Every problem VibeCheck catches — and how to fix it

Runtime problems that make an AI-built page slow, janky, or heavy — DOM bloat, memory leaks, layout shift, long tasks, oversized images and bundles. Each one is caught live by a VibeCheck detector as it happens in the browser.

• Excessive DOM sizeerrorA huge DOM slows style, layout and memory on every frame. Virtualize long lists and flatten wrappers to get node count back under control.
• JavaScript memory leakerrorA leaking heap grows until the tab janks or crashes. Find the listeners, timers and subscriptions that are never cleaned up and release them.
• Cumulative layout shift (CLS)warningContent that jumps as the page loads fails Core Web Vitals and mis-taps users. Reserve space for images, ads and late content to stop the shift.
• Long tasks blocking the main threadwarningLong tasks freeze the page and wreck INP. Break up heavy work, defer it, or move it to a worker so the main thread stays responsive to input.
• Console.log spamwarningDebug logs left in production leak internals, slow hot paths, and bury real errors. Strip them at build time and add a no-console lint rule.
• Duplicate network requestswarningThe same endpoint fetched several times wastes bandwidth and risks race conditions. Deduplicate with a query cache or a shared in-flight request.
• Unoptimized imageserrorImages without dimensions shift layout; oversized ones waste bandwidth. Add width/height, lazy-load below the fold, and serve the right size.
• Large image fileserrorMulti-hundred-KB images blow your LCP and mobile data budget. Compress to WebP/AVIF, resize to display size, and serve from an image CDN.
• Large JavaScript bundleswarningOversized JS bundles delay interactivity and waste mobile data. Code-split by route, tree-shake imports, and lazy-load heavy components.
• Heavy dependencieswarningHeavy libraries add weight and known pitfalls. Lazy-load them, import a lighter subset, or swap for a native API to cut bundle size and jank.

Discoverability problems that keep search engines from ranking — or even reading — your pages: missing titles and descriptions, absent Open Graph and canonical tags, heading and alt-text gaps, and missing sitemap/robots files.

• Missing page titlewarningWith no <title>, search results and browser tabs fall back to your URL. Add one unique, descriptive title under 60 characters.
• Page title is too longwarningGoogle truncates titles past ~60 characters, hiding the end in search results. Trim your <title> so the whole headline shows.
• Page title is a framework defaulterrorShipping a default title like “React App” or “Vite App” tells Google your page is an unfinished template. Replace it with a real title.
• Page title is very shortwarningA one- or two-word <title> wastes your strongest ranking signal. Write a descriptive 30–60 character title with real keywords.
• Missing meta descriptionwarningNo meta description lets Google write its own snippet from page text — usually a worse pitch. Add a 150–160 character summary.
• Meta description is too longwarningSearch engines cut descriptions off past ~160 characters. Trim yours so the call to action is not lost in the ellipsis.
• Missing social preview image (og:image)warningWith no og:image, links to your page show a blank box on Slack, X, and iMessage. Add a 1200×630 preview image so shares look real.
• Missing og:titleinfoWithout og:title, shared links fall back to your raw <title>, which may not be share-optimised. Set an explicit og:title.
• Missing og:descriptioninfoWith no og:description, shared links show no supporting text under the title. Add one to make link previews compelling.
• Missing og:urlinfoog:url tells platforms the canonical address to credit shares to, even with tracking params. Set it so shares consolidate.
• Missing Twitter/X cardinfoWithout twitter:card, links on X show a plain URL instead of a rich preview. Add summary_large_image for a full-width card.
• Missing canonical linkinfoDuplicate URLs (trailing slash, query params) split your ranking. Add <link rel="canonical"> to point search engines at one address.
• No H1 headingwarningThe <h1> is your page’s strongest on-page topic signal. If it is missing, add exactly one that names what the page is about.
• Multiple H1 headingsinfoSeveral <h1>s scatter your page’s topic signal. Keep one <h1> and demote the rest to <h2> so the outline is clear.
• Images missing alt textwarningImages with no alt attribute are invisible to screen readers and search engines. Add descriptive alt text (or alt="" if decorative).
• Vague link textinfoLinks that say “click here” or “read more” tell search engines and screen readers nothing. Use text that names the destination.
• Unfriendly URL sluginfoURLs with IDs, underscores, or capital letters are hard to read, share, and rank. Use clean, lowercase, hyphenated keyword slugs.
• Page is set to “noindex”errorA noindex tag tells search engines to drop the page from results entirely. If it is not deliberate — often a staging leftover — remove it.
• Missing or invalid sitemap.xmlwarningWithout a sitemap, search engines discover pages the slow way. Generate a /sitemap.xml listing your URLs and reference it from robots.txt.
• Missing robots.txtinfoA robots.txt tells crawlers what to index and where your sitemap is. Add one so crawling is intentional, not guessed.

Answer-engine problems that stop AI assistants (ChatGPT, Perplexity, Claude, Google AI Overviews) from reading, trusting, and citing your content — missing structured data, client-only rendering, blocked AI crawlers, and absent agent-discovery signals.

• Missing structured data (JSON-LD)warningAnswer engines extract facts from schema.org JSON-LD. Without it they guess from prose. Add JSON-LD so AI and Google read your page correctly.
• Invalid structured data (JSON-LD)warningMalformed JSON-LD is silently skipped by search and answer engines. Fix the syntax and schema.org @context so your structured data actually counts.
• Missing <main> landmarkinfoWithout a <main> landmark, assistants and screen readers can’t find your primary content. Wrap the main content in a single semantic <main>.
• Missing author and date signalsinfoAnswer engines weigh authorship and freshness when choosing sources. Add author and published-date signals so your content is trusted and cited.
• Content only renders with JavaScriptwarningCrawlers and AI agents that don’t run JS see an empty page. Server-render or prerender your content so it’s in the HTML, not built client-side.
• Missing llms.txtinfollms.txt hands AI assistants a clean markdown summary of your site so they read it accurately. Add one at /llms.txt to guide LLMs to your best pages.
• No markdown content negotiationinfoServing a markdown version to agents that ask for it lets them read your content without parsing the DOM. Add Accept: text/markdown negotiation.
• robots.txt blocks AI crawlerswarningIf robots.txt disallows GPTBot, ClaudeBot or PerplexityBot, assistants can’t read or cite you. Allow the AI crawlers you want to appear in answers.
• No agent interface (MCP) advertisedinfoSites that want agents to take actions can advertise an MCP server so assistants discover their tools. Add a /.well-known/mcp.json for agent discovery.

The document-head fundamentals every page needs — viewport, charset, language, favicon. AI scaffolds routinely skip them because the generated component owns the body and nobody edits the base HTML document.

• Missing viewport meta tagerrorWithout a viewport meta tag, mobile browsers render your page at desktop width. Add one line to make the layout responsive again.
• Missing charset declarationwarningA missing <meta charset> lets browsers guess your encoding, mangling emoji and accented text. Declare UTF-8 as the first tag in <head>.
• Missing lang attribute on <html>warningA missing lang attribute stops screen readers picking the right voice and search engines detecting language. Add lang to your <html> element.
• Missing faviconwarningNo favicon means a blank browser tab and a 404 on every load. Add a <link rel="icon"> so your site is recognisable in tabs and bookmarks.