Who is behind Primafonte

Who pays for a scanner that charges nothing?

The scan is free and complete: all 33 scored checks, with no signup and no email address. It is paid for by the implementation work bought by those who would rather we did the job for them. Nobody pays to appear, nobody pays to raise their score, and scan data is never sold. The result is kept for 24 hours against the domain, only so the work is not repeated.

Why publish the whole methodology?

A tool that diagnoses a problem and also sells the cure has an obvious conflict of interest, and the only honest way to settle it is to let anyone check how every point is worked out. So all 33 checks are documented with their threshold, their weight and their reasoning, across four layers weighted 25, 30, 30 and 15. The methodology is versioned, and old scores are never recalculated.

How is the score split up?

The index splits 100 points across four layers, and not evenly: what stops a model reading you at all weighs more than what shades how it describes you. Each layer groups several checks with their published threshold, and a website can be comfortable in one and fail another without that being a contradiction.

• —Discoverability, 25 points: whether crawlers find you and know what they may do with your content.
• —Content accessibility, 30 points: whether what a bot sees resembles what a person sees.
• —Semantic structure, 30 points: whether your content can be split and quoted without losing sense.
• —Entity authority, 15 points: whether models identify you as one specific, consistent entity.

What data are the thresholds built on?

The thresholds do not come from a hunch: they were calibrated against a sample of our own, 321 domains from Spain, the United Kingdom, the United States and the rest of Europe. Of those, 226 could be measured in full, 51 had firewalls that cut the analysis off, and 41 never answered. The sample averaged 61 out of 100, and the full study is published with its method in the open.

What do we not do?

Some decisions depend neither on the price nor on the client, and it is worth stating them before anyone asks. These four come up the most, and the answer has been the same since the first day the scanner went public. The history of each domain is kept for 24 months and then deletes itself, without anyone asking.

• —We do not sell placement: nobody can pay to appear or to change their score.
• —We do not publish rankings of other people’s websites. Diagnosing whoever asks for it is a service; publishing lists of failures is a problem.
• —We do not promise that any particular model will cite you. Nobody can guarantee that, and anyone who does is selling something else.
• —We set no cookies, neither our own nor third-party ones, and we ask for no personal data to scan.