NotionCue
AI Visibility Platform
All systems live
Dashboard →
AEO GuideCluster Mapllms.txt GeneratorRobots.txt GeneratorBLUF TemplatesBlogChangelogAboutPrivacyTermsContact
Free AEO Tool

Page Speed
AEO Impact Checker

Enter any URL to get real Core Web Vitals data from Google PageSpeed Insights — then see exactly how your load performance affects citation rates across ChatGPT, Perplexity, Gemini, and Claude.

Free — no account needed
Real PageSpeed Insights data
AEO impact per metric
How it works

Three steps to your
speed audit.

This tool pulls live data from Google's PageSpeed Insights API — the same data source Google uses internally — then maps each metric against known AEO thresholds to show you which speed issues are actively costing you AI citations.

01
Enter your URL
Paste any public URL — homepage, blog post, landing page. The tool runs a live PageSpeed Insights test against the real URL, so what you get back reflects your actual current performance, not a cached snapshot.
02
We pull your Core Web Vitals
The tool fetches LCP, TTFB, CLS, FID, TBT, and your overall Performance score for both mobile and desktop. Each metric is compared against the thresholds that matter for AI crawler compatibility, not just standard Google search benchmarks.
03
Get your AEO impact report
Each metric gets a plain-English AEO impact rating — how likely it is to cause retrieval timeouts, extraction errors, or deprioritisation by specific AI engines. Fixes are ranked by impact so you know what to tackle first.
Why it matters

Speed is a citation signal,
not just a UX metric.

Most SEO professionals know page speed affects Google rankings. Fewer realise it directly affects whether AI engines can cite your content at all. The mechanism is different from traditional search — and the threshold is less forgiving.

Retrieval-based engines like Perplexity and ChatGPT browse mode don't work from a pre-built index the way Google does. They fetch and parse your page in real time, on demand, as part of answering each query. If your page takes 6 seconds to load, there's a real chance the crawler gives up before it finishes reading. Your content simply doesn't make it into the answer — not because it wasn't relevant, but because the server was too slow.

Google's Gemini is less sensitive to real-time fetch speed because it draws primarily on Google's existing crawl index. But Gemini uses mobile-first indexing, which means pages with poor mobile performance — even if desktop is fine — get lower quality signals in the index Gemini draws from. A page in the "Poor" Core Web Vitals band consistently gets fewer citations from Gemini than an equivalent page in the "Good" band, independent of content quality.

Claude and other LLMs that use web retrieval tools follow similar patterns. The common thread across all of them: a page that loads fast is a page that gets read. A page that doesn't get read doesn't get cited.

AEO thresholds by metric
LCP
< 2.5s Good2.5–4s Warn> 4s Poor
Perplexity crawlers time out around 4s. LCP under 2.5s leaves a safe margin. Above 4s, real-time citation is unreliable.
TTFB
< 600ms Good600ms–1.8s Warn> 1.8s Poor
Server response time is the first bottleneck any crawler hits. A high TTFB delays every subsequent metric and compounds retrieval risk.
CLS
< 0.1 Good0.1–0.25 Warn> 0.25 Poor
High layout shift doesn't cause timeouts, but it can fragment text blocks during extraction — the model receives garbled or incomplete content.
TBT
< 200ms Good200–600ms Warn> 600ms Poor
Heavy JavaScript blocking the main thread prevents headless crawlers from reading dynamically rendered content entirely.
FCP
< 1.8s Good1.8–3s Warn> 3s Poor
First Contentful Paint affects how quickly text content is visible to crawlers that don't wait for full page load.
Engine breakdown

How each AI engine
handles slow pages.

Not all AI engines are equally sensitive to page speed. The impact depends on whether the engine uses real-time retrieval, a crawl index, or a hybrid of both. Here's what actually happens when your page is slow, for each engine.

Perplexity
Very High sensitivity
Real-time retrieval on every query
Perplexity fetches URLs live as part of every answer. If your page doesn't load within its timeout window, it won't be cited regardless of relevance. TTFB and LCP are the critical metrics here. A server that responds in under 600ms and renders primary content under 2.5s will be reliably retrieved. Above those thresholds, citation becomes unreliable.
ChatGPT Browse
High sensitivity
Real-time browser-based retrieval
ChatGPT's browse mode uses a headless browser to fetch pages. Heavy JavaScript that blocks the main thread (high TBT) is a significant problem here — the browser waits for JS to resolve before reading content, and pages with TBT above 600ms frequently fail to extract cleanly. Render-blocking resources and deferred scripts are the main culprits.
Gemini
Medium sensitivity
Google crawl index + real-time signals
Gemini draws on Google's existing search index rather than fetching pages live per query. This makes it less sensitive to real-time load speed and more sensitive to Core Web Vitals as measured over time in the Chrome User Experience Report (CrUX). Mobile performance matters more here than for other engines — Gemini uses mobile-first indexing.
Claude (web retrieval)
Medium–High sensitivity
Tool-based retrieval on demand
When Claude uses web retrieval tools, it fetches pages through an API call that behaves similarly to a headless browser. Pages with server-side rendering and fast TTFB tend to extract cleanly. Client-side rendered apps that require JavaScript execution to display content are the highest-risk format — Claude may receive an empty shell rather than the actual content.
Who benefits

Not just for developers.
Any team losing AI citations to slow pages.

Page speed is often treated as a developer problem. In practice, the teams that lose the most AI citations to slow pages are content and SEO teams who don't have visibility into performance data — until now.

📈
SEO Professionals
Identify speed issues before they tank AEO performance
Benchmark client sites against competitors on Core Web Vitals
Build the performance case for dev team prioritisation
Track metric improvements after CDN or hosting changes
✍️
Content Marketers
Check whether heavy images are costing your posts citations
Identify pages that rank well in search but get skipped by AI
Prioritise which content pages to optimise first
Understand the speed gap between your content and competitors'
⚙️
Developers & Agencies
Translate Core Web Vitals into client-facing AEO impact
Identify render-blocking JS causing crawler read failures
Validate speed improvements with before/after AEO scoring
Present performance work in terms of citation rate, not just ms
Practical fixes

The fastest ways to improve
your AEO speed score.

Most speed problems that affect AI citation rates fall into four categories: server response time, image optimisation, JavaScript blocking, and render strategy. Each fix below maps to a specific metric and an AEO impact level.

TTFBLow effortVery high impactMove to a CDN with edge caching enabled. Cloudflare's free tier alone cuts TTFB by 40–70% for most sites by serving cached responses from a node geographically close to the requester. This is the single highest-leverage speed improvement available for the least effort.
TTFBMedium effortHigh impactEnable full-page caching for static or semi-static pages. WordPress sites using WP Rocket or Cloudflare Cache Rules, Next.js sites using ISR — any approach that eliminates server-side processing on repeat requests cuts TTFB dramatically.
LCPLow effortHigh impactPreload your LCP element. Add <link rel="preload"> for the hero image or above-the-fold text block. This is a one-line HTML change that consistently moves LCP by 0.5–1.5 seconds on pages where the LCP element is an image.
LCPMedium effortHigh impactConvert images to WebP or AVIF and serve them with responsive srcset. A 1MB hero JPEG is the single most common cause of LCP above 4s on mobile. Converting to WebP typically reduces file size by 25–35% with no visible quality loss.
TBTMedium effortHigh impactAudit and defer third-party scripts. Analytics, chat widgets, ad scripts, and heatmap tools loaded in the <head> are the most common source of TBT above 600ms. Move them to load after the main content using async or defer attributes, or load them on user interaction.
TBTHigh effortVery high impactSwitch client-side rendered (CSR) pages to server-side rendering (SSR) or static generation. CSR pages — built with vanilla React or Vue without SSR — serve an empty HTML shell until JavaScript runs. AI crawlers that don't execute JS see a blank page. Next.js, Nuxt, or Astro with SSR fix this at the framework level.
CLSLow effortMedium impactSet explicit width and height attributes on all images and iframes. The browser can then reserve the correct space before the asset loads, eliminating the layout shift that happens when images pop in and push content down. This is a two-attribute HTML fix per image element.
FAQ

Common questions.

Why does page speed affect AI citations?
Which Core Web Vital matters most for AEO?
What PageSpeed score should I aim for?
Is this tool free?
Does desktop speed or mobile speed matter more for AEO?
Will improving PageSpeed directly increase my AI citation rate?
What causes a high TTFB and how do I fix it?