Programmatic SEO, and the line between a family and a doorway

Every programmatic page answers a search with a part that stays the same and a part that changes. In "[product] alternatives" the product changes. In "[service] company in [city]" the city does. The dataset holds one row per page, and the template holds everything that's shared.

Heddle keeps twelve of these patterns as specs, from locations and industries to comparisons, integrations and calculators. Each lists the fields a row must have before its page may be published. Those fields are the part a template can't write, and they're where the work goes.

Google's spam policies name the failure twice. Scaled content abuse is many pages generated mainly to manipulate rankings, including pages generated with AI tools without adding value for users. Doorway abuse includes pages targeted at specific regions or cities that funnel users to one page.

Neither policy bans a page per city or per comparison. They ban pages that only differ by the swapped word. So the question to ask of any pattern is whether row forty-one has something to say that row forty doesn't.

A row missing its required material isn't published thin. heddle validate lists it as skipped. After the build, the uniqueness check compares each page with its siblings and fails one whose eight-word sentences repeat beyond the family's ceiling, 80% for most families.

That ceiling came from a real failure. In September 2026 two of techfabric.com's city pages shared 67% of their sentences, which passed review because each page read fine on its own.

Measure before writing. Score each candidate on four questions in order. Is the searcher a buyer? Can the company give them something useful? Is there distinct material for every page? Does demand measure? Then read three live results pages. If the vendor's own documentation fills page one, pick a longer phrasing. Ship one pattern, wait six weeks for it to be read, then decide whether to expand.