pSEO at a small scale: when it pays, and when it's junk

Two companies do the same thing: both generate a few hundred subpages from one template. One ends up with pages people actually use, the other with junk it quietly deletes six months later. With pSEO, the generating is never what decides which way it goes. What decides it is whether there is data behind the template that nobody outside your company has.

What is pSEO?

pSEO, or programmatic SEO, means you stop writing pages one by one. You write a single template, and a database fills it in: every row becomes a page of its own. The method itself is neutral. Its value comes entirely from the data behind it, never from the page count.

The idea is old, only the name is new. A web shop's category pages, an estate agent's town-by-town listings, a job board's role pages: one structure, one table behind it, and as many pages as there are rows.

A template has two halves. One is the frame: the heading, the order of the paragraphs, the internal links. The other is the variable part, the bit that changes row by row. If that variable part is a single word, say a town name, the pages barely differ. If it is ten fields and every one of them carries weight (price, size, condition, availability, limit), each page becomes its own answer.

Here is the difference in practice. “Machine hire in Győr”, “Machine hire in Sopron”, “Machine hire in Pécs”: three pages, one sentence. Against that: “Which machine fits through a 90 cm doorway”, “What deposit does a cement mixer need”, “What a household socket will actually run”: three pages, three different sets of data, out of one template.

So the question with pSEO is never how many pages you can generate. It is how many real questions you can answer with them, and those two are rarely the same number. pSEO is a production method, not a genre: every one of those pages could be written by hand, it is just that nobody sits down to do it eighty times. The first half of the decision is therefore about substance, whether there are eighty questions worth a page each. The second half is technical, and that is the easier half.

When does it pay off for a small company?

When you hold data nobody else in your market can write down: your own price list, your own service area, your own stock, your own measurements. In that case every generated page answers a real question and the template only supplies the shape. Without data like that, pSEO is not worth doing at any size.

Start by listing what only you know. Not about the trade in general, about you:

• Your own price list or spec table. Size, output, material, pack size, deposit, minimum order. The things you say on the phone to every caller.
• Your own service area. Which towns you drive to, which ones you do not, and on what terms. That is written down nowhere else.
• Your own stock or capacity. What is on the shelf, what can be ordered, how long delivery takes, when someone is free.
• Your own measurements. Whatever you record on jobs as a matter of routine: consumption, service life, failure rate, turnaround.

If at least one of those already lives in a table, pSEO pays off even at a small scale, because the questions you answer are ones no competitor can answer. If none of them does, generating pages only magnifies what is missing.

There is a simple test for this. Write the very first page by hand, before you generate anything. If that single page stands up on its own, and the next twenty differ from it only in the data they carry, you have something to work with. If the hand-written one already feels empty, the rest will not be better.

Small scale here does not mean a thousand pages. Twelve well-chosen questions are enough, as long as each of them has a row of your own table behind it. That is why our visibility work starts with a look at the data, and the template comes after.

When does it turn into junk?

When the pages differ only in the town name printed on them: the same paragraph, a few words swapped, no new fact anywhere. A page like that tells the reader nothing, and it is exactly what Google names in its spam policy as scaled content abuse. The page count goes up, the value does not.

Junk is recognisable before it is built. One question is enough to put to the template: if you delete the variable word, is anything left that is true, specific and interesting? If nothing is left, then generating does not multiply content. It multiplies an empty sentence.

The second sign is maintenance. On four hundred generated pages, one price rise means four hundred stale numbers unless a live source sits behind them. And the visitor meets the old figure on exactly the page the search engine likes showing most.

The third sign is that your own pages start competing with each other. If five pages answer the same question in the same words, the search engine picks one, the rest stay on the site carrying no weight, and the strength of your internal links gets split between them. The fourth sign is the simplest one: show three randomly picked pages to a colleague. If they cannot say what the difference is, neither will a visitor.

With AI answers the effect is sharper still: they quote a section, and only one that contains a fact. Our piece on GEO optimisation goes into that. Whatever says nothing to anyone gets quoted by nobody.

Where does the data come from?

From where it already lives: the price list, the stock records, the delivery zones, old quotes, the service log. The bigger half of the job is not the generating, it is getting the data into shape: one table where every row means the same thing, and someone owns it. The generator knows only what that table says.

In most small companies the data exists, it just is not in one place. Part of it sits in the invoicing tool, part in a spreadsheet, the rest in somebody's head. That last part is the expensive one, because generating cannot reach it.

• One table, one meaning. One row, one page, and every column has a precise name. If the same column holds a net price on some rows and a gross price on others, that mistake ends up on every generated page.
• One owner. Someone whose job the table is, and who knows where the numbers in it come from.
• One rhythm. Weekly or monthly updates, written down. Whoever changes the price list refreshes the pages with the same motion.

This is the same work our piece on building a company knowledge base describes, except that here the result shows up immediately: what goes into the table stands on the site as a page at the next run. That is what makes pSEO sustainable instead of a one-off campaign.

At a small scale this is not a database project. A spreadsheet does the job, as long as it has a header row and everyone fills it in the same way. Tooling starts to matter when the number of rows grows, until then what matters is order. The riskiest column is the one two people fill in differently, because the generator believes both of them equally.

Where does Google stand on this?

Google does not ban automation: a machine-made page can still be a good page. What its spam policy names is content produced at scale for the search engine and worth nothing to a reader, called scaled content abuse. The test is not how a page was made, but what a person gets on it.

In practice that distinction is a relief. There is no need to fear that the search engine pushes a page down purely because a machine produced it. What you cannot do is cover a gap with volume: the same sentence with the town name swapped, several hundred times over, is precisely the case the policy describes.

A practical filter for every generated page: think through what the person who clicks gets. If there is a number, a condition, a limit or an availability on it that matters to them, the page stands on its own. If only the keyword repeats, the algorithm is not the problem.

A mass of pages like that does not live apart from the rest of your site either. Same domain, same judgement. Which is why playing for volume rarely beats ten pages written properly.

At our end pSEO is part of the Visibility service, and we always start with the data, not with the template. What we measure afterwards is not the page count: it is how many generated pages bring in impressions at all, and how many of those lead on to an enquiry. If most of them bring nothing months later, that is not a matter of patience, it is feedback about the table.

A missing fact means a missing block, not an invented placeholder.