7% of dental practice websites are broken right now
I checked 971 of them across eleven US metros. 71 were broken on the day I looked.
This number gets quoted at practices constantly and I could not find anyone who had actually counted, so I did. The method is below and every step is reproducible.
The count
| Fault | Count | What it means |
|---|---|---|
| Dead domain | 38 | No A record and no NS record. The site is gone rather than slow, usually an expired domain. The practice very often still trades. |
| Certificate fault | 28 | Chrome shows a full-page red warning before the site loads. The practice is open and still paying for hosting. |
| No working HTTPS | 5 | The server answers on plain HTTP but the secure connection fails outright. |
71 of 971, which is 7.3%. Eleven metros: Portland, Seattle, Boston, Denver, Jersey City, St. Louis, Chicago, San Diego, Philadelphia, Atlanta and Minneapolis, taken from practices that publish a website in OpenStreetMap. An earlier version of this page said 8.1% from a smaller sample. That figure took its numerator and its denominator from scans made on different days, which I could not reconcile, so I pulled it and re-ran the population and the checks in a single pass. This number comes from one pass over one list.
How each one was checked
- A DNS lookup against two independent resolvers, Cloudflare and Google, so one resolver having a bad day cannot create a false positive.
- A plain HTTP request.
- The page loaded in a real headless Chrome. Only faults Chrome itself raises are counted.
Anything that loaded, or failed ambiguously, was dropped rather than counted. That is deliberate: an undercount is a far cheaper mistake here than an overcount.
What the data does not show
I expected the broken ones to cluster on a few bad hosting platforms. They do not. Grouping the practices by the IP address they resolve to, there are 19 clusters holding five or more practices, covering 193 of them between them. Those 19 clusters contain 3 broken sites in total. Sixteen of the nineteen have none at all.
So 68 of the 71 broken sites sit on an address hosting nobody else, or have no address at all because the domain has lapsed. A broken dental website is overwhelmingly an individual practice problem rather than a platform problem. That surprised me and it is the opposite of what I set out to show.
One dental-specific website vendor does appear repeatedly in the broken set, but in this population it runs one broken site out of ten hosted, roughly the background rate. It shows up often because it is large. I went looking for a villain and the data did not provide one.
A warning if you try to reproduce this by reading certificates: you cannot get a per-vendor failure rate that way. A correctly provisioned site serves a certificate for its own domain, so a vendor’s shared certificate is visible only when provisioning has failed. Group by that and every vendor comes out at 100% broken, which is how I know the method is wrong rather than the vendors. Grouping by IP address works, because an address is there whether the site is healthy or not.
The sample is also limited to practices that publish a website in OpenStreetMap, and coverage there varies enormously by city. Jersey City returned 598 businesses and Philadelphia returned 29. So this is five well-mapped metros, not America.
Why nobody notices
The practice cannot see it. Staff open the site from a bookmark on a machine that stored a certificate exception the day it broke, so from the front desk it looks entirely normal. The warning is shown only to somebody arriving fresh, which is every prospective patient clicking a Google result. Several of the sites in this count have been broken for over a year, and one was serving a certificate that expired in 2019.
Check yours
Type your domain in here and you get what certificate it actually serves, whether a browser would refuse it, and whose job the fix is. Nothing stored, no signup. Or email the domain to ishubham2101@gmail.com and I will look at it properly. The one command that tells you whose fault it is if you would rather run it yourself. If you manage a list of practices or sites and want all of them checked, that is the thing I sell.