Programmatic SEO has a reputation problem. Publish forty location pages from one template and the worry is automatic: is this the thin content Google warns about? The honest answer is that scale is not the thing that gets penalised. Thin pages are. A site can publish hundreds of pages and rank, and another can publish three and get suppressed. The line between the two is substance per page, not page count.
This guide is the quality-first method: a repeatable way to generate pages at scale where every page earns its place. It assumes you have already accepted the premise that data, not town names, is what makes a local page real. If you want that argument in full, the defensible version of programmatic SEO makes it, and the scaled content policy guide answers the policy question in detail. Here we take both as read and focus on the method you run.
The policy, in one line
Google’s spam policy targets “scaled content abuse”, which it defines as “when many pages are generated for the primary purpose of manipulating search rankings and not helping users”. One of its own examples is “using generative AI tools or other similar tools to generate many pages without adding value for users”. Read it carefully and the test is about outcome, not method. The tool you use to make a page is not what gets judged. Whether the page helps a real person is. Templated generation is fine. Templated emptiness is not.
That single line is the whole brief for the method below. Every step exists to make sure each generated page adds value for a user, so that scale never becomes the problem.
The four-step method
The method has four steps, run in order: put unique substance on each page, measure how similar the pages actually are, apply a quality gate, and publish in tranches. The first three keep individual pages honest. The fourth keeps the whole estate honest, and it is the step most programmatic projects skip.
Step one: unique substance per page
A location page needs a reason to exist beyond the town name in its title. The reason is local substance: facts that are true for this area and not the next one. Travel time from your base. Jobs you have actually done nearby. A parking or access quirk of the high street. A council scheme or a building stock that changes the work. None of this comes from a template, and that is the point. The template carries the structure; the substance carries the page.
A useful discipline is to leave a deliberate slot on every page for the part only you can write, and treat the page as unfinished until that slot holds something real. If the only difference between two pages is the town, you have one page published twice.
Step two: measure similarity
Intuition is a poor judge of how similar two pages are, because the template fools your eye. The fix is to measure it. Strip the location names first, then compare what is left, because “boiler repair in Lewes” and “boiler repair in Uckfield” are the same sentence wearing different hats. Compare each page against its closest sibling rather than the whole site, because a single near-twin is what gets a page collapsed in search.
The mechanics of this, including word shingles, Jaccard overlap and where to set the threshold, are covered in full in the guide on duplicate content across location pages. The principle for the method is simpler: you cannot manage similarity you have not measured, so measure it before you publish, not after a ranking drop tells you.
Step three: a quality gate
A quality gate is a rule you apply consistently before a page goes live: enough unique substance, enough local specificity, a genuine reason to exist. A page that fails the gate is not deleted. It is sent back for more substance. The gate flags weak pages before they reach the public, which is the entire job. It turns “we think these pages are fine” into “these pages cleared a bar we wrote down”.
The gate only works if it is a number you trust and a rule you do not quietly override. The moment you wave a thin page through because you are in a hurry, the gate has stopped doing anything. Consistency is the feature.
Step four: publish in tranches
This is the step that separates a defensible programmatic project from a risky one, and it is the one almost nobody does. Do not publish forty pages on a Tuesday. Publish five, then watch them for a few weeks. Do they get indexed? Do they hold a position, attract impressions, earn the occasional click? If five good pages behave like good pages, publish the next ten. If they sit unindexed or never gain traction, that is a signal to fix the substance before you scale the problem.
Publishing in tranches does three things at once. It gives Google a natural, earned growth curve instead of a wall of pages appearing overnight. It gives you real feedback on whether your substance is actually working before you commit the whole estate to it. And it makes any problem cheap to fix, because you find it at page five, not page forty-five. A worked example makes the rhythm clear.
A worked example
Say you cover the Sussex Weald and want pages for Uckfield, Lewes and Crowborough among a dozen towns. You start with three, one per town, because three is enough to test the substance without betting the estate on it.
The Uckfield page leads with the fact that your base is twelve minutes away and you carry stock for the older housing on the Bell Lane side, so same-day calls are realistic there. The Lewes page is honest that the one-way system and resident parking on the steep streets means you quote a fixed call-out rather than hourly, and it names two recent jobs in the conservation area where listed-building rules changed the fix. The Crowborough page notes the higher ground and harder winters, the boilers you see most on the Beacon estates, and a typical travel time that is longer than the other two. Strip the town names and these three pages do not collapse into one, because the facts underneath them genuinely differ. That is the similarity check passing for the right reason.
All three clear the quality gate, so they go live together. Three weeks later the Uckfield and Crowborough pages are indexed and picking up impressions; the Lewes page is indexed but flat. Reading that, you do not rush the next tranche. You look at Lewes, find the local section is thinner than the other two, add a genuine paragraph on the parking-and-access reality that customers there actually ask about, and only then publish the next three towns. By the time you reach town twelve, every page has been pressure-tested against the four steps, and the estate grew at a pace that looks earned because it was.
Where Townsmith fits
Townsmith builds the first three steps of this method into the generation itself, so the discipline does not depend on you remembering it. Every page is scored against its closest sibling before publishing, and the score withholds its largest credit until the page carries real local substance, so a thin page shows up as a low number rather than slipping through unnoticed. The scoring is documented so you can defend a page to a client or to yourself. The gate in the free plugin is advisory: it scores each page and flags the weak ones, and the decision to hold a page back stays yours. The fourth step, publishing in tranches, is a habit no plugin can have for you, but a quality score per page is exactly what tells you whether the last tranche earned the next one. You can see the full set of features if you want the detail.