How to Audit Your Own Shopify Store in an Afternoon (Free Tools Only)

Everything in this series comes back to one idea: the storefront is the variable nobody re-tests. Agencies change, ad budgets change, the theme built in 2022 stays. This is the piece that lets you test it yourself, in an afternoon, with tools that cost nothing. Three tools, one real phone, and a short list of numbers worth writing down.

The storefront is the variable nobody re-tests.

You'll finish with a page of measurements rather than a verdict. A verdict from a tool is easy to argue with. Your own numbers on your own store are not.

Tool One: Lighthouse, on the Mobile Profile

Open Chrome, go to your busiest collection page, open DevTools (F12), click the Lighthouse tab, choose Mobile, and run it. Do the same on your best-selling product page and the homepage. Run each twice and keep the second result.

The mobile profile matters. Lighthouse's mobile run slows the CPU down four times over and simulates a slow 4G connection, roughly 1.6 Mbps with 150 ms of latency. That isn't your office wifi on a MacBook. It's closer to a customer on a two-year-old Android on the train, which is who visits your site.

Ignore the big performance number. On a heavily customised Shopify theme with a normal app stack, the overall score bounces around and rarely tells you what to change. Write down three things instead: Largest Contentful Paint (LCP), Cumulative Layout Shift (CLS) and Total Blocking Time. Then scroll to the diagnostics and note which element Lighthouse names as the LCP element. On most Shopify stores it's the hero image or the first product image, and knowing which one is the first useful fact you'll get all afternoon. While you're there, check the "properly sized images" audit; the most common finding is a 750px image being served into a 285px card (You're Sending 750px Images Into a 285px Space), and the "reduce unused JavaScript" list will usually show two apps doing the same job (Are You Paying Twice For the Same Job?). Why the headline score itself is a trap is covered in Stop Chasing a Green Lighthouse Score.

Tool Two: Shopify's Web Performance Reports

Lighthouse is a lab test. Shopify's own reports are field data from your real visitors. In admin, go to Online Store > Themes and look at the performance summary, or go to Analytics > Reports and type "contentful", "next" or "cumulative" into the search box.

You'll see three Core Web Vitals, each at the 75th percentile, each ranked Good, Moderate or Poor. The thresholds match Google's: LCP at or under 2,500 ms, Interaction to Next Paint (INP) at or under 200 ms, CLS at or under 0.1. Data lags by up to 36 hours and only covers the last 90 days, so treat this as a trend tool, not a "did my change work this morning" tool.

Two reports are worth the extra ten minutes. The Page Type reports tell you whether it's the product template or the collection template dragging you down. The Over Time reports carry event markers for app installs and theme changes, so if CLS jumped in March you can see what went in that week. That's often the whole diagnosis. Layout Shift Is Why Customers Tap the Wrong Thing and When Your Support Chat Widget Is the Slowest Thing on Your Homepage cover the two most common culprits.

Tool Three: The Rich Results Test

Go to search.google.com/test/rich-results and paste in a product URL. This shows what Google can read about the product from your structured data, which is roughly what an AI shopping assistant can read too.

You're looking for two result types. Product snippets means Google understands it's a product and may show price and rating. Merchant listings is the one that matters for a store that sells: it needs the fields that make a listing transactable, including shipping and return details. Most Shopify themes emit the first. Far fewer emit the second, because the delivery and returns fields aren't in the default theme output (Delivery and Returns Are Missing From Your Structured Data).

Then read the details Google parsed. Is the name the title you'd want in a shortlist, or a SKU label (Your Product Titles Don't Say What the Product Is)? Is the description the one on the page, or a truncated meta description from three years ago (The Description an AI Reads Is Not the One on Your Page)? Is the brand field populated, and does it match the brand on the page (Your Storefront Says One Brand, Your Data Says Another)? For the category side, see Shopify Categories and Category Metafields.

While you have a browser open, paste a URL from your old domain, if you had one, into a redirect checker such as httpstatus.io. A chain of 302s is a leak, and the fix is small (A 302 Is Not a 301, and Your Old Domain Is Leaking). For the agent-side view of all this, Can an AI Agent Actually Buy From Your Store? walks through the accessibility checks underneath it.

Tool Four: A Real Phone, Ad Blockers Off

Now put the laptop away. Pick up a phone, not the newest one in the office, turn off wifi, disable any content blocker, and open your store in a normal browser tab rather than from a link inside an app that opens its own in-app browser.

Do four things and write down what you see.

Open your busiest product page and scroll until the add-to-cart button appears. Note roughly how many screens it took. On a 390 by 844 viewport, good is on the first screen or a short scroll below it (How Far Down Is Your Add-to-Cart Button? Go and Measure It has the full method). While you're there, count how many buttons share the same colour as add-to-cart (One Colour for the Action You Want), and check whether a review score is visible without scrolling, or only at the bottom of the page (Four Hundred Reviews and None of Them Where They Count).

Open a collection page. Count the screens before the first product card, and note whether the cards show price, rating and enough of the product to tell items apart (Your Collection Pages Are Hiding the Products, and Three Things Missing From Your Collection Page Product Cards).

Watch the homepage load with the screen held still. Did anything move after you started reading? Did a popup, a chat bubble and a sticky bar all arrive at once (The Popup, the Chat Widget and the Sticky Bar Are Fighting Each Other)?

Pinch-zoom the hero image. Is there text baked into it that a screen reader or a search engine can't see (Text Baked Into an Image Can't Be Read by Anyone, and Alt Text Isn't an Accessibility Checkbox)?

This is the check a desktop review never catches.

That's the whole argument of Your Desktop Site Looks Fine. That's the Problem.

Twenty Minutes on the Catalogue

Open your product export in a spreadsheet and sort by title. Look for the same product spelled three ways (The Same Product Has Four Different Names on Your Site), titles that describe the pack instead of the product, and variant option names that differ between products. Open your product with the most variants and count how many are shown at once before any guidance (Progressive Disclosure, and Fifty-Five Combinations and No Guidance). Then look at how many products have sold fewer than a handful of units in the last year (Does Your Catalogue Breadth Work Against You?). This is the tedious part and it's where most of the AI-visibility work lives.

If You Have Another Hour

Read your five most-visited blog posts and note the date on each (Content That Stopped in 2023). Find every health, safety or money claim on your product pages and check whether a named, qualified person stands behind it (Health, Safety and Money Claims: What Needs an Expert Name Against It). Then go through a first purchase as a new customer and see whether the brand story ever appears and whether you were asked a single useful question at checkout (Your Best Story Is the One a New Customer Never Finds, and Capture the Data at the First Purchase, Not the Third). If you sell consumables, check whether "subscribe and save" is a checkbox or a product (Subscription Is a Product, Not a Checkbox). And before anyone proposes a personalisation engine, reread the catalogue section above (Why the AI Personalisation Project Should Be Last, Not First).

What the Numbers Are Worth

On a customised theme with a normal app load, chasing a 90-plus Lighthouse score is usually not worth the money. Getting LCP under 2.5 seconds at the 75th percentile on your product template usually is, because it's measured on real customers. CLS under 0.1 is almost always worth it and often cheap. INP is the one most stores can leave until the other two are green.

The structured data findings are worth more than the speed findings for most stores. They're fixable without touching the theme's performance budget, and they compound as more shopping moves through assistants.

Where This Leaves You

Start with Five Years of Ad Spend and a Flat Revenue Line, which is where this series began, and What a Chatbot Can't Tell You About Your Store, which explains why your numbers matter more than a general answer.