Almost nobody fills schema in

Structured data is how an engine understands what you are, not just what you write: a garage, a restaurant, a dental practice, an online shop. Google uses it for rich results, and AI assistants use it to cite you with some idea of who you are rather than at random.

The trouble is that filling it in by hand is specialist work: you need to know schema.org exists, pick the right type out of hundreds, and get the hierarchy right. Most people leave the default, and the default says almost nothing.

It reads the site, weighs the clues, decides

The business type no longer comes from a list of words found in the text, but from signals with different weights, added up into a confidence score.

Behind it sits a register of 171 trades — from bakery to car dealership, from plumber to online shop — each with its schema.org type, its parent type, and a sensible fallback when the evidence runs short.

It also knows when it isn’t sure

Automatic detection that gets it wrong quietly is worse than no detection at all. So the confidence score is not an internal number: it decides how much the plugin allows itself to do.

And above all it shows its reasoning: the panel lists the clues it recognised with their weights — “Google Business: auto repair shop +6 · Brand Context: officina meccanica +4” — and the ones it rejected. Disagree, and you override it: your manual choice beats everything.

Three traps, three rules

The premises rule. If the detected type implies a place you can walk into — a restaurant, a shop — but the site carries no trace of an address, phone number, opening hours or Google profile, the type drops to its fallback. Declaring premises you cannot evidence is an invitation to be ignored.

What you make is not what you are. A site selling a plugin, an app or a course stays an organisation, and the product is declared separately as an offering. Before, a site like that risked describing itself as a piece of software.

One graph, and it stands aside when it should

Everything lands in a single JSON-LD block with no duplicate identifiers: site, organisation, page and content linked together, the way engines expect to find them.

And if another SEO plugin on the site already builds its own graph, AISA does not add a second one: it steps back. Two graphs contradicting each other are worse than one imperfect graph — tested with Yoast installed and then removed, with no duplicate nodes either way.

Meanwhile the SEO Doctor does not just tell you what is missing: logo, social profiles, opening hours and contact details are fixed from there, in one click, without hunting for the right page.

Since 2.0. AISA no longer steps aside: with Yoast, Rank Math or All in One SEO active, its nodes go into the SEO plugin’s graph and fill only the missing keys, with no second block and without overwriting your values.