Did improving Core Web Vitals actually boost our AI citations? Here's our data from 6 months of testing

Your results align with what we’re seeing in the industry. Here’s why this makes sense:

Why AI crawlers care about performance:

1. Timeout limits - AI crawlers have strict time budgets. Slow pages get skipped.

2. Render completion - AI needs fully rendered content. Slow LCP means incomplete parsing.

3. Quality signal - Fast sites correlate with well-maintained, authoritative content.

The research backs this up:

A study of 2,138 websites found:

• CLS ≤ 0.1: 29.8% more AI mentions
• LCP ≤ 2.5s: ~50% more likely to appear in AI results
• TTFB < 200ms: 22% higher citation density
• Pages > 1MB: 18% abandoned by AI crawlers

The bottom line:

Core Web Vitals aren’t just about user experience anymore. They’re crawlability signals for AI systems.

We did a similar experiment but focused specifically on Perplexity crawling.

Perplexity-specific findings:

Perplexity crawls in real-time to answer queries. This makes performance even more critical:

• Pages loading > 3s: 60% lower citation rate
• Pages with render-blocking JS: 40% lower citation rate
• Pages with CDN: 2.1x higher citation rate vs. origin-only

The real-time factor:

Unlike Google (which pre-indexes), Perplexity fetches pages live when answering queries. Every millisecond matters because users are waiting.

What we prioritized:

1. Edge CDN (Cloudflare)
2. Minimal JS dependencies
3. SSR/Static generation over CSR
4. Pre-connect to common resources

Result: Perplexity citation rate increased 3.2x over 3 months.

We’ve been tracking this across 40+ client sites. Some patterns:

Performance tiers and AI citations:

The biggest single improvement:

LCP optimization has the highest ROI. Moving from 4s to 2s LCP typically doubles AI citations without other changes.

Quick wins we recommend:

1. Image optimization (biggest impact, easiest fix)
2. Remove unused CSS/JS
3. Use CDN for static assets
4. Preload critical resources

Start with LCP. It has the highest correlation with AI citation improvement.

I’ve been analyzing the relationship statistically. Some findings:

Correlation coefficients:

Interpretation:

LCP has stronger correlation with AI citations than backlinks or domain age. This is notable because we traditionally consider those authority signals more important.

Caveat:

Correlation isn’t causation. But it suggests performance optimization deserves more attention than many AI visibility strategies give it.

WordPress-specific perspective here. Most WordPress sites have terrible Core Web Vitals by default.

Common WordPress performance killers:

1. Unoptimized themes (bloated CSS/JS)
2. Too many plugins
3. No caching
4. Images uploaded at full resolution
5. Shared hosting

What we do for clients:

1. Switch to lightweight theme (GeneratePress, Kadence)
2. Audit and reduce plugins
3. Add caching (WP Rocket, LiteSpeed)
4. Implement image optimization plugin
5. Move to quality hosting

Typical results:

• LCP: 6s -> 2s
• CLS: 0.25 -> 0.05
• AI citations: 2-3x improvement

WordPress sites are often the worst performers and have the most room for improvement.

Enterprise perspective: performance optimization for AI is different from traditional performance optimization.

Traditional optimization:

• Focus on user-facing pages
• Optimize for first visit
• Consider returning visitors

AI crawler optimization:

• ALL content pages matter (not just top traffic)
• Crawlers don’t cache like browsers
• Bots may hit many pages rapidly

Enterprise-specific recommendations:

1. Server capacity for burst traffic - AI crawls can be aggressive
2. Consistent performance across all pages - Not just homepage
3. API response times - If content loads via API, optimize that too
4. Monitor crawler behavior - Log analysis shows AI crawl patterns

We found our documentation pages (historically deprioritized for performance) had the highest AI citation potential once optimized.

I build SEO tools and have been analyzing AI crawler behavior.

Interesting patterns:

1. GPTBot is less patient than Googlebot - Times out faster
2. PerplexityBot focuses on text - Skips heavy interactive elements
3. ClaudeBot respects robots.txt strictly - Very polite crawler

What this means practically:

• Ensure critical content renders quickly
• Don’t rely on client-side rendering for important text
• Heavy interactive features can be lazy loaded
• Alt text and structured data matter (text-based understanding)

Testing tip:

Use Puppeteer or Playwright with aggressive timeouts (5s) to simulate AI crawler behavior. If your content doesn’t appear in that window, AI crawlers might miss it.

This thread has validated our findings and added useful context.

Key takeaways:

1. Core Web Vitals DO impact AI citations significantly
2. LCP is the highest-priority metric (50% more AI inclusions)
3. AI crawlers have stricter timeouts than traditional crawlers
4. Heavy pages (>1MB) may be abandoned by AI crawlers
5. WordPress sites have the most room for improvement

Our optimization checklist:

1. LCP under 2.5s (ideally under 2s)
2. CLS under 0.1
3. INP under 200ms
4. Page weight under 1MB
5. CDN for all static assets
6. Preload critical resources

Monitoring:

• Track Core Web Vitals in Search Console
• Monitor AI citations with Am I Cited
• Correlate improvements to measure impact

The ROI on performance optimization just increased. It’s not just about user experience anymore - it’s about AI visibility.

Thanks everyone for the data and insights.