Does a distinctiveness score demonstrate anything about a produced page
The question comes up every time a spam update rolls out, and it shows up in the same form: someone has a collection of generated pages, a tool that flags them as highly distinct from one another, and a feeling that the number is not answering the question they actually have. They are correct to be doubtful, and the reason is easier to prove than to argue about, because both halves of it can be measured on the same set of pages in about a second.
I generate location pages for one service, one page per town, from a skeleton with a lot of synonym slots. My checker says the pages hardly overlap, which everyone tells me is what you want. But the search engine’s own policy names synonymising as an example of the thing it acts against, so which one am I supposed to trust, my score or their page?
Both, because they are not measuring the same thing. The policy wording is about what a batch of pages is for and how little it adds for the reader; the method is clearly not the violation, and the same sentence says so whether a human or a machine wrote it. A similarity score, meanwhile, measures variation density over short stretches of words. A pool can move that score as far as it goes and still say one thing in many costumes, which is the pattern the policy example describes rather than the opposite of it. The first thing to establish is not how different the pages look but where their differences originated.
Why the score moves in the wrong direction when the documents carry facts
Construct the same pool twice. Once with synonyms and nothing else varying, and once with no synonyms at all but a number of real values tied per page, the sort of thing a listing would contain. The synonymised pool rates as hardly overlapping. The pool carrying values scores worse, because facts recur across pages in the same sentence patterns and short word runs persist. By the score, the pages designed to say something specific look like the questionable ones, and the pages stating every town the same thing in different adjectives look innocent. That inversion is the whole answer to the original question: the score cannot distinguish a pool that says something from one that merely looks varied, so a good number is not evidence of anything the policy weighs and a poor one is not an charge. It measures the surface, which is exactly the part a synonymiser is built to move.
There is a test that reveals where the differences came from, and it requires one pass. Replace every value that arrived from data, the town name included, with a substitute, and check the pool again. If what remains is many pages that are still distinct, the template manufactured them and the reader gets one page every time. If what remains collapses to a single page repeated, the data was carrying the difference, and the pages are worth as much as that data behind them. That second part the test cannot settle for you, and no measurement will.
The report, with both pools assessed and the file that reproduces them: is spintax spam
Run the deletion pass before the publication decision rather than after, because it answers a question the dashboard never answers: what does a reader find on the second page that they did not see on the first. If the answer is a different configuration of the same sentence, more variation will not fix it, and the number that claims otherwise is measuring the wrong surface.