytespider, CCBot — these are the major AI crawlers. Default robots.txt configurations often don't explicitly allow them, and some platforms (Cloudflare's default AI block) actively deny them. Explicit allow-listing is required for visibility.

Citation-friendly content patterns. Specific numbers (not "many businesses" but "37% of businesses in our 2025 cohort"). Named sources for claims. First-person experience language ("In our 2025 retainer cohort..."). Dates in references. AI engines preferentially cite content that reads as authoritative source rather than rewritten summary.

Real author entities. Same E-E-A-T signal that Google rewards, but AI engines lean even harder on it. Articles attributed to a named author with a public bio, verifiable credentials, and sameAs links to LinkedIn / X get cited at much higher rates than anonymous content.

We've watched our own articles get cited in AI answers within 7-14 days of publishing — and not just the long-form ones. Short, specific, claim-dense paragraphs from our service pages get pulled into AI answers daily. The leverage is real, the implementation is straightforward, and most agencies aren't doing any of it yet.

The pillar-cluster content architecture

Content marketing in 2026 is a topical-authority game, not a keyword-stuffing game. The structural model that wins:

Pillar pages (3,000-5,000 words) — comprehensive, authoritative coverage of a major topic. Each pillar links to 8-15 supporting articles and is linked from all of them. This article you're reading is a pillar.

Supporting articles (1,500-2,500 words) — focused on a specific sub-topic of the pillar. Each links back to the pillar and to 2-3 sibling supporting articles. Title patterns: "How to X" / "Why Y is happening in 2026" / "X vs Y: what to choose."

Glossary entries (300-600 words) — definitions of category terminology. Each glossary term links to relevant pillar pages and gets linked back when the term is used in pillars/articles. This site has 70+ glossary terms backing the pillars currently published.

Service / product pages — the conversion destinations. Link from pillars and articles into relevant service pages with intent-matched anchor text.

The compounding mechanic: as you publish supporting articles, they internally link to the pillar, which builds the pillar's authority. The pillar then ranks higher, which spills authority back to the supporting articles. By month 6 of a serious content engine, the pillar starts ranking for ambitious head terms, and the supporting articles start ranking for long-tail variants. Both compound.

What this looks like in practice (we ship this for every retainer client):

• Month 1: Pillar 1 + 3-4 supporting articles + 10-15 glossary entries published. Internal linking complete.
• Month 2: Pillar 2 + 3-4 new supporting articles. Internal linking updated.
• Month 3: Pillar 3 + 3-4 new supporting articles. First measurable ranking improvement appears.
• Months 4-6: Continue 1 pillar / 4 supporting per month cadence. Begin updating earliest articles with new data.
• Months 7-12: Compound. Update pillars quarterly. Build out clusters around any pillar that starts ranking.

By month 12, a serious content engine has shipped 12 pillars + ~48 supporting articles + ~80 glossary entries — call it 130+ pieces of interlinked content, all anchored by topical depth. The pillar + cluster strategy is how new sites compete with established domains, full stop. The reason most agencies don't do this is because it takes discipline over 12-18 months and the work isn't glamorous on a deck.

Migration without losing rankings

Replatforming a site is the highest-risk SEO operation any business ever runs. Done badly, it erases years of organic traffic in a single weekend. Done well, it lifts traffic substantially because the new stack performs better.

The non-negotiables for a safe migration:

Complete URL inventory before anything else. Crawl the existing site. Export every URL, every status code, every meta title, every canonical, every backlink target. This is the baseline.

1:1 redirect map. Every old URL gets a 301 redirect to its closest equivalent on the new site. Old /blog-category/post-name to new /insights/post-name. Old /services-old/seo to new /services/website-development. Every redirect documented before launch.

Schema and meta migration. Don't rewrite meta titles and descriptions during a replatform. Keep them identical. The replatform is the change; changing metadata simultaneously confounds your ability to diagnose problems.

Sitemap submitted on day one. New sitemap submitted to Google Search Console immediately after DNS flip. IndexNow ping for fast Bing/Yandex/Seznam discovery.

Watch Search Console daily for 30 days. Look for spikes in 404s, sudden drops in indexed pages, surge in soft-404 reports. Catch problems inside 48 hours, not when revenue collapses six weeks later.

Don't change top-performing content. The temptation during a replatform is to "improve" your highest-traffic pages. Don't. Migrate them identical. After 60 days of stable rankings on the new platform, then iterate.

We've migrated dozens of sites. The pattern of safe migrations: meticulous preparation, 1:1 redirects, conservative content changes. The pattern of failed migrations: rushed timeline, "we'll set up redirects later," and bulk content rewrites during the replatform. The blast radius of a failed migration is months of lost organic traffic. Take the time.

What a real audit looks like

When a prospect comes to us with an existing site, the audit covers:

Technical:

• Mobile + desktop LCP, INP, CLS measured via PageSpeed Insights + real Chrome UX data (CrUX)
• Render path: SSR / SSG / CSR breakdown
• Schema markup inventory and validation
• Sitemap structure and Search Console submission status
• Canonical handling and URL structure consistency
• Mobile usability flags
• Indexable pages count vs Search Console indexed count

Content:

• Topic coverage map (what's covered, what's missing, where the cluster gaps are)
• Internal linking architecture (orphan pages, under-linked pages, over-linked pages)
• Author entity strength (named authors, real bios, sameAs links)
• Content recency (last meaningful update on top-traffic pages)

Off-page:

• Backlink profile (quality, anchor diversity, toxic flag review)
• Brand mentions vs link conversions (unlinked brand mentions are recovery opportunities)
• Local citations (NAP consistency on category-specific directories for local businesses)

AI search:

• llms.txt presence and quality
• Robots.txt AI crawler allow-list
• Citation rate in ChatGPT, Perplexity, Claude (manual sampling for top 20 keywords)
• AI Overview presence on top-target keywords

The audit takes 5-8 hours and produces a 12-15 page report. We charge $2,400 for it as a standalone engagement, or include it in the first 30 days of any website + SEO retainer.

What it actually costs

The honest numbers for marketing site + SEO work in 2026:

Website build only (no ongoing SEO):

• Small marketing site (5-8 pages): $4,500-$12,000
• Mid-size marketing site (10-25 pages): $12,000-$28,000
• E-commerce on Shopify theme: $4,500-$15,000
• E-commerce custom Shopify: $18,000-$45,000
• E-commerce headless / Hydrogen: $45,000-$150,000+

SEO + content retainer (monthly):

• Small business: $1,800-$3,500/mo
• Mid-market: $3,500-$8,000/mo
• Enterprise: $10,000-$30,000/mo

Our pricing falls in the boutique-to-mid-tier range across both. We don't take standalone SEO retainers below $1,800/mo because the production capacity needed (pillar + supporting articles + glossary at the cadence above) doesn't fit smaller budgets.

The trap most businesses fall into: spending $30,000 on a custom website build and then spending $0/mo on SEO afterwards. The website on its own doesn't generate organic traffic. The website + 12 months of content engineering generates organic traffic. Budget for both or the website investment compounds at zero.

When website + SEO isn't worth doing

We tell prospects to skip a custom website build about 10% of the time. The categories where the standard playbook isn't right:

• Pre-launch businesses without product-market fit. Build a landing page on Carrd ($19/year), focus on paid for customer discovery, defer the real site until you know what you're selling.
• Hyperlocal services with stable referral pipelines. A plumber with 95% repeat-customer business doesn't need an elaborate site. Google Business Profile + a simple Squarespace landing page is sufficient.
• Businesses planning to be acquired inside 12 months. Custom website investment doesn't compound on that timeline. Keep what you have.

Likewise, we tell prospects to skip SEO retainers about 15-20% of the time:

• Businesses in categories where organic search is dominated by Wikipedia, Reddit, or YouTube. Some categories have organic search markets you genuinely can't break into without 5+ years of effort. Allocate to paid instead.
• Businesses targeting under-100-search-volume-per-month long-tail. The economics don't work; the engineering cost of SEO exceeds the traffic value.
• Businesses in restricted industries where AI engines block citations. Cannabis, kratom, firearms — these face platform-specific SEO challenges. Different playbook (covered in the restricted-vertical marketing pillar).

Picking a website / SEO agency without getting burned

If you're going to outsource this, the diligence checklist:

1. Ask for the live URL of a recent site they built. Look at it on mobile. Run it through PageSpeed Insights. If their portfolio site has an LCP above 2.5s, they can't get yours below it.

2. Ask about their schema implementation. A real SEO agency answers with specific schema types they implement and why. An agency that says "we add schema" without specifics is bluffing.

3. Ask about their AI search optimization approach. If they don't have a documented approach for GEO, llms.txt, and AI crawler allow-listing, they're stuck in 2022.

4. Ask about content production volume and structure. A serious SEO agency commits to pillar + cluster content cadence. An agency that pitches "we'll write 4 blog posts a month" without a topical-authority strategy is selling activity, not outcomes.

5. Ask about their migration approach. Real agencies have a documented checklist for replatforming. Bad agencies promise "we'll handle everything" without specifics.

6. Ask which CMS they recommend and why. A real agency has an opinion and can defend it. An agency that says "whatever you prefer" doesn't have enough conviction to do this well.

7. Ask for content samples. Read what they've written for other clients. Does it have specific numbers? First-person experience language? Or does it read like content optimized for word count?

What's coming in 2027

Three things worth watching:

AI Overview share of click-throughs is the biggest variable. In 2026 it's roughly 35-45% for informational queries; commercial queries are less affected. By 2027 we expect this to climb to 50-60% on info queries and start meaningfully impacting commercial queries. The implication: GEO will move from a nice-to-have to a primary SEO discipline.

Core Web Vitals threshold tightening. Google has indicated CWV thresholds will continue to tighten. Sites currently passing at 2.4s LCP may not pass by mid-2027. Continued performance investment matters.

Image and video search are increasingly important entry points. We're starting to see meaningful traffic from Google Lens and image search on specific verticals (home services with portfolio work, jewelry, fashion). Optimizing for image search (alt text discipline, structured image metadata, image schema) is an emerging discipline worth watching.

What we'd do if you handed us your website + SEO tomorrow

Days 1-14 — audit and prioritization.

• Full technical audit (12-15 page report)
• Topic coverage map
• Backlink profile review
• AI search citation baseline
• Prioritized 90-day roadmap

Days 15-30 — technical fixes.

• Schema implementation across all pages
• LCP optimization (typically dropping from 3s+ to under 1.5s)
• Sitemap split + Search Console resubmission
• robots.txt AI crawler allow-list
• llms.txt + llms-full.txt published

Days 31-60 — content engine launch.

• Pillar 1 published with 3-4 supporting articles
• Glossary expansion (10-15 terms)
• Author bio pages with proper E-E-A-T signals
• First measurable ranking improvement appears

Days 61-90 — compound.

• Pillar 2 published with 3-4 new supporting articles
• First quarterly content review (what's working, what's getting cut)
• Backlink outreach to category-relevant high-authority targets
• First measurable AI search citation appears

By day 90, an account on this engagement has: a measurably faster site, a complete schema implementation, the start of a real topical authority moat, and the first inbound organic traffic uplift. By month 12, the moat is meaningful — the kind of compounding traffic that doesn't disappear when ad spend pauses.

If you'd like an audit on your specific site, open the intake and we'll come back inside one business day with a detailed report covering everything above. Whether you end up working with us or not — the audit is yours either way.

For the rest of the playbook on what to do with that traffic once you have it, see the companion guides on paid advertising in 2026 and social media in 2026.