Shopify SEO Expert

Duplicate collection and filter URLs

This is the classic Shopify trap. The same product exists at /products/x and at /collections/anything/products/x, and Shopify's default canonical usually handles the simple case — but the moment you add filters, sort orders and tag pages, you get /collections/shoes/blue, /collections/shoes?filter.v.option.color=blue and /collections/shoes/blue?sort_by=price-ascending all serving near-identical content. Google crawls all of them, splits the ranking signals between them, and often picks the wrong one to show. I map what is actually being crawled, set the canonical rules, and cut the rest off.

Index bloat from variants and tag pages

A store with 1,200 products and eight tags per product can generate tens of thousands of crawlable tag URLs, none of which anyone searches for and all of which consume the crawl budget your real pages need. The fix is not one setting — it is deciding which of those pages have genuine search demand behind them, keeping those, and removing the rest from the index properly rather than with a blanket robots block that also hides the signals.

Canonicals, redirects and the host problem

A canonical tag that points at a URL which redirects is not a hint, it is a contradiction, and Google resolves it by not indexing the page. Same for a sitemap full of redirecting URLs. This is exactly the failure I found and fixed on my own site this week, and it is far more common than it should be — it happens quietly every time a store switches between www and non-www, adds a domain, or migrates and keeps an old base URL in a template.

Pagination and infinite scroll

If page two of a collection cannot be reached without JavaScript, the products on it may never be discovered. Infinite scroll implemented badly hides most of a catalog from crawlers while looking perfect to a human. I check what is reachable from a plain HTML fetch, because that is what the crawler sees.

Internal linking and site architecture

Internal links are how importance moves through a site, and most Shopify stores leak it — every product in one flat collection, no hub pages, navigation that links to twelve things and body copy that links to nothing. I design the collection tree around how people search rather than how the warehouse is organised, and build the collection-to-product and blog-to-collection links that make the structure legible.

Search Console as the source of truth

Not a third-party score out of 100. The Page indexing report tells you which of your URLs Google has excluded and why, in Google's own words. Most stores have never had anyone read it line by line. That report is where I start on every engagement, and it is where you will see the first results.

What actually earns a richer listing

Product with a valid Offer — price, currency and availability — is what puts price and stock into the search result. BreadcrumbList replaces the raw URL with a readable path. FAQPage can earn expandable questions under your listing. AggregateRating can show stars, but only when the reviews are genuinely on the page and genuinely from customers. Organization ties the store to a logo and social profiles in the knowledge panel.

What is quietly ignored

Marking up content that is not visible on the page. Review markup you wrote about yourself. Stuffing keywords into schema properties. Article markup on a product page. Google either drops these or, for the review case, applies a manual action. Plenty of Shopify SEO apps emit exactly this kind of markup by default and report it to you as a green check.

The double-schema problem

Your theme emits Product schema. Your review app emits its own. Your SEO app emits a third. They disagree about price or availability, and Google gets to choose which one to believe. I audit everything a page is emitting, consolidate it into one @graph the theme controls, and turn off the duplicates at the source.

Written into the theme, not injected by an app

App-injected schema breaks when you change themes, disappears when you cancel the subscription, and usually renders after the page loads. Markup in the Liquid template ships in the HTML, survives, and is yours. It also costs nothing per month.

Validated, then re-validated

Every change goes through the Rich Results Test and the Schema.org validator before it ships, and through Search Console's enhancement reports after. Schema that validates in a tool but throws errors on live URLs is common, and the only way to know is to check the live URLs.

What Shopify gives you by default

An automatically generated robots.txt that already blocks the cart, checkout, account and internal search pages, and an auto-generated sitemap.xml that splits into child sitemaps for products, collections, blogs and pages. For a small store this is fine. It is genuinely a good default.

Where the defaults fall short

The default rules do not know about your filter parameters, your tag pages, or the thousands of near-duplicate URLs your specific merchandising creates. The generated sitemap includes every published product and page whether or not it should be indexed, so a store with a large archive of thin or discontinued products is asking Google to spend its crawl budget on them. And nothing in the default reconciles the sitemap against your canonicals.

robots.txt.liquid customisation

Shopify lets you override the generated file with a robots.txt.liquid template, which is the correct place to add crawl rules, block parameter patterns that produce duplicates, control specific crawlers, and add sitemap references. It is also a file where one careless line can deindex a store, so it gets version-controlled, reviewed and tested against real URLs before it goes live.

AI crawlers are a separate decision

GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot and a dozen others each identify themselves and each can be allowed or blocked independently. That is a business decision, not a technical one: blocking them protects your content and removes you from AI answers, allowing them does the reverse. Most store owners have never made the decision consciously — they inherited whatever the default was. I make sure it is a choice.

Sitemap discipline

Every URL in the sitemap should return 200 and should be the canonical version of itself. No redirects, no noindexed pages, no URLs pointing at a host that forwards somewhere else. It sounds obvious. It is the single most common thing I find broken, and it was broken on this site until I fixed it.

Why a brand wants to be the source an AI names

When someone asks an assistant to recommend a product, the model answers from what it can retrieve and it attributes the answer to a handful of sites. If your store is not one of them, a competitor is — and the customer never sees a search result page where you could have appeared. Worse than being absent is being present and wrong: models will state your price, your shipping policy or your return window from whatever they can find, and if that information is not clearly structured on your site, they will infer it. Being citable is partly about traffic and mostly about controlling what is said about you.

llms.txt and llms-full.txt

A plain-text file at the root of your domain that states, in the model's own preferred format, who you are, what you sell, what is true about your policies, and which pages matter. llms.txt is the summary; llms-full.txt carries the detail. There is no search volume behind the term yet and no search engine requires it — that is precisely why it is cheap to do now. This site publishes both, and you can read them.

Content an answer engine can actually lift

Models quote passages, not pages. That rewards a specific shape of writing: a direct question as a heading, a complete answer in the first two sentences under it, facts stated as sentences rather than implied by a table, and no critical information trapped in an image or loaded by JavaScript. Most product pages fail this on the third point alone. FAQ sections with real questions and self-contained answers are the highest-leverage single change for most stores.

Structured data does double duty

The same Product, Offer and FAQPage markup that earns rich results in Google also gives a retrieval system an unambiguous statement of your price, availability and answers. You are not doing two jobs. You are doing one job that pays out in two places, which is the main reason GEO work is not a separate budget line.

Crawler access, checked rather than assumed

None of the above matters if GPTBot is getting a 403 from your CDN or your bot-protection app is silently blocking ClaudeBot. I verify what each crawler actually receives instead of assuming robots.txt is being honoured by the layers in front of it.

How this gets measured, honestly

There is no Search Console for AI answers yet. What can be measured: referral traffic from chatgpt.com, perplexity.ai and similar sources in GA4, whether the crawlers are fetching you in your server logs, and manual spot-checks asking the assistants your own category questions and recording who they name. That is a real, repeatable measurement — it is just not a rank tracker, and I will not pretend it is one.

Clusters, not scattered posts

One substantial hub page for a topic your store should own, surrounded by articles that answer the specific questions around it, all linked to each other and all linked into the collection they support. Twelve unrelated posts on twelve unrelated topics build nothing. Twelve posts around one topic build an argument that your site is the place to read about it.

Topics chosen from demand, not from opinion

I pull real search volume and difficulty data before anything gets written, and I would rather target a 90-searches-a-month term where you can plausibly rank than a 20,000-a-month term where the first page is Amazon, Wikipedia and three publishers with twenty years of links. The map of what is worth targeting comes before the calendar.

Every article routes to something buyable

In-body links from articles into the exact collection or product they discuss, with anchor text that describes the destination. This is how blog traffic turns into revenue instead of into a chart you show at a meeting, and it is also how a collection page accumulates the internal links it needs to rank.

Product and collection copy counts as content

Collection pages with a real introduction and a real answer to what a shopper is choosing between will outrank a bare product grid. Product descriptions that answer the questions your support inbox gets will outrank a manufacturer's copy pasted onto four hundred other stores. This is usually the cheapest content work available and almost always the last thing anyone does.

Days: the fixes that are binary

Some things are not a matter of degree. A page is either eligible to be indexed or it is not. If your canonicals point at a host that redirects, or your sitemap lists URLs that 301, those pages will not be indexed — and no content change alters that. Fixing it flips a page from excluded to eligible, and Search Console's Page indexing report reflects it typically within a few days to two weeks after a recrawl. That part is verifiable and you can watch it happen in your own account. What position the page then takes is a completely separate question with a completely different answer.

Days to weeks: structured data and rich results

Valid markup shows up in Search Console's enhancement reports shortly after a recrawl, usually within days. Whether it changes how your listing looks in the results is Google's call, not mine — rich results are an eligibility, not an entitlement, and Google withholds them from plenty of pages with perfectly valid schema. When they do appear it is typically a matter of a few weeks. Expect improved eligibility with certainty and improved appearance with confidence, not certainty.

Weeks to months: ranking movement

Once pages are indexable and the duplication is resolved, movement usually shows up first on long-tail and mid-tail terms — the specific product and category phrases where the competition is thinner. Typically that begins somewhere in the four-to-twelve-week range and continues after. Head terms move more slowly and usually need links and brand signals rather than more code, which is honest work but not work I can do for you in a sprint.

Months: content and topical authority

A content cluster takes roughly three to nine months to carry meaningful traffic, because it needs to be written, indexed, linked and then trusted. If someone tells you an article will rank in two weeks, they are describing a keyword nobody searches for. This is the part of SEO that requires patience, and the part that keeps paying after the engagement ends.

The size of the outcome is set by your store, not by me

Two results already published on this site: an international luxury fashion brand at +455% sessions, +114% orders and +74% revenue over the engagement, and a US-based beauty and lifestyle brand at +254% in total sales. Those are real numbers on real stores. They are not a forecast for yours. They happened in specific niches with specific catalogs, and technical SEO was one of several things moving at the same time — catalog data, page speed and marketing all changed too. I publish them as evidence that I have done the work, never as an average you should expect.

Some number one results cannot be won, and I will tell you which

In several of the highest-intent Shopify queries, the result sitting at number one is a page the platform itself owns — shopify.com/partners/directory, apps.shopify.com. In adjacent categories it is the same story with the vendor's own site. Those positions are brand property and no amount of content outranks a platform on its own terms. The realistic target is a position beneath them, or the long tail where the platform has no page at all. I know this precisely because I measured it on my own site and it was not the answer I wanted. I would rather tell you that on the first call than sell you a number one you were never going to get.

What I will actually commit to

Not rankings, and not a pay-when-you-rank arrangement, because both require pretending I control an algorithm I do not. What I commit to is the work and the reporting: a full audit with findings you can verify yourself, the fixes implemented rather than recommended, documentation of every change, and a straight answer every month about what moved, what did not, and what I got wrong.

How a Shopify SEO engagement runs

1. 01

   Free 30-minute call

   You tell me the store URL and what you think is wrong. I look at it while we talk — index coverage, what is actually being crawled, whether the schema validates, whether the sitemap is honest. If the problem is not SEO, or if you do not have enough traffic yet for SEO to be the right spend, I say so on that call.

2. 02

   Shopify SEO audit

   A fixed-price technical audit: crawl of the whole store, Search Console indexing analysis, structured data validation on real URLs, robots and sitemap review, internal link structure, and an AI-crawler and citation check. You get a prioritised list with effort and expected impact against each item, and you own it whether or not you continue with me.

3. 03

   Implementation, highest impact first

   The indexability and duplication fixes ship first because they are the ones that gate everything else. Everything is built on a duplicated theme, previewed, and pushed with a written change log. Nothing goes to your live store before you have seen it.

4. 04

   Verify, then keep going

   After launch we watch the Page indexing and enhancement reports together — those move first and confirm the fixes landed. From there it is either a handover with documentation, or a monthly retainer for the slower work: content clusters, internal linking, and the next round of what the data says to do.