Programmatic SEO, Goldmine or Landfill

Programmatic SEO, the practice of generating hundreds or thousands of pages from a template and a database, has minted some real winners and buried a lot of sites under their own thin content. Whether it turns into a goldmine or a landfill comes down to a few things most teams get wrong long before they write a single page. The mistake is rarely the templating itself. It is the assumption that volume is the goal, when volume is only ever a byproduct of having something real to say at scale.

When it works

It works when each page genuinely answers a distinct question and has something real behind it, data, inventory, prices, availability, the kind of thing a person actually came looking for. Travel sites, real estate, job boards, and marketplaces win here because the underlying data is the value, and the template is just a way to present it at scale. The page for flights from Chicago to Lisbon in October is useful because the prices, the airlines, and the layovers are different from the page for flights from Chicago to Tokyo. Nobody is reading those pages for the prose. They are reading them for the answer, and the answer changes from one page to the next.

That difference is the whole game. A good programmatic page is one where you could not have written it without the data, because the data is the content. Swap in a different row from the database and you get a meaningfully different page, not the same page wearing a different hat. When that condition holds, scaling is honest. You are not manufacturing pages, you are exposing information that already exists and that people are already trying to find.

When it fails

It fails when the pages are spun from nothing, the same paragraph with a city name swapped in, no real information underneath. Search engines have gotten good at spotting that, and AI answers ignore it entirely, because there is nothing worth lifting. You end up with thousands of pages that dilute the good parts of your site and earn nothing.

The damage is worse than zero. Thin pages do not just fail to rank, they drag down the pages around them. Crawlers spend their time budget wading through filler instead of finding the work that actually deserves attention. The site starts to read, in aggregate, like a content farm, and that reputation is sticky. Cleaning it up later means deleting most of what you built, which is a harder conversation than not building it in the first place.

A worked example

Picture two job boards, both generating a page for every job title in every city. Board A pulls live listings, salary ranges from real postings, the number of openings this week, and which companies are hiring. The page for warehouse jobs in Columbus is different from the page for warehouse jobs in Reno because the openings, the pay, and the employers are genuinely different. A person searching for exactly that lands on exactly that, and an AI answer summarizing local pay can quote it because there is a number to quote.

Board B generates the same grid of pages but has no live data. Each page reads, there are many great warehouse jobs in Columbus, browse openings today, with the city name dropped into a sentence that was written once and reused ten thousand times. The Columbus page and the Reno page are identical except for two words. There is nothing to rank for and nothing to cite, because the page does not actually know anything. Same template, same ambition, completely different outcome, and the only variable that changed was whether real data sat underneath.

What this means for AEO

The rise of AI answers raises the bar rather than lowering it. When a model assembles a response, it pulls the specific, verifiable detail and skips the filler, because filler adds nothing it cannot already generate itself. A page that states a concrete number, a current price, a real availability window, gives the model something it cannot invent, and that is what gets surfaced and attributed. AEO rewards the same thing good search always did, just more strictly. If your page is the only place a particular fact lives in a clean, structured form, you become the source. If your page is a rewording of common knowledge, you are competing with the model's own training, and you will lose.

This is why the programmatic sites that hold up under AI answers are the ones sitting on proprietary or freshly updated data. The template is incidental. What earns the citation is that the page knows something the model does not, and presents it clearly enough to be lifted.

The test before you build

Ask whether a single one of these pages would be worth making by hand. If the answer is no, scaling it will not fix that, it will multiply it. Programmatic only works when the data is the substance and the template is the convenience, never the other way around.

A useful second question is whether you would be comfortable showing one of these pages to the person you built it for, with no apology and no plans to improve it later. If you find yourself thinking it is fine for now, or that you will flesh it out once it ranks, you have already answered the first question and just do not want to admit it. The pages that win were worth publishing on day one. Everything else is landfill with good production values, and the search engines, the AI answers, and eventually your own analytics will all tell you the same thing.