The 90+ checks we run

Without a viewport tag, mobile browsers render the page at desktop width and scale it down — unreadable, and penalised under mobile-first indexing.

A noindex directive removes this page from search results entirely. Usually intentional on staging — frequently shipped to production by accident.

The lang attribute tells search engines which regional results to serve, and screen readers which pronunciation to use.

Google shows favicons beside organic results on mobile. Without one you get a generic globe and lose brand recognition in the SERP.

A malformed canonical URL can make Google discard the canonical hint entirely, leaving duplicate resolution to guesswork.

A cross-domain canonical tells Google to rank a different site instead of this page. Occasionally deliberate — usually a syndication mistake.

Non-UTF-8 encodings corrupt user-submitted text, break copy-paste, and mangle emoji and non-Latin scripts.

The HTML spec requires the charset within the first 1024 bytes. Declared later, browsers may ignore it and guess instead.

A <title> under ~50 characters wastes SERP real estate and rarely captures the keywords people actually search.

Google truncates titles around 60 characters on mobile. Anything past the cutoff is invisible to searchers.

A description under ~140 characters leaves SERP space empty, and Google may pad it with random page text.

Google truncates descriptions after ~160 characters — including any call-to-action you put at the end.

Partial Open Graph data produces inconsistent link previews: some platforms scrape it, others fall back to the title tag.

Without an explicit charset the browser guesses the encoding, which garbles non-ASCII text and emoji.

X falls back to Open Graph, but declaring twitter:card explicitly gives you control over the preview layout.

Every page needs a unique, descriptive <title>. It is the strongest on-page ranking signal and the headline shown in search results.

A 140–160 character meta description summarises the page for search engines and click-through.

The canonical tag tells Google which URL to consolidate ranking signals to when duplicates exist.

OG tags control how your page appears when shared on Slack, LinkedIn, Facebook, WhatsApp, X, etc.

Without @type there is nothing to interpret — the structured data block is silently discarded.

QAPage markup is only valid for forum-style pages with a single question that users can submit answers to. Using it on a self-authored FAQ, a blog post, or any page with multiple questions is a structured-data spam violation.

JSON-LD is how you qualify for SERP rich results (stars, FAQs, prices, breadcrumbs) and AI search citations.

78% of structured-data errors are syntax (missing comma, smart quotes, unclosed bracket). A single typo invalidates the entire block.

Multiple H1s split the topical signal and leave crawlers guessing which one describes the page.

Alt text like "image1.jpg" or "photo" describes nothing — it fails both screen-reader users and image search.

Anchors like "click here" and "learn more" pass no topical signal and are useless to screen-reader users navigating by link.

When the H1 and <title> share few words, Google may rewrite your SERP title — often worse than your own.

Jumping from H2 straight to H4 breaks the document outline for screen readers and muddies content hierarchy for crawlers.

Internal links distribute ranking signals and are how crawlers discover related pages. Orphan pages consistently underperform.

Pages with very little body text rarely rank, and Google's helpful-content guidance treats them as low value.

Every page should have exactly one H1 that names its primary topic.

Alt text is required for accessibility (WCAG) and image SEO. Use alt="" for purely decorative images.

CSP restricts which scripts may run. Without it, any reflected or stored XSS executes unmitigated.

Without it your pages can be embedded in an invisible iframe for clickjacking attacks.

Referrer-Policy controls how much of your URL leaks to other sites in the Referer header.

nosniff stops browsers MIME-sniffing a response into an executable type.

HTTPS is mandatory in 2026 — Chrome marks HTTP pages "Not Secure" and Google uses HTTPS as a ranking signal.

HSTS forces browsers to upgrade HTTP requests to HTTPS automatically, preventing downgrade attacks.

A Lighthouse performance score under 50 means severe Core Web Vitals problems — most mobile users leave before the page renders.

Uncompressed HTML can be 4–10× larger over the wire, hurting Largest Contentful Paint and mobile data costs.

When several pages share a title, Google picks one to surface for a query — and you do not get to choose which.

Identical descriptions waste the chance to differentiate each page in the SERP, and Google often rewrites them anyway.

A blanket Disallow: / removes the entire site from search. Correct on staging, catastrophic on production.

Malformed directives are handled inconsistently between crawlers — some ignore the line, others misread the whole file.

Old lastmod values tell Google there is nothing new to fetch. Fresh content behind a stale date gets skipped.

x-default tells Google what to serve when no language matches. Without it, unmatched regions land on the wrong locale.

Only relevant for multi-language or multi-region sites. Single-locale sites can ignore this safely.

The Sitemap directive is how every search engine — not just Google — discovers your URL inventory automatically.

robots.txt is the standard way to tell crawlers which paths to skip and where to find sitemaps.

Sitemaps give search engines a definitive URL list with lastmod hints.

The page returns different content to GPTBot or PerplexityBot than to a normal browser. Usually a CDN or firewall rule, not a deliberate choice.

Only a fraction of this page exists in the raw HTML. AI crawlers do not execute JavaScript, so they see very little of it.

The spec expects an H1 site name and a blockquote summary. Note that Google states llms.txt neither helps nor harms visibility — treat this as hygiene, not a ranking lever.

AI engines favour content that cites primary sources. Linking to .gov, .edu and peer-reviewed pages signals research-backed writing.

When content is buried under framework markup, AI engines and crawlers struggle to extract the actual text.

LLMs lift tables directly into answers because they are structured and easy to extract. Comparison content without a table gets cited less.

Your robots.txt disallows GPTBot (OpenAI) across the whole site, so it cannot read or cite any of your pages.

Your robots.txt disallows ClaudeBot (Anthropic) across the whole site, so it cannot read or cite any of your pages.

Your robots.txt disallows PerplexityBot across the whole site, so it cannot read or cite any of your pages.

Your robots.txt disallows OAI-SearchBot across the whole site, so it cannot read or cite any of your pages.

Your robots.txt disallows Google-Extended across the whole site, so it cannot read or cite any of your pages.

The raw HTML contains almost no text — the content only exists after the JavaScript bundle runs. AI crawlers do not run JavaScript, so they receive an empty page.

llms.txt is the emerging convention (2024-26) for telling AI crawlers what your site is about and what pages matter.

GPTBot, ClaudeBot, PerplexityBot, etc. need access to your content to cite you in AI search answers.

AI engines cite pages that state the answer plainly in the first paragraph. Pages that "set up" the topic before answering get skipped.

AI engines lift individual paragraphs as citations. Paragraphs relying on "this", "it", "as mentioned" are less likely to be extracted.

Google's E-E-A-T framework and AI engines both prefer cited content with clear authorship.