We measured 88 Italian sites. SEO holds up. GEO doesn't.

We ran 88 real Italian business sites through our analysis engine: 15 law firms, 19 physiotherapy practices, 13 beauty salons, 13 real estate agencies, 28 restaurants and trattorias. We didn't pick them to find disasters. We took them from businesses Google already shows on page one, searching for the trade together with the name of five, six or seven Italian cities, and excluding directories and portals. These are sites that work, run by businesses that pay someone to build and maintain them.

The result is the clearest thing we've ever measured: in all five sectors the SEO score holds up and the GEO score collapses, with no exception.

The numbers, before the explanations

The SEO score tells you how readable a page is for a traditional search engine. The GEO score tells you how readable it is for an assistant, meaning how much of what's written is also stated in a form a machine can cite without having to interpret it.

The column that matters is the first one: the median SEO score sits between 62 and 67 across all five sectors. Five different trades, sites built by different agencies in different years, and the same score within five points of each other. The median GEO score, on the other hand, ranges from 30 to 46, moving three times as much.

Why the SEO number is so stable

Because SEO has twenty years of tools behind it. The title tag, the meta description, heading structure, the sitemap, page speed: these are things a WordPress theme handles on its own, that a plugin checks, and that anyone who built these sites has seen at least once on a green or red screen. The result is a common floor. Nobody scores 90, and almost nobody scores 40.

GEO doesn't have that. GEO has no floor yet, because almost nobody measures it, and what it asks for isn't visible just by looking at the site. A site can be beautiful, clear and complete for a person, and still be silent to a machine.

The eight things missing in every single sector

Comparing the five groups, eight shortcomings recur across all five, regardless of trade:

• The address isn't in the structured data. It's on the contact page, where a person reads it. It's not in JSON-LD, where a machine reads it. Across the 15 law firms: fifteen out of fifteen.
• No geographic coordinates. Latitude and longitude cost two lines of code and remove any ambiguity around a street with a common name.
• No business schema. Schema.org has types built for this purpose, from LegalService to Restaurant. Declaring them means stating what business you're in instead of leaving it to be guessed.
• No identifiable author. Who wrote this page, and what qualifies them to write it.
• The city isn't in the title tag. For a business chosen based on proximity, this is the detail that decides whether you're even a possible answer.
• The city isn't in the meta description. Same reason, second place to look.
• Not enough conversational language. The pages talk about "areas of expertise" and "treatments," while people ask "can I fire an employee who's on sick leave" and "how much does laser hair removal cost."
• Few sector-specific questions spelled out explicitly. The answers are often there, scattered through the text. What's missing is the question written the way a customer would actually ask it.

None of these eight is a design, copy or server problem. They're all information the site already has and simply doesn't state.

What this means if you manage clients' sites

It means that the sites you already manage most likely carry a gap of 18 to 37 points, one the client has never seen because nobody's ever measured it for them, and one that doesn't require rebuilding the site to close.

It's an awkward thing to sell and an easy thing to deliver. Awkward because the client will tell you "but my site is fine," and they'll be right: it is, for Google. Easy because most of those eight points are half-hour-per-page fixes, not projects.

And there's an implicit deadline. When a client asks you why their competitor gets mentioned by ChatGPT and they don't, the answer "we worked on SEO" will be true and it won't be enough.

The limits of these numbers, stated before you ask

This is 88 sites, not eight thousand: from 13 to 28 per sector. These are medians, not averages, and we report them as medians on purpose, because in small samples a single extreme case can shift an average. These are Italian sites for local businesses: they say nothing about an e-commerce store or a SaaS product. And they're measured with our engine, using our criteria, which are public but remain ours.

What these numbers support is one thing, and we stand by it firmly: the gap between SEO and GEO isn't a case of a few neglected sites, it's the rule across every sector we looked at. What they don't support is a national percentage, and we won't write one.

Where to start

Take a client's site and check three things, which you can see in five minutes by looking at the source code: whether the address appears in an application/ld+json block, whether there's a schema declaring the type of business, and whether the city appears in the homepage's title tag. If all three answers are no, you're looking at the gap we measured, on a site that's doing perfectly well on Google.