How Lombard small businesses show up online
6 cities, 8 trades and one batch of measurements on 1 and 2 November 2026. We publish totals only and name no business.
What we measured and on whom
We took the businesses that PagineGialle lists first in Milano, Brescia, Bergamo, Como, Varese and Pavia, in 8 trades: hairdressers, restaurants, dentists, heating engineers, garages, estate agents, jewellers and builders. We read the listings on 1 November 2026, on Tuttocittà , which serves the same records.
PagineGialle shows paying advertisers first. So this is a sample of small businesses that pay PagineGialle to be listed first, not of every small business in Lombardy.
Minafy, our audit engine, measured all 50 sites in one batch on 1 and 2 November. Every figure here is a count out of those 50, or out of the sites a check could read, and the spreadsheet holds all of them.
Email is where most of them fall short
Email authentication is 3 records in a domain's settings, SPF, DKIM and DMARC, that let a receiving server check that mail using the business's address really came from it. Without them, anybody can send mail that looks like the business's own.
1 of 50 sites has it in full: an SPF record, DMARC set to quarantine or reject rather than only to report, and a DKIM key if the check found one. 43 have part of it, and 6 have none of the 3.
HTTPS is nearly universal and security headers are not
48 of 50 sites serve a valid HTTPS certificate, and 2 don't serve HTTPS at all. Of the 49 where we could test it, 44 send a visitor who types http:// on to https://.
Security headers are instructions a server sends with every page, telling the browser what it may load and when to refuse. Minafy looks for 6 of them: 3 of 50 sites send all 6 and 20 send none.
A phone score is a single reading
Google's Lighthouse scores a page from 0 to 100 on a simulated phone. 9 sites scored 0 to 49, 37 scored 50 to 89 and 3 scored 90 to 100. For 1 it returned no score.
It is one lab run per site, and a score moves between runs, so the bands say more than any single number.
| Score | Sites |
|---|---|
| 0 to 49 | 9 |
| 50 to 89 | 37 |
| 90 to 100 | 3 |
| No score returned | 1 |
Google publishes speed measured on real visitors only for sites with enough of them. It had that data for 14 of the 50, and 7 of those were fast on every measure.
Most pages have the basics and few have all of them
48 sites have a title on every page we read, and 35 have no 2 pages with the same one. A meta description, the line under the title in search results, is on every page of 25 sites and on none of 7.
Alt text, which describes a picture to someone who can't see it, is on every image of 4 sites. On the 4 pages the accessibility check loaded, 50 of 50 sites had at least one issue rated moderate or worse.
What search engines and assistants can read
38 of the 50 carry structured data, code that tells search engines and AI assistants what a page is. Of the 35 where it could be read, 33 describe the business itself.
1 site links that data to 4 or more of its profiles elsewhere, the links that tell a machine which Google listing and which social pages are the same business. 41 link none.
17 sites show a price with a currency sign in the first 6,000 characters of a page's text, which is where the check reads. None of the 50 shuts out GPTBot, ClaudeBot, PerplexityBot or the 3 other crawlers the check names in its robots.txt.
Every count fits in one table
| Check | Fully | Partly | Not at all | Not assessed |
|---|---|---|---|---|
| Security and email | ||||
| A valid HTTPS certificate | 48 | 0 | 2 | 0 |
| http:// redirects to https:// | 44 | 0 | 5 | 1 |
| All 6 security headers | 3 | 27 | 20 | 0 |
| Email authentication in full | 1 | 43 | 6 | 0 |
| Speed | ||||
| Desktop speed: 100 out of 100 | 3 | 46 | 0 | 1 |
| Real-visitor speed: all fast | 7 | 7 | 0 | 36 |
| Homepage sent compressed | 47 | 0 | 3 | 0 |
| Every page | ||||
| A title on every page | 48 | 2 | 0 | 0 |
| No duplicate titles | 35 | 14 | 1 | 0 |
| A meta description on every page | 25 | 18 | 7 | 0 |
| An H1 heading on every page | 27 | 20 | 3 | 0 |
| Alt text on every image | 4 | 43 | 3 | 0 |
| No page under 300 words | 36 | 12 | 2 | 0 |
| No broken link found | 37 | 10 | 1 | 2 |
| A viewport tag on every page | 47 | 0 | 3 | 0 |
| No accessibility issue above minor | 0 | 14 | 36 | 0 |
| What machines read | ||||
| Any structured data | 38 | 0 | 12 | 0 |
| Organization or LocalBusiness data | 33 | 0 | 2 | 15 |
| LocalBusiness data | 25 | 0 | 3 | 22 |
| LocalBusiness: name, address, URL | 24 | 1 | 3 | 22 |
| LocalBusiness: all 6 recommended | 1 | 24 | 3 | 22 |
| 4 or more sameAs links | 1 | 8 | 41 | 0 |
| A price with a currency sign | 17 | 0 | 33 | 0 |
| The city named on the site | 49 | 0 | 1 | 0 |
| No AI crawler barred by robots.txt | 50 | 0 | 0 | 0 |
The Minafy version that ran, rubric 8 at commit 56cc3f9, scored 3 empty readings as 0, and the table counts those sites as not assessed. So 3 counts differ from Minafy's own: it put 10 sites in the lowest phone band, 1 at zero for desktop speed and 6 without a redirect. One of the 3 sites with no alt text had no image on the pages read, which Minafy scores as 0.
How the numbers were taken
For each site, Minafy loaded up to 12 pages in a desktop browser, 0.8 seconds apart, then 6 at phone size for screenshots and 4 for the accessibility check. robots.txt, read as the rules for every crawler, decided which 12 pages it read. For 2 sites the crawl read only the homepage.
It also checked up to 150 of each site's own links and 30 outside links, asking for the headers only where it could, and made a handful of single requests: the homepage, robots.txt, the sitemap, llms.txt, the headers and the http:// address. Those did not consult robots.txt. Every request carried a Chrome user agent ending in Minafy-study and our address, and Google's Lighthouse loaded each homepage twice for the speed scores.
Each check gives a site full points, part of them or none, and the table counts the sites in each. A check that got no reading counts as not assessed, never as a fail. The 6 security headers are Strict-Transport-Security, Content-Security-Policy, X-Frame-Options, X-Content-Type-Options, Referrer-Policy and Permissions-Policy.
The study gives no figure per city or per trade: with 8 or 9 sites a city and 6 or 7 a trade, the groups are too small to compare. Of the 38 readings in the spreadsheet, 4 are a model's judgement, 2 are tactics we no longer treat as levers, 3 depend on the day's search results and 3 the tool can't read on Italian sites, so none of those is published as a figure.
Send us your address and we will tell you which of these checks your site passes, in the same terms. Get a free audit.
Updated