When the Normal Rules Don't Apply: E-Commerce, Video and Scale

Everything in the first ten articles assumes a certain kind of website: a service business with a few dozen pages, where each page is written by a person and each one matters.

Some sites aren't that. A shop with 4,000 products. A business whose customers want to watch rather than read. An app. A site generating thousands of pages from a database. In each case the fundamentals still hold, and the practical work looks different enough that following general advice will lead you wrong.

Four situations, and what changes in each.

E-commerce: your category pages are the money pages

The instinct is to optimize product pages. It's the wrong instinct.

Commercial queries like "men's trail running shoes" overwhelmingly return category pages, not individual products. Search for anything you sell and look at what ranks. It'll be a listing page nine times out of ten.

So treat categories as first-class content. Keyword-mapped like any other page. A genuinely useful intro that helps someone choose, not a wall of SEO text below the fold that nobody reads and everybody can tell was written for a robot. Crawlable subcategory links.

Then faceted navigation, which is where most e-commerce sites quietly destroy themselves. Every filter combination generates a URL. Colour, size, price, sort order, and suddenly a 400-product shop has 90,000 crawlable URLs, almost all of them near-duplicates. Google burns its crawl budget on nonsense and your actual pages get visited less.

The decision is per-facet, not global: index the facet pages with real search demand, because "red trail running shoes" is something people search, then canonicalize or noindex the rest and block the crawl traps in robots.txt.

Product pages win on uniqueness and schema. Manufacturer descriptions duplicated across every retailer that sells the item rank for nobody. Rewrite with detail only you have: how it actually fits, what it's genuinely good for, what to buy instead. Add your own photos. Collect on-page reviews, which give you unique text plus rating stars. Implement Product, Offer and AggregateRating schema completely, because that data powers rich results and the free product listings in Google Shopping.

Inventory lifecycle is the silent equity leak. This is the part nobody plans for and everybody pays for. Temporarily out of stock: keep the page live, mark it clearly, suggest alternatives. Discontinued with a successor: 301 to the successor. Discontinued forever: 410, or 301 to the closest category. Seasonal: keep the URL permanent and reuse it annually, so /black-friday/ rather than /black-friday-2025/, which throws away a year of accumulated authority every January.

And link from content to commerce. Buying guides point at categories, blog posts point at products. That's the informational-supports-transactional structure from the keyword mapping article, running at retail scale.

YouTube: the second-largest search engine, and a text corpus

YouTube is where how-to, review and demo intent goes. It holds a permanent position in Google's video carousels. And, through transcripts, it's a body of text that both Google and language models mine constantly.

That last point is why even a business with no interest in being on camera should care. A YouTube transcript is retrievable text on a domain answer engines trust absolutely. AI engines cite YouTube relentlessly.

What actually drives YouTube rankings, in practice: keyword-researched titles, using YouTube's own autocomplete as your keyword tool because it reflects what people search there rather than on Google. Descriptions where the first two lines do the selling. Tags barely matter.

Then the behavioural signals, which dominate everything else. Click-through rate on the thumbnail and title pair, which is why thumbnails deserve genuine design effort. The thumbnail is the title tag of video. And watch time, which means hooking the first fifteen seconds and then delivering without padding.

Upload real caption files rather than relying on auto-captions, which are mediocre and get indexed as-is. Write keyword-natural chapter titles, because they become "key moments" in Google's results.

On your own site: embed relevant videos on matching pages, which lifts engagement time. Add VideoObject schema. And publish the transcript as page text, which gives you relevance, accessibility and quotable material for AI engines in one move.

If video is the format your customers want but not the work you want to do, that's what the short-form video service exists for. Four a month, captioned and formatted per platform.

App stores: their own search engines, their own rules

Brief, because most readers won't need it, but the top of the market gets asked.

If you ship a mobile app, the stores are search engines with their own indexes. On the Apple App Store, the app name, subtitle and a hidden keyword field carry the indexable text; conversion runs on screenshots and preview video; ratings weigh heavily. On Google Play, the short and long descriptions are indexed, so write them keyword-naturally, and Play also weighs engagement, retention and technical quality including crash rates.

Both stores care enormously about ratings velocity and recency, so prompt at moments of genuine delight and respond to reviews. Localized listings expand reach.

The connection to everything else: external signals lift store rankings. A ranking website, press coverage, and social demand driving branded store searches all feed it. Your SEO and brand machine is also your app store substrate, and an app landing page on your own domain captures Google demand and deep-links into installs.

Programmatic SEO: powerful, and the fastest way to get flattened

Programmatic SEO means generating pages at scale from structured data. City by service matrices. "Average rent in [city]" pages. Template plus variable, multiplied out.

It works, and it's the technique behind a lot of very large sites. It's also one core update away from being classified as scaled content abuse, and the line between the two is exactly one thing: unique value per page.

Three tests before you generate anything. Does the data genuinely differ per page in a way that matters to a reader? Is there real search demand for each variant, checked as a keyword cluster rather than assumed? And are the pages meaningfully different from each other, or are they the same page with a name swapped?

Then ship a pilot batch, not the whole matrix. Watch indexation rates in Search Console. If Google declines to index the batch, that's your answer, delivered free and early. Scale only what earns indexation and clicks.

The small-business version of this is service-area pages, one page per city you serve, and the same test applies with no softening. A page for every town within 100km, generated by find-and-replace, is doorway spam whether you made twelve or twelve hundred. I said this in the local article and it's worth repeating here because the temptation scales with ambition: one real page about a city you genuinely serve beats twelve fake ones, and the twelve fake ones can cap the whole domain.

News and publisher sites, in one paragraph

If you run a content property rather than a business site: Google News and Top Stories eligibility runs on NewsArticle schema, clear bylines and dates, an established publication entity, and speed of publication, since freshness decays in hours rather than days. Build topic-page hubs for recurring stories. And if you syndicate, enforce canonical discipline in the agreement or your syndication partners will outrank you with your own work.

The thing all four have in common

None of these are exceptions to the fundamentals. Crawl, index, rank still governs. Intent still decides format. Quality still decides whether you hold a position once you get it.

What changes is the failure mode. E-commerce fails at scale through crawl waste and duplication. Video fails through treating YouTube as a hosting service rather than a search engine. Programmatic fails through generating pages that have no reason to exist.

In each case the failure is the same shape: doing more of something without asking whether each additional unit earns its place. That question is the whole discipline, and it's the one the automated SEO platforms can't ask on your behalf, which is where this series ends in two days.

Common questions

Should I optimize product pages or category pages for e-commerce SEO?

Category pages, primarily. Commercial queries like "men's trail running shoes" overwhelmingly return category pages rather than individual products, so treat categories as first-class content with keyword mapping and genuinely useful intro copy. Optimize product pages for uniqueness and complete Product, Offer and AggregateRating schema, but expect categories to carry your commercial rankings.

What should I do with out-of-stock product pages?

It depends on why. Temporarily out of stock: keep the page live, mark it clearly and suggest alternatives. Discontinued with a replacement: 301 to the successor. Discontinued permanently: return a 410, or 301 to the closest relevant category. Never mass-redirect discontinued products to the homepage, which Google treats as a soft 404.

Does YouTube help SEO?

Yes, in three ways. YouTube videos appear in Google's video carousels, embedding relevant video on a matching page raises engagement time, and transcripts are retrievable text that both Google and AI engines mine. Publishing the transcript as page text gives you relevance, accessibility and quotable material for AI citations at once.

What ranks a video on YouTube?

Behavioural signals dominate: click-through rate on the thumbnail and title pair, and watch time. Thumbnails deserve real design effort because they function as the title tag of video. Support them with keyword-researched titles using YouTube autocomplete, a description whose first two lines sell, real uploaded caption files, and keyword-natural chapter titles that become key moments in Google's results.

Is programmatic SEO safe?

Only when each generated page carries unique value, has genuine search demand of its own, and is meaningfully different from its siblings. Without those, it's the scaled content abuse pattern Google's spam policies target. Ship a pilot batch and watch indexation in Search Console before scaling. Google refusing to index the pilot is your answer.