Free every Thursday · 59,694 Readers
Fastlane Insider Newsletter: What's working for Shopify in the AI era?
Free every Thursday · 59,694 Readers
Fastlane Insider Newsletter: What's working for Shopify in the AI era?
AI will happily tell you why your store is slow. Some of that advice is excellent. Some of it quietly breaks your checkout, and customers rarely tell you. They just leave.
Store owners ask AI for speed help because it turns dense speed reports into plain steps in seconds, without the cost of a developer. More owners now paste a PageSpeed Insights report into ChatGPT or Claude and follow whatever it says. I understand why. The reports are full of terms like “render blocking resources,” developers are expensive, and the AI answers in plain English straight away.
I’ve built websites since 2005 and worked in SEO for more than 18 years, and I’ve seen this go both ways. AI can take a store owner from confused to confident in an afternoon. It can also take a working checkout offline with one confident suggestion, because it doesn’t know which scripts on your site take payments. It has never seen your checkout, your theme or your gateway settings unless you tell it.
When the advice is right, the gains are real. One homewares store I worked on had a mobile Largest Contentful Paint (LCP) of 5.8 seconds on its product pages. The main product image was a 2.4MB JPEG, and the theme was lazy loading it. We converted the images to WebP, resized them to the theme’s actual display size, and stopped lazy loading the main image. LCP dropped to 2.1 seconds, with no other changes. That is one store’s result, and it depended on the image being the bottleneck, but it shows the kind of fix an AI can explain well and that carries almost no risk to checkout.
Speed matters more on a store because a lost visit is a lost sale, not just a lost page view. A widely cited 2016 Google study found that 53% of mobile visits are abandoned when a page takes longer than three seconds to load, a figure Google repeated in its mobile page speed industry benchmarks report. The study is old and covered mobile sites in general, so treat it as direction rather than a forecast for your store. Slow pages also cost trust: eCommerce Fastlane’s guide to the website features that calm anxious buyers explains how a page that hangs or jumps makes shoppers wonder whether the checkout will work at all.
Google measures page experience with three Core Web Vitals, judged at the 75th percentile of real visits:
Table 1: Google’s “good” thresholds for Core Web Vitals.
AI genuinely helps with four speed jobs: explaining reports, finding heavy plugins and scripts, fixing images, and writing small, specific code. Used well, an assistant is like having a patient developer explain your speed reports and answer the follow up questions you would feel silly asking a person.
Paste in a PageSpeed Insights or GTmetrix report and ask what the three biggest problems are and why. AI is good at turning “reduce unused JavaScript” into something like “your page builder loads a large script on pages that don’t use it.” Ask it to rank the problems by likely impact, and to say which ones touch the cart or checkout.
Give it your active plugin list and the slowest requests from a waterfall chart. It will usually spot the usual suspects: sliders, social feeds, chat widgets, page builders and duplicate analytics tags. This is often the cheapest win of all, because a plugin you remove costs nothing to maintain and cannot conflict with checkout later.
Product images are the most common cause of slow stores, and the safest thing to fix. AI can explain WebP and AVIF formats, suggest image sizes for your theme, and write a script to batch compress a folder. It can also explain why your main product image should not be lazy loaded: Google’s research on the performance effects of too much lazy loading shows how deferring the largest image delays the metric that matters most. eCommerce Fastlane’s five step product page speed audit is written for Shopify, but its rules on image sizing and hero images apply to a WooCommerce product page just as well.
Ask for one thing at a time, such as a caching header, a rule to preload your main font, or a WP-CLI command to clear expired data. Narrow requests get safe, testable answers. Broad requests (“make my site faster”) get broad answers, and broad answers are where the next section’s problems come from.
Six AI speed suggestions commonly break WooCommerce stores: caching every page, delaying all JavaScript, disabling cart fragments, removing unused CSS and JavaScript, combining files, and cleaning the database. These suggestions are fine on a blog. On a store, each one can break carts, logins or payments. None of them is wrong in principle. They just need exclusions the AI doesn’t know to add.
Table 2: Common AI speed suggestions, what breaks on a store, and the safe version.
WooCommerce’s own documentation backs up the caching row. Its guide to configuring caching plugins for WooCommerce says the cart, checkout and My Account pages must stay out of the cache, lists the cookies to exclude, and even recommends avoiding JavaScript minification with one popular caching plugin. Many caching plugins apply those page exclusions automatically when they detect WooCommerce, but server level caches and CDNs set up separately often don’t, so confirm the exclusions rather than assuming them.
The cart fragments row needs one update that many AI answers miss. Since version 7.8, WooCommerce only loads the cart fragments script on pages with the Mini-Cart widget, or where a theme or plugin asks for it, as its post on best practices for the cart fragments API explains. If an AI tells you to disable it everywhere, check first whether your theme still uses it, and test the mini-cart after any change.
Payment providers often need their scripts to load early. Stripe, for example, recommends including Stripe.js on every page of a site, not just checkout, because it uses browsing signals to help detect fraud. Stripe is fine with the script loading asynchronously; the problem is a “delay all JavaScript until interaction” setting, which can stop express checkout buttons from appearing, or make a card form fail at the final step. Customers rarely report it. They just leave.
In my own WordPress and WooCommerce development work, payment and cart scripts are the first exclusions I add to any speed plugin, before any other setting changes. It takes five minutes and prevents the most expensive kind of speed fix: the one that makes the store faster and the checkout broken.
The safe workflow for AI speed fixes is five steps: measure a baseline, ask with full context, change one thing on staging, test checkout, then ship and watch. If checkout fails at any point, roll back and try a narrower change.
Figure 1: A safe workflow for AI speed fixes. If checkout fails, roll back and try a narrower change.
The phone test matters most. Express pay buttons and delayed scripts often behave differently on mobile, and mobile is where most speed problems show up first.
A prompt gets safer answers when it names your setup and asks the AI to list exclusions and an undo step for every suggestion. The quality of the advice depends on what you tell the AI. Compare “how do I speed up my site?” with a prompt like this:
I run a WooCommerce store on [theme], with [caching plugin].
Payment gateways: [Stripe, PayPal]. Page builder: [none / Elementor].
Here is my PageSpeed Insights report for a product page: [paste].
Suggest the three changes with the biggest impact.
For each one, list anything that must be excluded so the cart, checkout, My Account page and payment scripts keep working, and tell me how to undo it.
Asking for exclusions and an undo step changes the answers. The AI starts thinking about what could go wrong, not only about the score. Keep the final decision with a person, too: eCommerce Fastlane’s look at AI tools that draft changes and wait for approval makes the same case for promotions, where the safest design has the AI fill in the form and a human press save.
If the answer comes back vague, ask a follow up: “Which of these changes touches the cart or checkout, and how would I know if it broke?” A good answer names the page to test and the symptom to look for. A weak answer tells you to clear your cache and try again.
The workflow stays the same at every size, but who runs it and how much you test changes as revenue grows. A store doing a few orders a day can afford to learn by doing; a store doing hundreds cannot afford a broken checkout for even an hour.
If you’re just starting, focus on images and plugins first. They are the safest fixes, AI explains them well, and they usually deliver the biggest gain on a young store. Leave advanced settings like JavaScript delay switched off until you have a staging site.
If you’re growing and a freelancer or agency helps with the site, use AI to understand the reports before you brief them. You’ll ask better questions, spot padded quotes, and know which changes need a staging test. Ask whoever makes the change to show you the checkout test, not just the new score.
If you’re scaling, treat every speed change like a release: staging, a written test plan, a rollback ready to go, and a check of real visitor data in Search Console’s Core Web Vitals report after launch rather than a single lab score. At this stage AI is most useful for reading reports and drafting the test plan, not for deciding what ships.
AI is now the fastest way for a store owner to understand why a site is slow, and I’d encourage every owner to use it. Just don’t treat a higher PageSpeed score as the goal. The goal is a faster store that still takes money.
Treat AI like a sharp junior developer. It’s quick, it knows a lot, and it has never seen your checkout. Give it context, make it explain itself, and test everything before your customers do.
About the author: Fahad Raza is a WordPress developer and SEO consultant who has built websites since 2005 and worked in SEO for more than 18 years, including eight and a half years on the IKEA Australia and New Zealand website. He runs WordPress Developer Sydney, building fast, custom WordPress and WooCommerce sites for small businesses, and WordPressSEO.com.au, which offers sprint based SEO for WordPress sites.
ChatGPT can help fix a slow WooCommerce store by explaining speed reports, spotting heavy plugins and oversized images, and writing small, specific code, but it should not decide what goes live. It cannot see your checkout, your payment gateways or your theme unless you describe them, so its broad advice can break carts or payments. Paste in a PageSpeed Insights report with your theme, caching plugin and gateways, ask for the three highest impact changes with their exclusions and an undo step, then apply one change at a time on a staging copy and test checkout after each one.
The cart, checkout and My Account pages should never be cached on a WooCommerce store, because they show information specific to each shopper. WooCommerce’s own caching guidance also lists cart and session cookies to exclude, so a cached page is never served to someone with items in their cart. Many caching plugins add these exclusions automatically when they detect WooCommerce, but server level caches and CDNs configured separately often do not. After setting up caching, add a product to the cart as a guest, check the cart count updates, and confirm the My Account page loads for a logged in customer.
Delaying JavaScript on WooCommerce is safe only when payment, cart and checkout scripts are excluded from the delay. A setting that holds every script until the shopper interacts can stop express pay buttons from appearing, break card forms at the final step, or stop add to cart from working. Payment providers such as Stripe recommend loading their script on every page to support fraud detection, so it should load normally or asynchronously, never wait for a tap. Add the exclusions before you switch the setting on, then run a full test order in test mode on desktop and on a phone.
You should disable WooCommerce cart fragments only on pages that have no mini-cart, and only after testing. The script keeps the mini-cart count up to date without a page refresh, so removing it where a mini-cart exists can leave shoppers looking at the wrong count. Since version 7.8, WooCommerce already limits the script to pages with the Mini-Cart widget or where a theme or plugin requests it, so many stores need less work here than older advice suggests. Check whether your theme still loads it, change one page type at a time, and confirm the cart count updates after adding a product.
Test checkout after a speed change by running a full purchase as a guest and as a logged in customer, on desktop and on a phone. Add a product and confirm the cart count updates, apply a coupon, change the shipping address, and complete a test order with your gateway in test mode, including express buttons such as Apple Pay or PayPal. Then check the My Account page. Do this on staging before going live, and again on the live store. For the next seven days, compare orders and checkout completion against the week before the change.
