Most store audits I run open the same way. Hundreds of URLs, a respectable traffic line, and almost nobody on the team able to tell me which page is supposed to do which job. That confusion is the real problem, not the keyword list.
Ecommerce SEO works when every page type has one job: categories capture demand, products convert it, guides answer the questions people ask before they buy, and technical controls stop filters and duplicates from eating crawl attention. Architecture and commercial priorities come first. The content comes after.
Prioritise by business value
Before I touch a single page, I build one list of categories and products and score each on a few things at once: real search demand and buyer intent, margin, whether stock is reliable, conversion rate, seasonality, how strong the competition is, and what organic visibility already exists. No single number wins this argument on its own. A category can have great demand and terrible margins, or sell beautifully but never come back in stock.
Your first SEO target should sit where commercial value meets a credible ability to win. A high-volume query attached to an unstable, low-margin product looks great on a slide and rarely pays the bill.
Design a clean store architecture
A practical hierarchy is:
home → department → category → subcategory → product
Keep important pages within a small number of clicks, use descriptive navigation, and add breadcrumbs. Avoid creating multiple indexable paths to the same inventory.
Assign query ownership:
- category pages own product-family intent;
- product pages own model or SKU intent;
- guides own questions, comparisons, and use cases;
- policy pages own delivery, return, warranty, and payment questions.
Build category pages for decisions
The query that lands on a category page is almost always a shortlist query, not a single-product query. Someone searching "running shoes for flat feet" or "split air conditioner for a small room" has not chosen yet. They are asking the page to narrow the field. A grid of forty products with no guidance does the opposite of that job.
So a category page needs more than a product grid. Include:
- a concise description of who the category is for, and who it is not for;
- useful subcategories and filters that match how buyers actually narrow (size, room, use case, budget band), not just how your catalogue is structured internally;
- the buying criteria that matter for this category, named plainly;
- the differences between the important product types, so a beginner can self-select;
- stock and delivery information where it changes the decision;
- links to the guides and policies a hesitant buyer reaches for;
- a short FAQ built from real buyer friction, not invented questions.
Keep the main shopping experience near the top. Long copy should support choice, not bury products. A common pattern that works: a tight intro and the grid above the fold, then the criteria, comparisons, and FAQ below it for the people who scroll because they are still deciding. The buyers who already know what they want never read past the grid, and that is fine.
One test I apply on every category page: could a person who knows nothing about the category read this and confidently pick one product over another? If the answer is no, the page is a catalogue listing, not a decision page, and it will lose to the competitor that helps.
Make product pages unique and complete
Use manufacturer facts as inputs, not final copy. A strong product page answers:
- what the product is;
- who it suits and does not suit;
- important specifications;
- compatibility, dimensions, materials, or care;
- what is included;
- delivery, returns, and warranty;
- genuine reviews and questions;
- alternatives when unavailable.
Do not invent experience or benefits. I have cut more "premium craftsmanship" lines than I can count, and nobody misses them. Specific factual copy beats generic persuasion almost every time, because a shopper deciding between two products is hunting for a reason to choose.
The richest source of product-page content is usually free and already sitting in your inbox: the pre-sale questions customers ask before they buy, and the reasons they give when they return something. If three buyers a week ask whether a charger is included, that question belongs on the page in plain words, not buried in a spec table. If a product gets returned because the sizing runs small, a sizing note on the page prevents the next return and the next support ticket at the same time. Treat the returns log and the support queue as a content brief that writes itself.
A product page that is genuinely complete also reduces the work after the sale. The store that explains compatibility, dimensions, and what is in the box accurately gets fewer "this isn't what I expected" returns than the store that wrote three adjectives and a hero image. In commerce, content quality and operational cost are the same problem viewed from two ends.
Give the page schema it can prove
Structured data does not create rankings or AI citations by itself, but it removes ambiguity for anything reading the page — search engines, shopping feeds, and AI answer engines alike. Mark up Product with the SKU, brand, and a description that matches the visible copy, paired with an Offer that pulls price, currency, and availability from the same source that renders the page — never a static value a developer set once and forgot. The moment schema says "in stock" and the page says otherwise, every system reading that data learns to distrust the rest of it.
Add AggregateRating and Review only when the ratings are real and the underlying reviews are visible on the page. A number pulled from a private ops sheet with no matching on-page reviews is a fast way to fail a manual spot-check and gain nothing from the rich result it was chasing. On category pages, a well-marked Breadcrumb and, where it fits, an ItemList clarify hierarchy without dressing up a listing as a single product. Validate schema against live prices and stock on a schedule, not once at launch — catalogues drift, and a rich result built on stale data spends the click it earns on a bad first impression.
The reviews worth marking up are the same ones worth asking for in the first place: post-purchase prompts, the questions customers actually send before buying, and photos buyers submit unprompted. That is genuine user-generated content, and it does two jobs at once — it gives the product page the specific, first-hand detail generic manufacturer copy cannot fake, and it gives you material to mark up honestly instead of manufacturing a rating that will not survive scrutiny.
Control filters, variants, and duplicates
Faceted navigation is where most large-store crawl budget dies quietly. Every combination of colour, size, brand, and price filter can mint its own URL, and a catalogue of a few hundred products can balloon into tens of thousands of near-identical pages that search engines waste time crawling and that dilute the few pages you actually want ranked.
The decision is which filter combinations earn an indexable landing page and which do not. A useful rule: index a combination only when it has real, distinct search demand and you would write unique guidance for it. "Running shoes" plus "for flat feet" is often a real query worth its own indexable page. "Running shoes" plus a specific hex colour and a price slider position usually is not. Everything else should stay crawlable for users but kept out of the index through your technical setup, so the index holds the pages that deserve to compete and nothing else.
Also review:
- product variants and canonical rules;
- parameters for sort and tracking;
- duplicate category paths;
- out-of-stock handling;
- discontinued product redirects;
- pagination or load-more behaviour;
- internal search pages;
- structured data matching visible price and availability.
Use content to resolve buying friction
Here is the part most agencies leave out of the proposal: the majority of stores do not need more content. They need to stop publishing for a quarter and fix the commercial pages that already carry buyer intent. Publishing is the easy work. Repairing the category tree is the boring work that actually moves revenue.
So create content only when it supports a commercial question:
- size or compatibility guides;
- product comparisons;
- “best for” decision guides with transparent criteria;
- care and maintenance;
- delivery and COD explanations;
- replacement and upgrade paths.
Link guides to the relevant category and products, and link back from commercial pages where the guide reduces uncertainty.
Returns and COD content carries more weight than people think
In cash-on-delivery markets, the returns and delivery explanation is not a footnote. It is part of the conversion decision and part of the margin. A buyer who is unsure whether they can refuse the parcel at the door, or what happens if it does not fit, hesitates, and in COD that hesitation often becomes a refused delivery you still paid the courier to attempt.
Here the operational reality and the content task connect directly. In one anonymised store, 137K EGP in ad spend produced 564K EGP in sales on paper, a 4.1× gross return. After delivery, the real picture was a 1.9× delivered return with roughly a third of COD orders coming back. Some of that gap is logistics and audience quality. But part of it is avoidable confusion that clear content removes: a returns page that states the policy without hedging, a delivery page that explains COD steps and timing, sizing and compatibility notes that stop the wrong product from being ordered at all. Most of that store's conversation happened on WhatsApp and Messenger, which means the same questions were being answered by hand, one chat at a time, when a clear page could have answered them once.
That is the lesson to carry, not the multiple. Your return rate, margin, and delivery performance differ. The principle holds: in COD, returns and delivery content is conversion content and cost content at the same time. For the full reporting picture behind numbers like these, see real Facebook ROAS and reconciling Meta, CRM, and collected revenue.
Internal linking that earns its place
Most store internal linking is decoration: a "related products" widget and a footer full of links nobody clicks. Linking that actually moves rankings and revenue does three jobs deliberately. It passes authority from your strongest pages to the commercial pages you want to win with, so a popular guide should link down to the category it supports. It routes buyers along the decision path, so a comparison guide links to the products it compares and the category links to the guide that reduces hesitation. And it tells search engines which page owns which intent, so you stop two pages quietly competing for the same query. Link with descriptive anchor text that names the destination, not "click here," and link from the body where the link is useful, not only from templated modules.
Arabic product search behaves differently
Search behaviour in Arabic-language GCC and Egypt markets is not a translated mirror of the English pattern. Buyers mix Arabic and English within the same session — searching a product category in Arabic and a brand name in Latin script, or the reverse — and colloquial dialect terms often carry more real search volume than the formal Arabic label a catalogue export uses by default. A category page that only targets the textbook Modern Standard Arabic term for a product type will miss a meaningful share of the way people actually type the query.
Build the Arabic keyword list the same way you would the English one — from real query data and the language customers use in support chats, not from a direct translation of the English category names — and make sure product titles, category copy, and breadcrumbs in the Arabic version of the store reflect that. Sizing, compatibility, and delivery language matter even more here, because Arabic-first buyers in COD markets are already weighing whether to trust an unfamiliar store, and unclear or obviously translated copy adds to that hesitation instead of resolving it. If you are building this out for a Saudi audience specifically, the mechanics of ranking Arabic commercial pages are covered in more depth on the Saudi SEO service page.
Prepare for AI-search citations
Product discovery may begin inside a generated answer. Make factual passages easy to identify:
- define the category clearly;
- state comparison criteria;
- show specifications consistently;
- explain policies without ambiguity;
- expose author or reviewer accountability for guides;
- keep availability and price claims current;
- support original recommendations with a visible method.
No markup guarantees citation. Useful information, crawlability, entity clarity, and conventional SEO remain the base. Genuine reviews and answered buyer questions are exactly the kind of first-hand, verifiable evidence an AI answer engine is more likely to lean on — another reason a manufactured rating is a bad trade even before you factor in the trust risk to shoppers. If you want to track whether your products and guides are actually surfacing in AI answers, treat it as a measured surface with dated evidence, covered in measuring AI search visibility.
Technical checklist
Before the 90-day plan below, run the store against this list. Most of it is boring, and that is the point — this is the layer that decides whether the commercial and content work above ever gets seen.
- Category and product pages return a real 200 status and are not gated behind JavaScript the crawler cannot execute.
- Canonical tags point to one URL per intent, and filtered or sorted variants of the same inventory do not compete with the canonical.
- The XML sitemap is segmented (categories, products, guides) and stays free of redirects, noindex pages, and 404s.
- Structured data on products and categories matches the visible price, availability, and review count.
- Core Web Vitals hold up on the actual product template, not just the homepage.
- Mobile rendering matches desktop content — nothing gated behind a "load more" tap that hides text from crawlers.
- Hreflang and language or region targeting are correct for every market and language pair the store serves, including Arabic.
- Out-of-stock and discontinued product URLs redirect to a live alternative instead of dead-ending.
- Log files or crawl stats show search and AI crawlers actually reaching category and product pages, not only the homepage and top navigation.
If you would rather have this audited hands-on than run it yourself, that overlaps with what a dedicated SEO agency engagement covers for a UAE-based store — but the list above is worth running internally first regardless of who does the deeper work.
A 90-day action plan
Days 1–30: fix the map
Choose priority categories, assign query owners, review indexation, and repair internal links, canonicals, duplicates, and broken pages.
Days 31–60: improve commercial pages
Rewrite priority categories and products, add missing decision information, improve images and structured data, and connect policies.
Days 61–90: build supporting demand
Publish a small set of buyer-led guides, strengthen internal links, monitor query movement, and improve pages based on real search and conversion behaviour.
Measurement that matters
Track:
- organic revenue and collected revenue where available;
- non-brand traffic to categories and products;
- conversion rate by organic landing page;
- assisted conversions from guides;
- indexed pages versus pages intended for indexation;
- out-of-stock organic landings;
- query ownership conflicts;
- cited product or guide pages in relevant AI answers.
Do not celebrate traffic that lands on unavailable products or cannot convert in the target market.
Next step
If your store has many pages but weak category ownership, duplicate URLs, or low organic conversion, request a systems diagnostic. For the broader SEO/GEO scope, see AI SEO and GEO.
Want a working ecommerce SEO checklist to run against your own category and product pages? Ask for the ecommerce SEO checklist and I will send it over.