The Technical SEO Checklist for 2026

Technical SEO in 2026 spans six layers: crawl access, indexation control, performance, rendering, structured data, and — the new one — AI-crawler readiness. This is the working checklist we run on every audit, published in full. Work top to bottom; each layer assumes the one above it.

Layer 1: Crawl Access

• robots.txt permits everything you want crawled — and explicitly considers GPTBot, PerplexityBot, ClaudeBot, Google-Extended (blocking AI crawlers is opting out of AI search)
• XML sitemap: current, error-free, submitted, containing only canonical 200-status URLs
• No accidental crawl traps: infinite calendars, session parameters, faceted-filter explosions
• Server returns proper status codes; no soft-404s; redirect chains ≤1 hop
• Crawl-budget check on large sites: log-file review of where bots actually spend time

Layer 2: Indexation Control

• Canonical tags: self-referencing on originals, correct on variants, never conflicting with noindex
• Index bloat audit: parameter URLs, tag pages, thin archives out of the index
• Pagination and faceted navigation with deliberate index/noindex rules
• hreflang correct and reciprocal, if international
• One indexable version per resource: protocol, www, trailing-slash all consolidated

Layer 3: Performance (Core Web Vitals)

• LCP < 2.5s on real-user (CrUX) data, not just lab runs
• CLS < 0.1: dimensions on images/embeds, no late-injected banners
• INP < 200ms: long-task audit, third-party script diet
• Images: modern formats (WebP/AVIF), responsive sizes, lazy-loaded below the fold
• Fonts preconnected/preloaded; render-blocking CSS/JS minimized

Layer 4: Rendering

• Critical content and links present in rendered HTML — verified with URL Inspection, not assumed
• No client-side-only navigation invisible to crawlers
• Key content not gated behind interaction (tabs/accordions that never enter the DOM)
• Remember: most AI crawlers execute little or no JavaScript — if it's not in the initial HTML, LLMs may never see it

Layer 5: Structured Data

• Organization schema with alternateName and sameAs — your entity's ID card
• Page-type schema where genuinely applicable: Article, Service, Product, FAQPage, BreadcrumbList
• Person schema for authors, linked to bio pages
• Everything validated; no markup describing content that isn't visibly on the page

Layer 6: AI Readiness (the 2026 layer)

• llms.txt at root: a curated map of your key pages for machine readers
• Answer-first structure on commercial pages: 40–60 word liftable openings
• Visible "last updated" dates and a real refresh cadence on pillar content
• Entity consistency check: does asking three AI engines "what is [your brand]?" return the same, correct answer?

Every item here is checkable in an afternoon with Search Console, a crawler, and honesty. Or have us run the 200-point professional version — the technical SEO audit — which ends in fixes, not just findings. We eat our own checklist: this site's robots.txt and llms.txt are live examples.

How to Actually Run This Checklist (The Two-Afternoon Version)

Afternoon one — evidence gathering: crawl the site with Screaming Frog (the free tier covers 500 URLs), export Search Console's indexing and Core Web Vitals reports, pull a CrUX check on your top templates, and fetch robots.txt, sitemap.xml, and llms.txt. Afternoon two — triage: walk the six layers against the evidence, logging every failure with three fields: severity (does it block crawling/indexing, degrade it, or merely offend perfectionism?), effort, and owner. You'll finish with a prioritized register instead of anxiety — which is the entire difference between an audit and a worry list.

The Five Failures We Find Most Often

1. Redirect chains from historical migrations — http→https→www→trailing-slash, three hops deep, quietly taxing every crawl and every link's equity.
2. Index bloat from parameters and tags — thousands of near-duplicate URLs diluting the crawl while money pages wait days for re-indexing.
3. Client-side content invisible to non-Google crawlers — fine in Googlebot's renderer, absent for GPTBot and friends, which is why a site can rank yet never be cited.
4. Schema that contradicts itself — one Organization name in the homepage markup, another in the footer, a third on the About page. Entity ambiguity, self-inflicted.
5. The 2023 AI-crawler blocklist nobody revisited — a panic-era robots.txt still refusing GPTBot in 2026, silently opting the brand out of AI search.

Severity Triage: What Blocks vs What Degrades

Fix this week: anything preventing crawl or index of commercial pages — robots errors, noindex accidents, 5xx instability, canonical conflicts. Fix this month: Core Web Vitals failures on template level, redirect chains, schema invalidity, rendering gaps. Fix this quarter: index bloat cleanup, internal-linking improvements, llms.txt and answer-structure passes. Sequencing by severity is what separates a useful audit from the 60-page PDF that dies in a drawer — and it's why our professional audits ship as a register with owners and dates, not a document with a cover page.

Keeping It Fixed: The Regression Problem

Technical SEO decays at the speed of your deploy cycle: every release can reintroduce a noindex, break a canonical, or ship a layout shift. The countermeasures are cheap — a pre-launch checklist item in your release process, automated weekly crawls diffed against baseline, and Search Console alerts actually routed to a human. Sites that institutionalize these three stay clean for years; sites that don't re-buy the same audit every eighteen months.

Tooling: The Minimal Stack That Covers All Six Layers

You don't need enterprise spend to run this checklist properly. The working minimum: a desktop crawler (Screaming Frog's free tier for small sites; paid for anything real) covering layers 1–2 and much of 5; Google Search Console — free and non-negotiable — for indexation truth, Core Web Vitals field data, and rendering checks via URL Inspection; PageSpeed Insights / CrUX for layer 3 at template level; the schema validator for layer 5; and your server log files — the most under-used free dataset in SEO — for where crawlers actually spend time, including which AI bots visit and what they fetch. Layer 6 needs no tooling at all beyond honesty: open your robots.txt, check for an llms.txt, and paste your key page into a plain-text extractor to see what a non-rendering crawler receives. The pattern worth internalizing: every layer has a free source of truth, and paid tools mostly add scale, scheduling, and prettier exports. Teams fail this checklist not for lack of budget but for lack of cadence — which is why the last section of this post exists, and why it's the one to actually implement.

Key Takeaways