Programmatic SEO
Programmatic SEO, built on data rather than swapped place names.
Generating thousands of pages from a template is trivial. Generating thousands of pages that each answer a real query, index, and do not drag the rest of the site down with them is the entire job. Most programmatic SEO fails at that second step, because the data behind it was never there. This page is about telling the two apart before you generate anything.
Straight answer
The technique is neutral. The data behind it decides whether it works.
Google does not penalise a page for being generated. It ignores pages that are thin and near-identical, and at scale it reads that as a quality problem across the whole domain. The difference is entirely in what each page actually carries.
Gets ignored, or worse
- A template with the town name swapped and nothing else
- Every city times category combination a database allows
- Pages built for keywords with no real searches
- Invented data manufactured to justify the URL
- No internal links, so nothing gets crawled
- Volume treated as the goal in itself
Earns its place
- Real data that genuinely differs from page to page
- Only the combinations with actual search demand
- A specific query answered completely per page
- Indexation and crawl rules decided before generating
- Internal linking that concentrates authority
- Pruning treated as most of the work
The uncomfortable part
The valuable work is subtractive: deciding which pages should not exist. A page per city times category times attribute produces a combinatorial explosion of URLs, most with no searches and nothing to say, and it spends the crawl budget that should reach the pages that matter. Anyone quoting you on page count rather than on the data behind it has skipped the only step that decides whether this works.
Scope
What a programmatic SEO build covers here.
A development project as much as an SEO one, which is why it sits with someone who does both rather than a writer or a plugin.
Data model
Auditing the source data for coverage and quality, and deciding which fields carry genuine per-page value. If the data is thin, this is where we say so, before a single page is generated.
Demand mapping
Which of the possible combinations have real search demand and distinct intent. The rest are folded together or dropped, so the site does not fill with URLs nobody searches for.
Template design
A page pattern that presents the data usefully, with enough genuine substance and structure that each page stands on its own rather than reading as a fill-in-the-blank form.
Indexation control
Deciding what is indexable, what is canonicalised, and what is kept out of the crawl entirely, so crawl budget reaches the pages that earn rather than the long tail of empty combinations.
Internal linking
A linking structure that connects the generated set to the rest of the site and to each other, so the pages are discovered and authority is concentrated rather than spread to nothing.
Monitoring
Watching indexation and performance after launch, pruning the pages that do not earn their place, because a programmatic set is maintained rather than shipped once and forgotten.
When not to
When the honest answer is not to generate pages at all.
If you have to invent the data to fill the template, the searches are not there either, and the pages will be thin because the underlying information is. A smaller set of genuinely useful hand-built pages will do more.
If the site already struggles to get its existing pages indexed, adding thousands more competes for the same crawl budget and usually makes indexation worse, not better. That gets fixed first.
And if the combinations you are excited about have no measurable demand, generating them is effort spent building pages to rank for searches nobody runs. The demand mapping is meant to catch this before anything is built, which is why it comes before the template rather than after.
Evidence
Data-driven page systems, actually shipped.
The parts that make programmatic SEO work, data modelling, templating and indexation control, are build work, and there are public examples of it on and around this site.
A faceted directory build
A custom listing and profile system for a client directory, with faceted search that generates pages on real logic rather than every combination, written up in the case studies.
Custom tools and data pipelines
The software side: data pipelines, API integrations and server-rendered pages that stay crawlable, which is the machinery a programmatic build runs on.
Programmatic pages for SaaS
Integration and comparison pages generated for software companies, where the data is real and each page answers a specific product query.
Cost
What it costs.
The same £95 blended hourly rate, since this is part SEO and part development. Scoped from the data, not the page count.
Data and demand review
From £650
An assessment of the source data and the real search demand behind the pages you want to generate, with an honest verdict on whether a programmatic build is the right move at all.
Build
Scoped
Data model, template, indexation rules and internal linking, implemented on the codebase and verified before the pages go live.
Maintained
Scoped
Post-launch monitoring and pruning, because a generated set that is never maintained slowly fills with pages that no longer earn their place.
If the data or the demand is not there, the review says so and you have spent a review fee rather than a build budget.
FAQ
Questions clients actually ask
What is programmatic SEO?
Programmatic SEO is generating a large set of pages from a data source and a single template, rather than writing each one by hand. A store creating a page per product attribute, a directory a page per city and category, a SaaS a page per integration pair: all programmatic.
Done on real data with genuine per-page value it is one of the most efficient things in search. Done as a template stuffed with swapped place names it produces thin, near-duplicate doorway pages that Google either ignores or, at scale, treats as a quality problem across the whole site. The technique is neutral. Whether it works is entirely about the data behind it.
Will Google penalise programmatic pages?
Not for being programmatic. Google does not care how a page was produced, only whether it is useful. What gets pages ignored or flagged is thinness and duplication: thousands of URLs that differ by a single word and answer nothing a searcher actually asked.
The line is genuine value per page. If each page carries data, comparisons or answers that exist nowhere else and match a real query, scale is a strength. If the only difference between two pages is the town name, you have built doorway pages, and volume makes that worse rather than better.
How many pages should we generate?
As many as you have real, distinct search demand and real, distinct data to serve it, and not one more. The failure mode is generating every combination a database allows: a page per city times category times attribute produces a combinatorial explosion of URLs, most with no searches and nothing to say, and it spends the crawl budget that should reach the pages that matter.
The work is deciding which combinations earn a page, which get folded together, and which should not exist. That pruning is most of the value, and it is the step the volume-first version skips.
Is programmatic SEO just for large sites?
No, but it needs a genuine data source. A directory with a few hundred real listings, a SaaS with a few dozen integrations, a supplier with a real part catalogue: all small in page count and all good candidates, because the data is real and each page answers a specific query.
What does not work is manufacturing the data to justify the pages. If you have to invent the rows to fill the template, the searches are not there either.
What do you actually build?
The data model, the page template, the indexation and internal-linking rules, and the monitoring, on the real codebase. This is a development project as much as an SEO one, which is why it sits with someone who does both rather than being handed to a writer or a plugin.
Where the data is thin or the demand is not there, the honest output is a smaller set of hand-built pages instead, and we will say so before anything is generated.
Keep reading
Related services
Book me
Have the data, not sure it is worth generating?
Tell us what the data is and which pages you are picturing. We will tell you whether the demand and the substance are there, or whether a smaller hand-built set would do more, before anything is generated.