Free website speed test, with an AI-readiness check
How fast is your site? Can an AI read it?
One check, two views. We load your page the way a mid-range phone does, then fetch it with no browser and no JavaScript, the way Google's crawler and ChatGPT do. Then we tell you, in plain English, what to fix first.
Free, no signup, about a minute. Ask your assistant: check_site_speed via our MCP server
What you get back
An example report.
Illustrative values, example-clinic.com
Lab score, simulated phone. Real-visitor numbers below.
48 /100 Poor
In plain English
Real visitors wait about 3.9 seconds for your main content. The single biggest cause: resize and compress the hero photo, worth up to 1.4 s. Separately, an AI assistant asking about you can read only 13 words of this page.
What your visitors experience, real Chrome users, last 28 days
What do these mean?Main content shows
Needs work3.9 s
Good is under 2.5 s. Google calls this LCP.
Responds to a tap
Needs work240 ms
Good is under 200 ms. Google calls this INP.
Stays still
Good0.04
Good is under 0.1. Google calls this CLS.
You pass when 75% of real visits are good on all three. This page fails on two.
What an AI assistant sees, fetched with no JavaScript
- Words readable in the raw HTML 13 words
- AI crawlers allowed by robots.txt All blocked
- llms.txt Not found
- Business facts as structured data Name and address only
Raw HTML, as an agent receives it
<div id="root"></div> <noscript>You need to enable JavaScript to run this app.</noscript> <script src="/bundle.4f2a.js"></script>
13 words. Your text is not in the page.
Fix first, ordered by time saved
- 01 −1.4 s
Resize and compress the hero photo
3.1 MB straight from the camera. It should be about 120 KB and will look identical.
- 02 −0.8 s
Load the chat widget after the page appears
It downloads 420 KB of script before your content can paint.
- 03 −0.6 s
Send service pages as HTML from the server
Fixes the agent view entirely and takes work off slow phones.
- 04 −0.3 s
Serve fonts from your own domain
Text waits for a font from another server before it shows.
Savings overlap, so don't add them up. 5 more issues are in the full report.
What up to 3.1 s might be worth, per year
At 2% conversion and $50 per order
| Monthly visitors | Estimated extra revenue |
|---|---|
| 5,000 | $2k to $6k |
| 20,000 | $7k to $22k |
| 100,000 | $37k to $112k |
Order-of-magnitude only. Widely cited industry studies: Walmart reported roughly 2% more conversions per second saved, Amazon about 1% per 100ms. The effect isn't linear and varies a lot by sector, so treat this as a range, not a forecast.
Get the full list by email
Every issue, how much time each saves, and the agent-view details.
Found something worth fixing? The audit turns this into a ranked plan your developer can quote against, for $249, and the fee comes off any fix.
See what a fix costsWhy is my website slow?
Almost always one of four things.
In rough order of how often they turn out to be the culprit. Testing finds out which one is costing you most, because fixing the wrong one is a lot of work for nothing.
01, the most common
Images that are too big
A camera photo can be several megabytes. Resized properly it's often 95% smaller and looks the same.
02
Third-party scripts
Chat, analytics, ad tags, A/B testing. Each one is a request to someone else's server, and you inherit their bad day.
03
A slow server
If the first byte takes a second, everything after it is late, however well the page is built.
04
Blocking CSS and fonts
The browser has your content but won't paint it until a stylesheet and a font have downloaded.
Doing it yourself
How to test your website speed
-
Step 1
Test the page that matters
Not just your homepage. Test the page people actually land on, and the one where they buy or sign up.
-
Step 2
Test on mobile
Most visitors are on a phone, on a worse connection than yours. A page that feels instant on your laptop can take eight seconds on a mid-range Android.
-
Step 3
Test more than once
A single run can be thrown off by a slow moment on a network. Two or three runs tell you what is normal.
-
Step 4
Write the number down
Today's score is only useful compared to last month's. Without a baseline you can't tell whether a change helped.
A test or a watch?
Speed testing vs website speed monitoring
Sites rarely get slow all at once. They get slow the way a room gets untidy: one plugin, one tracking script, one uncompressed image at a time. By the time it's obvious, you've lost months of visitors and you have no idea which change caused it.
That is the difference between a test and website speed monitoring. A test is a photograph of one moment. Ongoing website performance monitoring tells you when something regressed and what changed that day, which is usually the more useful thing to know.
See Speed CareCommon questions
The short answers.
A test is a photograph of one moment. Sites get slow one plugin at a time, so the more useful thing is knowing when it changed. That's what Speed Care is for.
Why is my website moving slowly?
Usually one of four things: oversized images, too many third-party scripts (chat widgets, analytics, ad tags), a slow server response, or fonts and CSS that block the page from rendering. The check above tells you which one is costing you the most time, so you fix the expensive thing first instead of guessing.
What is a common cause of slow website loading speed?
Images, more often than anything else. A single uncompressed hero image can weigh more than the rest of the page combined. Third-party scripts are the usual runner-up: each one is a request to someone else's server, and you inherit their bad day.
What is a reasonable page load time?
Your main content should appear within 2.5 seconds on a normal mobile connection. That's Google's “good” threshold. Past four seconds, people start leaving before they've seen anything.
Why can a site score 19 and still load fine?
The score is a lab test on a simulated slow phone, and it punishes everything a page ships. What Google ranks on is how fast the page was for real visitors over 28 days. A site on fast hosting with loyal visitors can feel quick while scoring badly. The long version is here.
How to pass Core Web Vitals assessment?
You pass when 75% of real visits are good on all three: main content within 2.5 s, a response to taps within 200 ms, and layout shift under 0.1. It's measured on your actual visitors, so a page that's fine on your laptop can still fail on their phones.
Why check what an AI assistant sees?
Because more people now ask ChatGPT, Claude or Gemini about a business before visiting it. Those assistants fetch your page without running JavaScript. If your words are only built in the browser, they find almost nothing, and can't describe or quote you.
How do I make my website run faster?
In the order that usually pays off: compress and correctly size your images, remove third-party scripts you no longer use, load the rest after the page renders, and make sure your host responds quickly. The first two are often enough to move a site from red to green.
How to monitor website performance?
Test the pages that matter, not just the homepage, on a schedule, and record the numbers each time so you can see the trend. A one-off test tells you where you are today; monitoring tells you when something broke, which is usually the more useful thing to know.
What are the top 3 website performance metrics to monitor?
Largest Contentful Paint (how long until the main thing appears), Interaction to Next Paint (how quickly the page responds to a tap or click) and Cumulative Layout Shift (how much the page jumps while loading). Together they cover: did it show up, does it respond, does it stay still.
Which tool is commonly used to track website performance?
Google's PageSpeed Insights and Lighthouse are the usual starting points, and both are free. They are built for developers, though, which is why the report above is written in plain language instead.
Does this test store my site?
No. The result is kept in memory for ten minutes, so emailing yourself the report doesn't run the check again, and then it's gone. If you ask for the full report by email, we keep your address and the URL you checked so we can reply, and nothing else.