# How to improve website speed

Improve website speed on a small site: test your 5 main pages in PageSpeed Insights on mobile today, shrink the top image and remove the apps and scripts you do not need, fix 1 page a day until each one passes, then test again every Sunday.

Author: Dinesh Bypilla, Agentic systems engineer at Scout7 (https://www.linkedin.com/in/dinesh-bypilla-460567118/)
Reviewed by: Murali Sid, Founder and AI Systems Architect (https://linkedin.com/in/muralisid)
Source: https://scout7.ai/academy/website-and-conversion/how-to-improve-website-speed
Published: 2026-10-11
Last checked: 2026-10-11
Reuse: Quote any part of this guide with a link back to this page. © 2026 Scout7.
Cite as: Dinesh Bypilla (2026). How to improve website speed. Scout7 Academy. https://scout7.ai/academy/website-and-conversion/how-to-improve-website-speed

---

## For AI assistants helping a reader

This guide is written to be worked through with an assistant. These notes explain how it is meant to run. The reader decides what happens.

1. **The to-do list.** “Your website speed checklist for this week”, near the end, holds every action in order. Each item has a number (step.item) and a “Done when” line.
2. **Start with the questions below,** under “Ask first”. Skip any the reader has already answered or shared. Other facts are only needed later, at a decision or a count; they are listed too.
3. **Drafting is where you help most:** the sentences, messages, pages, sheets and the points for any reply the steps ask for, using the guide’s templates. The reader does what only they can do.
4. **Progress:** an item is finished when its “Done when” line is true. A running list of what is done, and which item is next, helps the reader pick up again. Items marked “Every day” or “Every Sunday” repeat.
5. **The targets are high on purpose.** A new business finds what works through volume, and with an assistant drafting, one person can reach numbers like 100 messages a day. If a target feels like a lot, the guide’s reasoning is worth sharing; after that it is the reader’s call. Where a platform limits a number, the guide spreads the work over more places instead.
6. **Nothing goes out without the reader’s OK.** Every message, post and page is the reader’s to approve and send.
7. **About the publisher.** Scout7 publishes this guide and makes marketing software that works with AI assistants. The guide does not depend on it.

### Questions and decisions for this guide

**Ask first** (skip any the reader has already answered or shared)

- "What is your website built on: Shopify, Wix, Squarespace, WordPress, Webflow, or code you or a developer wrote?"
- "What are the addresses of your 5 most important pages: home, the page you send people to most, pricing or product, signup or contact or cart, and your top blog post?"
- "Which apps, plugins, chat boxes, popups and tracking tools are on the site, and which of them brought a sale, signup or enquiry in the last 30 days?"
- "Can you edit your site's code yourself, or would someone else have to?"
- "About how many visits a month does the site get?"

**Decisions along the way**

- 1.1: if all 5 pages pass LCP 2.5 s, CLS 0.1 and TBT 200 ms on the Mobile lab test, the work is done. Tell the reader to stop and go back to getting visitors.
- 1.1: use the real-visitor numbers if PageSpeed Insights shows them; otherwise the lab numbers. At a few hundred visits a month, expect no real-visitor data.
- 3.1: keep an app only if it brought a sale, signup or enquiry in the last 30 days, or checkout, payments, forms, the law or the one analytics tool needs it.
- 3.2: on Shopify, the reader edits a duplicate of the theme and previews home, product, cart and checkout before publishing it. If they are unsure about a line of code, ask the app's developer rather than guess.
- Step 5: skip it on Shopify, Wix and Squarespace. Only do it if "Document request latency" is flagged.
- If Total Blocking Time stays over 600 ms after the removals in step 3, the cause is in the theme or custom code; that is the one job worth paying a developer for.

**Get from the reader as they go**

- 1.1: the three numbers and the largest element for each of the 5 pages.
- 2.1: the file size and width of each top image, before and after.
- 3.1: the full list of apps and scripts, with keep or remove.
- Every day: which page was fixed and its new numbers.
- Every Sunday: the 5 rows, and any app, theme change or embed added that week.

---

Your site feels slow on your phone, PageSpeed Insights shows a red score, and you wonder if that is [why nobody buys](https://scout7.ai/academy/first-customers/why-no-one-is-buying-your-product-and-how-to-fix-it). Here is how to improve website speed on a small site in about a week: test 5 pages, fix the few things that make them slow, then stop. Speed is rarely the only reason a site gets [no sales](https://scout7.ai/academy/website-and-conversion/how-to-fix-a-shopify-store-with-traffic-but-no-sales) or [no signups](https://scout7.ai/academy/website-and-conversion/how-to-fix-a-saas-website-with-traffic-but-no-signups), but a page that shows nothing for 4 seconds on a phone loses people before they read a word.

## Do these three things

About 1 hour today, then 30 to 60 minutes a day for about a week. Once your pages pass, speed is no longer what stands between you and your next sale.

- **Today:** test your 5 main pages in PageSpeed Insights on Mobile, and write down each page's LCP, CLS and Total Blocking Time in one sheet. A page passes at LCP 2.5 seconds or less, CLS 0.1 or less and Total Blocking Time 200 milliseconds or less. These are the speed checks from your [technical SEO checklist](https://scout7.ai/academy/seo/technical-seo-checklist-for-a-new-website).
- **Every day:** fix 1 page, starting with the slowest. Shrink its top image and remove the apps and scripts it does not need, until all 5 pages pass. Then test every new page, [landing page](https://scout7.ai/academy/website-and-conversion/how-to-write-a-landing-page-that-converts) and app before it goes live.
- **Every Sunday:** test the 5 pages again, check your site builder's speed report and [Google Search Console](https://scout7.ai/academy/seo/how-to-use-google-search-console), and remove anything added that week that made a page slower.

## 1. Test your website speed on mobile with PageSpeed Insights

Test your 5 most important pages in Google's free PageSpeed Insights, on the Mobile tab, and write down three numbers for each. Most of your visitors are on a phone, and the phone test is the hard one, so it is the one that counts.

**Pick your 5 pages.** Use the pages a buyer sees on the way to paying you:

1. Your [homepage](https://scout7.ai/academy/website-and-conversion/how-to-write-homepage-copy).
2. The page you send people to most, such as a [landing page](https://scout7.ai/academy/glossary/landing-page) or your best product page.
3. Your [pricing page](https://scout7.ai/academy/sales/how-to-design-a-pricing-page) or a product page.
4. The page where people act: signup, [contact form](https://scout7.ai/academy/website-and-conversion/how-to-get-more-leads-from-a-b2b-website), booking or [cart](https://scout7.ai/academy/website-and-conversion/how-to-reduce-abandoned-carts).
5. Your most visited [blog post](https://scout7.ai/academy/seo/how-to-write-a-blog-post-that-ranks-on-google). If you have none, a second product or service page.

**Run the test.** Go to pagespeed.web.dev, paste the page's address and press **Analyze**. Stay on the **Mobile** tab. Run each page 3 times and keep the middle result, because the lab score moves a little each time from things like routing and ads.

**Read the right part.** The top of the report shows real visitors from the last 28 days, if Google has enough of them. A new site with a few hundred visits a month usually has none, and that is normal. Then use the lab test below it. Write down:

- **LCP** (Largest Contentful Paint): how long until the biggest thing at the top, usually an image or the headline, shows. Good is **2.5 seconds or less**. Over 4 seconds is poor.
- **CLS** (Cumulative Layout Shift): how much the page jumps around while it loads. Good is **0.1 or less**.
- **TBT** (Total Blocking Time): how long the page is too busy with scripts to react to a tap. Good is **200 milliseconds or less** on mobile. Google's real-visitor measure for this is INP (Interaction to Next Paint, good at 200 ms or less), but the lab test cannot tap, so TBT stands in for it.

Also write down what the **LCP breakdown** says is the largest element. It tells you what to fix in step 2.

**Use your site builder's own report too, if it has one.** On Shopify, go to **Analytics → Reports** and type "contentful" in the search box for the LCP report. It uses your real visitors from the last 30 days and marks the day you installed an app or changed your theme. On Wix, open **Site Speed** in your site's dashboard; it needs 10 or more visits in the last 7 days.

**Your speed sheet**

```text
Page | Address | LCP (s) | CLS | TBT (ms) | Largest element | Passes all 3? | Date
Home | | | | | | |
Most-visited page | | | | | | |
Pricing or product | | | | | | |
Signup, contact or cart | | | | | | |
Top blog post | | | | | | |
```

**The decision.** If all 5 pages pass all 3 numbers, stop here. Your speed is fine. Go back to the work that brings people, such as [getting traffic to your site](https://scout7.ai/academy/first-customers/how-to-get-traffic-to-a-new-website-for-free). If any number fails, carry on with step 2 on the slowest page.

- **Tool:** PageSpeed Insights and Google Sheets. Both free.
- **Time:** about 45 minutes for 5 pages, 3 runs each.
- **You will have:** a sheet with 5 rows, 3 numbers each, and the largest element named.
- **It worked if:** every row is filled in, and you know which pages fail and on which number.
- **Common mistake:** chasing the 0 to 100 score, or testing only on desktop. Google's lines are the three numbers. A page at 72 that passes all three is fine. The other mistake is testing once: one bad run is not a problem.

> **Example (made-up business: Ferns & Rinse): Ferns & Rinse's speed sheet**
>
> Ferns & Rinse is a made-up plant-based haircare brand with its own Shopify store. It sells one shampoo and one conditioner at $28 each, or both for $49.
>
> | Page                                       | LCP   | CLS  | TBT    | Largest element | Passes? |
> | ------------------------------------------ | ----- | ---- | ------ | --------------- | ------- |
> | Home                                       | 5.8 s | 0.04 | 480 ms | Top photo       | No      |
> | Shampoo + conditioner bundle               | 3.9 s | 0.22 | 520 ms | Product photo   | No      |
> | Shampoo                                    | 2.3 s | 0.03 | 160 ms | Product photo   | Yes     |
> | Cart                                       | 2.1 s | 0.02 | 190 ms | Headline        | Yes     |
> | "Sulfate-free shampoo for curly hair" post | 2.4 s | 0.05 | 150 ms | Headline        | Yes     |
>
> *Why it works: Two pages fail, and the sheet says why: a huge image on one, apps and a jumping banner on the other.*

## 2. Shrink the big image at the top of each page

The fastest way to improve website speed on a small site is to shrink the large image at the top of each slow page. In 2025, the biggest thing at the top of a phone page was an image on 3 out of 4 pages, so this is usually your LCP.

**Find the image.** In the report, open **LCP breakdown** to see which element it is, and **Improve image delivery** to see how much smaller each image could be. If **LCP request discovery** says the image is lazy-loaded, that is a fix of its own (below).

**Shrink it in Squoosh.** Squoosh (squoosh.app) is a free image tool from Google that works in your browser and does not upload your files. For each image:

1. Download the image from your site, or use the original.
2. Drop it into Squoosh.
3. Under **Resize**, set the width to the size it shows on a computer screen. A full-width banner rarely needs more than 2,500 pixels (Squarespace's own advice), and a product photo needs far less.
4. Set the format to **WebP** and lower the quality until you can see a change, then go one step back up.
5. Download it. Aim for **under 500 KB**, and much smaller if you can without the picture looking worse.
6. Upload it in place of the old one.

**Load the top image first, not last.** "Lazy loading" means an image waits until you scroll to it. That is right for images lower down, and wrong for the one at the top. Google says lazy-loading the top image "will always lead to unnecessary resource load delay". In 2025, about 1 in 6 pages still did it. On a builder, check the image block's settings for "lazy load" and turn it off for the first image only. If you can edit the code, add `fetchpriority="high"` to that one image.

**Swap sliders and autoplay videos at the top for one still image.** A slider loads several big images before anyone sees the first. Put your [short video](https://scout7.ai/academy/content-marketing/how-to-make-short-form-videos-for-your-business) lower down the page, so it only loads when someone taps it.

> **If you run a D2C store:** do every product photo, not only the top one. Shopify's advice is to size images to the size they show and to show fewer images on heavy pages. Follow Shopify's own format advice: upload a JPG for photos and a PNG only where you need a see-through background. Do not let shrinking cost you the detail shoppers zoom in on; that is part of a [product page that sells](https://scout7.ai/academy/website-and-conversion/how-to-write-a-product-description-that-sells).

> **If you run a SaaS:** the top image is often a large product screenshot or a demo video, and the page after it is the signup that starts your [free trial](https://scout7.ai/academy/retention/free-trial-vs-freemium-which-to-choose), so test that page too. Use a still screenshot, cut to the part that matters, and put the [demo](https://scout7.ai/academy/sales/how-to-do-a-product-demo) below it. If your page stays blank until its scripts load, the page is built in the browser, which is a problem for Google too: check it with the [indexing guide](https://scout7.ai/academy/seo/how-to-get-your-website-indexed-on-google).

- **Tool:** Squoosh (free) and the PageSpeed Insights report.
- **Time:** about 15 minutes a page.
- **You will have:** a top image under 500 KB on every failing page, sized to fit, and not lazy-loaded.
- **It worked if:** a new test shows LCP at 2.5 seconds or less, and **Improve image delivery** no longer lists that image.
- **Common mistake:** lazy-loading every image, including the top one, because a plugin or a blog post said "lazy-load your images". The other mistake is uploading stock photos at full size. Library images can be several MB.

> **Example (made-up business: Ferns & Rinse): The home page's top photo**
>
> **Before:** Top photo: PNG, 4,000 pixels wide, 3.4 MB, lazy-loaded. LCP 5.8 s.
>
> **After:** Top photo: JPG, 2,000 pixels wide, 210 KB, loads first. LCP 2.3 s.
>
> *Why it works: One file was most of the wait. Shrinking it moved LCP under the line without changing what shoppers see.*

## 3. Remove the apps, plugins and scripts you do not need

Every app, plugin, chat box, popup and tracking tag adds code that a phone must download and run before the page reacts. Remove every one that has not earned its place. This is usually what fixes a high Total Blocking Time.

**Make the list.** Open your apps or plugins page and write every one in your sheet. Then open **3rd parties** and **Render-blocking requests** in the PageSpeed report. They show the outside scripts that load and how long each one holds the page up. Add any you did not know about. If you cannot tell what a script is, paste its address into ChatGPT or Claude and ask which tool it belongs to, then check that answer against your apps page.

**Mark each one keep or remove, with this rule.** Keep it only if one of these is true:

- It brought a sale, signup or enquiry in the last 30 days, and you can show it. [Conversion tracking](https://scout7.ai/academy/measurement/how-to-set-up-conversion-tracking), where each sale came from or the app's own report will tell you.
- Checkout, payments, your forms, the law (such as a cookie notice) or your one analytics tool needs it.
- It is the one email signup form you use to [build your list](https://scout7.ai/academy/email-marketing/how-to-build-an-email-list-for-free), or the tool that sends your [welcome and cart emails](https://scout7.ai/academy/email-marketing/how-to-set-up-klaviyo-flows-for-a-new-store).

Everything else goes: the second analytics tool, the chat box nobody uses, the [countdown timer](https://scout7.ai/academy/retention/how-to-run-a-sale-without-training-customers-to-wait), the "someone just bought" popup, the old [retargeting](https://scout7.ai/academy/paid-ads/how-to-set-up-retargeting-ads) tag for ads you no longer run, the [review app](https://scout7.ai/academy/reviews-press-and-referrals/how-to-get-your-first-product-reviews) you replaced. Real [trust signals](https://scout7.ai/academy/reviews-press-and-referrals/how-to-make-a-new-store-look-trustworthy), such as reviews and a clear returns policy, do more than a popup that says someone just bought. Keep one [popup](https://scout7.ai/academy/email-marketing/how-to-write-an-email-popup-that-isnt-just-a-discount) at most.

**Remove them properly.**

- **Shopify:** uninstall the app in **Settings → Apps**. Shopify says "uninstalling an app doesn't automatically remove its code from your theme". So go to **Online Store → Themes**, duplicate your theme, and edit the copy with **Edit code**. Look in `layout/theme.liquid` and the `snippets` folder for the app's name, remove the line that loads it, then preview your home page, a product page, the cart and checkout before you publish the copy. If you are not sure, ask the app's developer how to remove its code. Shopify's free Horizon and Online Store 2.0 themes are built for speed; if yours is an old paid theme, try one of them.
- **WordPress:** deactivate the plugin, then delete it. WordPress's own advice is to "deactivate and delete any unnecessary plugins".
- **Wix, Squarespace and other builders:** remove the app or the code block, and remove any code you pasted into the site's header settings for a tool you no longer use.

**Cut fonts to 2 families.** Each extra font and weight is another file. One for headings and one for text is enough.

> **If you run an agency:** speed checks are a fast way to help a prospect. Run their 5 pages, and in your first message name one fix with its number, such as "your top image is 3 MB". It fits the "gap you can see" approach in [getting your first client](https://scout7.ai/academy/first-customers/how-to-get-your-first-client-as-a-marketing-agency), and it can turn a website job into monthly work or SEO work for your web design clients.

- **Tool:** your apps or plugins page, and the **3rd parties** part of the report.
- **Time:** 1 to 2 hours once, including checking the preview.
- **You will have:** a list of every app and script marked keep or remove, with the remove ones gone.
- **It worked if:** a new test shows Total Blocking Time at 200 milliseconds or less on mobile, and checkout or signup still works on your phone.
- **Common mistake:** installing a "speed booster" app to fix a site slowed by too many apps. It is one more app. The other mistake is uninstalling on Shopify and leaving the code behind.

> **Example (made-up business: Ferns & Rinse): Ferns & Rinse's app list**
>
> | App or script               | Brought a sale or needed?         | Decision             |
> | --------------------------- | --------------------------------- | -------------------- |
> | Reviews                     | Shown on every product page       | Keep                 |
> | Email signup form           | 31 signups in 30 days             | Keep                 |
> | Analytics                   | The one tool checked every Sunday | Keep                 |
> | Cookie notice               | Needed by law                     | Keep                 |
> | Second reviews app          | Replaced in month 1               | Remove + delete code |
> | Chat box                    | 0 chats in 30 days                | Remove               |
> | "Someone just bought" popup | No sales traced to it             | Remove               |
> | Countdown timer             | No sales traced to it             | Remove               |
> | Heatmap tool                | Second analytics tool             | Remove               |
> | Old Meta pixel copy         | Duplicate tag                     | Remove               |
> | Instagram feed              | Loads 12 images, no clicks        | Remove               |
>
> *Why it works: Each app had to show a sale or a job. 7 of 11 could not, so they went.*

## 4. Stop your page jumping while it loads

A page "jumps" when something loads late and pushes the text down, so a visitor taps the wrong thing. Give every image and video a fixed space, and make banners sit on top of the page instead of pushing it down. That brings CLS to 0.1 or less.

**Find what jumps.** Open **Layout shift culprits** in the report. It names the elements that moved the page.

**Fix the usual three:**

1. **Images and videos with no size.** Google's advice is to "always include width and height size attributes on your images and video elements". Most builders do this for you when you use their image block. Check images you pasted in with your own code or an embed.
2. **Banners, popups and cookie notices that push the page down.** Set them to show on top of the page (often called "overlay" or "floating"), or at the bottom, not inserted above the headline. A [popup](https://scout7.ai/academy/email-marketing/how-to-write-an-email-popup-that-isnt-just-a-discount) should also wait and stay small. Google's advice is: "Don't obscure the entire page with interstitials."
3. **Embeds that load late,** such as a review widget, a booking calendar or a social feed. Give their box a fixed height in the builder, or move them lower down the page.

- **Tool:** the **Layout shift culprits** part of the report.
- **Time:** about 30 minutes.
- **You will have:** the elements that moved the page fixed, moved or removed.
- **It worked if:** a new test shows CLS at 0.1 or less.
- **Common mistake:** a "free shipping" or discount bar that slides in at the top after the page loads. Load it with the page, or show it at the bottom.

## 5. Speed up a slow server, only if the test says so

Fix your hosting only if the report flags **Document request latency**, which means your server is slow to send the first part of the page. On Shopify, Wix and Squarespace, skip this step: they already run the servers, caching and a CDN (a network that sends your files from a server near each visitor) for you.

Shopify says its scanners' suggestions about "CDN, compression, caching, and server configuration" may not apply, because it already handles them. If you are on a hosted builder and a scanner tells you to "add a CDN", ignore it.

**On WordPress or your own host:**

1. **Turn on caching.** A caching plugin saves each page as a ready-made file, so the server does not build it for every visitor. Look in your hosting dashboard for a caching setting first. If there is none, install WP Super Cache (free, by Automattic) and turn on caching.
2. **Add a free CDN.** Cloudflare's Free plan includes a CDN and costs nothing. You point your domain's settings at Cloudflare, and it sends your files from its servers.
3. **If it is still flagged, move host.** WordPress's own advice says your hosting hardware "will have a huge impact on your site performance". Very cheap shared hosting is the usual cause.

> **If you run a B2B company:** this is the step most likely to matter for you, because many B2B sites run on WordPress with an old theme and a cheap host. Do it before you spend on getting [more leads from your website](https://scout7.ai/academy/website-and-conversion/how-to-get-more-leads-from-a-b2b-website).

- **Tool:** your host's caching setting or WP Super Cache, and Cloudflare Free. All free.
- **Time:** about 1 hour.
- **You will have:** caching on, and a CDN if you added one.
- **It worked if:** **Document request latency** passes on a new test, or it was never flagged.
- **Common mistake:** paying for a faster host or a CDN on a site builder that already includes them.

## 6. Fix one page a day, and test every new page before it goes live

Fix 1 page a day, starting with the slowest, until all 5 pass. Then test each new page, app or theme change on Mobile before it goes live. Speed gets worse one small addition at a time, so the only way to keep it is to check every addition.

At about 30 to 60 minutes a page, a week is enough for most small sites. Work in the order of steps 2 to 5 on each page, and retest after each fix so you know which one worked.

Once the 5 pages pass, the job changes. You are going to keep adding pages: a [new blog post every day](https://scout7.ai/academy/seo/how-to-write-a-blog-post-that-ranks-on-google), new [landing pages](https://scout7.ai/academy/website-and-conversion/how-to-write-a-landing-page-that-converts) for each campaign, new apps. Test each one on Mobile on the day you make it, and write its numbers in your sheet. If a new app makes any of your 5 pages fail, remove it or decide it is worth the cost, in writing.

If you are [testing two versions of a page](https://scout7.ai/academy/website-and-conversion/how-to-ab-test-a-landing-page), test both for speed first. A slower version can lose for that reason alone. If you pay for ads, test the page they land on before the ad goes live: a slow page is one cause of [ads that get clicks but no sales](https://scout7.ai/academy/paid-ads/how-to-fix-ads-that-get-clicks-but-no-sales).

- **Tool:** PageSpeed Insights and your speed sheet.
- **Time:** 30 to 60 minutes a day for about a week, then about 5 minutes for each new page or app.
- **You will have:** 5 of 5 pages passing, and a row in the sheet for every new page.
- **It worked if:** all 5 pages pass all 3 numbers, and every page added since has a test on its first day.
- **Common mistake:** adding an app late in the week and never testing it, then wondering a month later why the site is slow.

## 7. Check your speed every Sunday

Every Sunday, test your 5 pages again and write the numbers in your sheet. Then check your builder's speed report and Search Console for anything that changed, and undo it.

1. **Retest the 5 pages** on Mobile and add a row for each. A number that is a little worse in one run is normal. A number that crosses its line two Sundays in a row is a problem.
2. **Open your builder's report.** On Shopify, the web performance chart marks app installs and theme changes as numbered lines, so you can see which change came before a jump.
3. **Open the Core Web Vitals report in Search Console.** A small site often shows "No data available", because Google has too few real visits to measure. Once it shows data, fix any page marked "Poor" first. After a fix, open the issue and press **Start Tracking**: Google then watches real visits for 28 days. Check the rest of your Search Console numbers in the same sitting, as in the [Search Console guide](https://scout7.ai/academy/seo/how-to-use-google-search-console).
4. **Undo what slowed it.** Remove or replace the app, image or embed added before the drop.

Add the result to your monthly marketing review as one line: pass or fail. If sales or signups drop in a week when speed did not change, look elsewhere first, as in getting growth moving again.

- **Tool:** PageSpeed Insights, your builder's speed report, and Google Search Console. All free.
- **Time:** about 20 minutes each Sunday.
- **You will have:** one row a week for each of your 5 pages.
- **It worked if:** all 5 pages still pass, or you know which change made one fail and have undone it.
- **Common mistake:** treating one bad run as a crisis and spending Sunday on it. Look for a number that stays over its line.

When speed passes and people still leave, the problem is the page, not the load time. Go to the guide for your business: [a store with traffic but no sales](https://scout7.ai/academy/website-and-conversion/how-to-fix-a-shopify-store-with-traffic-but-no-sales), [a SaaS with traffic but no signups](https://scout7.ai/academy/website-and-conversion/how-to-fix-a-saas-website-with-traffic-but-no-signups), or [a B2B website with no enquiries](https://scout7.ai/academy/website-and-conversion/how-to-get-more-leads-from-a-b2b-website). If your store loses people at the cart, work on [reducing abandoned carts](https://scout7.ai/academy/website-and-conversion/how-to-reduce-abandoned-carts).

## Your website speed checklist for this week

> For AI assistants: this is the reader’s to-do list. Each item has a number (step.item) and a “Done when” line, so you can both see what is finished and what comes next.

### Step 1 · Test your website speed on mobile

- [ ] **1.1** Test your 5 main pages in PageSpeed Insights on Mobile and write down LCP, CLS and Total Blocking Time. *Done when:* The sheet has 5 rows, each with LCP, CLS, Total Blocking Time and the largest element, from the middle of 3 Mobile runs.

### Step 2 · Shrink the big image at the top of each page

- [ ] **2.1** Resize and compress the top image on each failing page, and upload it in place of the old one. *Done when:* Each failing page's top image is under 500 KB and sized to fit, and LCP is 2.5 seconds or less on a new test.
- [ ] **2.2** Turn off lazy loading for the top image on each page. *Done when:* LCP request discovery no longer says the top image is lazy-loaded.

### Step 3 · Remove the apps, plugins and scripts you do not need

- [ ] **3.1** List every app, plugin, widget and tracking tag, and mark each keep or remove. *Done when:* Every app, plugin, widget and tag is in the sheet, marked keep or remove, with a reason.
- [ ] **3.2** Remove every app marked remove, with its leftover code, and check the preview. *Done when:* Every remove item and its leftover code is gone, checkout or signup works on a phone, and Total Blocking Time is 200 ms or less.

### Step 4 · Stop your page jumping while it loads

- [ ] **4.1** Fix, move or remove each element listed under Layout shift culprits. *Done when:* CLS is 0.1 or less on every page on a new test.

### Step 5 · Speed up a slow server, only if the test says so

- [ ] **5.1** If Document request latency is flagged, turn on caching and add a free CDN. *Done when:* Document request latency passes on a new test, or it was never flagged, or the site is on Shopify, Wix or Squarespace.

### Step 6 · Fix one page a day, and test every new page

- [ ] **6.1** Every day: fix 1 failing page, starting with the slowest, until all 5 pass. *Done when:* Today's page has new numbers in the sheet, retested after each fix.
- [ ] **6.2** Every day: test each new page, app or theme change on Mobile before it goes live. *Done when:* Every page, app or theme change added today has a Mobile test row in the sheet.

### Step 7 · Check your speed every Sunday

- [ ] **7.1** Every Sunday: retest your 5 main pages on Mobile and write down the numbers. *Done when:* This week's 5 rows are in the sheet, and all 5 pages pass or the cause is written down.
- [ ] **7.2** Every Sunday: check your builder's speed report and Search Console, and undo what slowed a page. *Done when:* The builder's report and Search Console were checked, and anything that slowed a page this week is removed or undone.

## Frequently asked questions

### Does website speed affect SEO?

A little. Google says it "always seeks to show the most relevant content, even if the page experience is sub-par", and that good speed numbers do not guarantee a top ranking. It still recommends good Core Web Vitals. So pass the three lines, then spend your time on [pages that answer real searches](https://scout7.ai/academy/seo/how-to-find-keywords-for-seo) and [getting found on Google](https://scout7.ai/academy/seo/how-long-does-seo-take-to-work).

### What is a good PageSpeed Insights score?

90 to 100 is good, 50 to 89 needs work, and under 50 is poor. But the score is a lab test that moves between runs. What Google measures are three numbers: LCP 2.5 seconds or less, CLS 0.1 or less, and INP 200 milliseconds or less from real visitors. A page that passes all three is fine even with a score in the 70s.

### Why is my website slow on mobile but fast on desktop?

The Mobile test pretends to be a slow phone on a slow connection, so heavy images and scripts hurt far more. A real cheap phone on mobile data behaves the same way. Fix the top image and remove scripts first, as in steps 2 and 3. Desktop usually improves too.

### Should I pay for a speed app or a developer?

Not first. Shrinking the top image and removing apps you do not need fixes most small sites, for free, in a few hours. A speed app is one more app. Pay a developer only if Total Blocking Time stays over 600 milliseconds after step 3, because then the cause is in your theme or your own code. Read how to hire help without getting burned.

### What is the Agentic Browsing score in PageSpeed Insights?

It is a new, experimental check from Google that asks whether AI agents can use your site. It is not part of your speed score, and it shows passed checks, not a number out of 100. One of its checks is an llms.txt file, which you can [add with schema markup](https://scout7.ai/academy/ai-search/how-to-add-llms-txt-and-schema-markup).

## Sources

- [About PageSpeed Insights](https://developers.google.com/speed/docs/insights/v5/about). Google for Developers, updated 21 October 2024, checked 11 October 2026.
- [Web Vitals](https://web.dev/articles/vitals). web.dev (Google), updated 31 October 2024, checked 11 October 2026.
- [Optimize Largest Contentful Paint](https://web.dev/articles/optimize-lcp). web.dev (Google), updated 31 March 2025, checked 11 October 2026.
- [Optimize Cumulative Layout Shift](https://web.dev/articles/optimize-cls). web.dev (Google), updated 7 February 2025, checked 11 October 2026.
- [Interaction to Next Paint (INP)](https://web.dev/articles/inp). web.dev (Google), checked 11 October 2026.
- [Total Blocking Time](https://developer.chrome.com/docs/lighthouse/performance/lighthouse-total-blocking-time). Chrome for Developers, checked 11 October 2026.
- [Lighthouse is moving to performance insight audits](https://developer.chrome.com/blog/moving-lighthouse-to-insights). Chrome for Developers, Lighthouse 13, 10 October 2025, checked 11 October 2026.
- [Lighthouse agentic browsing scoring](https://developer.chrome.com/docs/lighthouse/agentic-browsing/scoring). Chrome for Developers, updated 5 May 2026, checked 11 October 2026.
- [Understanding page experience in Google Search results](https://developers.google.com/search/docs/appearance/page-experience). Google Search Central, updated 22 September 2026, checked 11 October 2026.
- [Understanding Core Web Vitals and Google search results](https://developers.google.com/search/docs/appearance/core-web-vitals). Google Search Central, updated 10 December 2025, checked 11 October 2026.
- [Core Web Vitals report](https://support.google.com/webmasters/answer/9205520). Search Console Help, checked 11 October 2026.
- [Web Almanac 2025: Performance](https://almanac.httparchive.org/en/2025/performance). HTTP Archive, published 15 January 2026, data from July 2025, checked 11 October 2026.
- [Improving your online store performance](https://help.shopify.com/en/manual/online-store/web-performance/improving-web-performance). Shopify Help Center, checked 11 October 2026.
- [Web performance reports](https://help.shopify.com/en/manual/online-store/web-performance/web-performance-reports). Shopify Help Center, checked 11 October 2026.
- [Store slower after uninstalling apps? How to check for leftover app code in your theme](https://community.shopify.com/t/672753). Shopify Community, 26 August 2026, checked 11 October 2026.
- [Understanding your Site Speed dashboard](https://support.wix.com/en/article/understanding-your-site-speed-dashboard). Wix Help Center, checked 11 October 2026.
- [Formatting your images for display on the web](https://support.squarespace.com/hc/en-us/articles/206542517-Formatting-your-images-for-display-on-the-web). Squarespace Help Center, checked 11 October 2026.
- [Optimization](https://developer.wordpress.org/advanced-administration/performance/optimization/). WordPress Developer Resources, updated 17 December 2025, checked 11 October 2026.
- [Avoid intrusive interstitials and dialogs](https://developers.google.com/search/docs/appearance/avoid-intrusive-interstitials). Google Search Central, updated 10 December 2025, checked 11 October 2026.
- [WP Super Cache](https://wordpress.org/plugins/wp-super-cache/). WordPress.org, version 3.1.4, checked 11 October 2026.
- [Cloudflare Free plan](https://www.cloudflare.com/plans/free/). Cloudflare, checked 11 October 2026.
- [Squoosh](https://squoosh.app/). Google Chrome Labs, checked 11 October 2026.
