What is AI Fix

What AI Fix Does

When VektorAI audits a page, it identifies issues across SEO, AEO, GEO, and technical categories. AI Fix takes those findings a step further: it generates actual code you can paste directly into your page source to resolve each issue.

Unlike generic fix templates that leave you with placeholder text, AI Fix reads your page's real content — the title, headings, body text, existing structured data — and writes fixes that reference your actual data. The output is ready to use immediately with no manual editing required.

AI Fix is available on saved audits for users on the Pro plan. For instance, if your page title is 'Best Coffee Shops in NYC', AI Fix generates a meta description referencing your actual content rather than a generic placeholder.

"On-device AI inference represents a paradigm shift in web privacy — user data never leaves the browser."

— W3C WebGPU Working Group

How It Works

AI Fix uses WebLLM to run AI inference directly in your browser via WebGPU. This means the language model runs entirely on your device — no page data is ever sent to an external server. According to the W3C WebGPU specification, browser-based GPU acceleration enables real-time AI inference without server round-trips.

During the audit, VektorAI extracts key context from your page and stores it alongside the audit results. When you click AI Fix on a specific issue, that context is passed to the local model, which generates a tailored fix based on your page's actual content.

Page Context Extracted

During the audit, VektorAI extracts the following signals from your page. This context is stored in the audit results and used by the AI model to generate relevant fixes.

• +Title tag
• +Meta description
• +H1 heading
• +H2 headings (all)
• +First paragraph of body content
• +Existing JSON-LD structured data
• +Open Graph tags (og:title, og:description, og:type)
• +Language attribute (html lang)
• +Author name (meta author)
• +Published date (article:published_time, meta date, or time element)
• +Site name (og:site_name)

Three Fix Tiers

Not every fix needs AI. VektorAI uses a tiered system to determine the best approach for each issue.

<!-- Example: Tier 1 deterministic fix -->
<meta name="viewport" content="width=device-width, initial-scale=1" />

<!-- Example: Tier 2 content-aware meta description -->
<meta name="description"
  content="Learn how to optimize your Next.js app for AI search
  engines with structured data and GEO best practices." />

<!-- Example: Tier 3 JSON-LD generation -->
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "What is AEO?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Answer Engine Optimization is the practice
        of optimizing content for AI-powered search..."
    }
  }]
}
</script>

Categories That Get AI Fixes

AI Fix covers the following audit categories. Each category uses different page context signals to generate the most relevant fix.

• +Open Graph — uses title, H1, meta description, first paragraph, og:type, site name, and language
• +Twitter Card — uses title, H1, meta description, and first paragraph
• +Author & Entity — uses title, H1, author name, published date, and site name
• +Technical — uses title, H1, and first paragraph to infer language and context
• +Meta — uses the page title for tag generation
• +Headings — uses the page title and existing heading content
• +Images — uses surrounding content to generate descriptive alt text
• +Structured Data — uses all available page context for JSON-LD generation

Performance Issues

How to Use AI Fix

Browser Requirements

AI Fix requires a browser that supports WebGPU, the modern GPU acceleration API used for local AI inference.

• +Google Chrome 113 or later
• +Microsoft Edge 113 or later
• +Other Chromium-based browsers with WebGPU enabled

Privacy

How does AI Fix protect my data?

AI Fix runs entirely in your browser using WebLLM and WebGPU. No page content, audit data, or generated fixes are transmitted to any server for AI processing. The model weights are downloaded once and cached locally. Your data stays on your device at all times during the fix generation process.