Skip to content
Umbra SystemsNetwork

Dental practices, United States, 2026

84,023 domains, split by how many findings each carries

Most dental practices have nothing wrong with their web presence

76.6% of 84,023 measured domains carry nothing we would state as fact. This page is the rest, and the denominator under every figure in it.

Exact domain, no guessing. If it is not one of the 84,023 measured, the next page says so and stops.

Corpus generated 2026-08-09 from 11 collection files. No practice, domain or clinician is named on this page.

01 The corpus

19,652
carry at least one finding we would state as fact
of 84,023 domains measured, which is 23.39%

One domain is one row. A practice can hold more than one domain and one record is not one company; nothing here counts companies. Of 84,145 lines read, 122 could not be parsed and are excluded from every denominator rather than counted as a pass.

24
checks per domain
question bank v1
312,522
readings we could not make
counted, never read as a No
84,023
judgement files read back
one file per domain

02 Where findings sit

Findings arrive on three separate axes, and the third one is very small

Technical axis

13.63% carry a technical finding
11,450 of 84,023

The domain, the mail path and the home page. 9,089 of them have their only finding here.

Google axis

12.19% carry a storefront finding
10,244 of 84,023

The card a patient sees before they ever reach a website. 7,912 have their only finding here.

Complaint record

289 have nothing else but this
of 320 matched records

This axis has no publishable rate at all, and the reason is printed with the check itself rather than left out.

Two numbers circulate here and they are not the same number. The union of the technical and Google axes alone is 19,363. Adding the 289 whose only finding is a complaint record gives 19,652. We print both rather than the flattering one, and the 289 is counted rather than subtracted, so the sentence cannot quietly start meaning a different group of people.

03 Every check, drawn

Every bar is the same length, because every bar is the whole corpus

24 checks, one bar each. The solid block at the left end is what we would state as fact. Everything to the right of it is the part of the corpus that check could not or did not say anything about, and it is drawn rather than dropped. A short bar would mean a missing denominator, and there are none.

The same ledger as a table, with every count spelled out
CheckFoundCould judgeRateCould not judgeNot applicable
Home page opens1,42777,9441.83%6,0790
Home page status code75579,0850.95%4,9380
https enforced5,44775,8617.18%8,1620
Domain expiry date2,84282,5173.44%1,5060
SPF record count86983,9691.03%540
Sending IP on a blocklist37365,1880.57%018,835
Printed address can receive mail15971,2890.22%12,7340
Claim prompt on Google10,24483,92012.21%1030
Unanswered complaints on BBB320320no base83,7030

One row reads no base rather than a rate. A complaint record only exists once a complaint has been filed, so that check never produces a passing case and its denominator is its own numerator. The percentage would read as one hundred, and one hundred percent of a population defined by the thing being counted is not a measurement.

An earlier version of the method page claimed no practice ever left the denominator, while the arithmetic on that same page was short by twelve thousand. The bars above are drawn so that sentence cannot be written again without looking wrong.

Softer readings, context only

Nothing in a reading is ever built on these.

Two checks that measure us

Analytics tag on the home page: our fingerprint table did not match on 17,389 of 71,389 locations. That is a hole in our table, so the check reports 0 findings and will keep reporting 0 until the table is better. No figure from either one is published as a fact about anybody.

Online booking entry point: our fingerprint table did not match on 62,685 of 71,389 locations. That is a hole in our table, so the check reports 0 findings and will keep reporting 0 until the table is better. No figure from either one is published as a fact about anybody.

04 Cut by specialty

The specialties patients pick for themselves keep the fullest storefronts

Share of each Google category whose profile carries a description. Category is the primary category on the Google profile, not the taxonomy code on the federal registry; the two disagree about how many practices are in each specialty, so the one being used has to be named. The four rows drawn heavy are the four named below.

  1. Dental implants provider91.1%316 / 347
  2. Cosmetic dentist83.8%1,557 / 1,857
  3. Oral and maxillofacial surgeon80.9%420 / 519
  4. Periodontist80.1%450 / 562
  5. Pediatric dentist78.2%3,175 / 4,061
  6. Orthodontist75.9%4,272 / 5,625
  7. Dentist66.8%39,890 / 59,724
  8. Dental clinic66.4%2,843 / 4,284
  9. Oral surgeon63.0%1,258 / 1,997
  10. Dental implants periodontist59.5%718 / 1,207
  11. Endodontist58.2%1,170 / 2,009
  12. Prosthodontist57.0%270 / 474

Orthodontists write a description 75.9% of the time. Endodontists do it 58.2% of the time. That is 17.7 points between two groups of clinicians with the same training in how to run a practice. A patient chooses their own orthodontist; an endodontist is usually chosen by the dentist down the road. This reads as an allocation of attention, not as a failure.

Two places the story does not hold

Periodontists (562 locations) reach 80.1% and oral and maxillofacial surgeons (519) reach 80.9%. Both are referral fed, and both sit near the top. If the explanation above were the whole explanation, they would not.

One column runs the other way entirely. On whether home page images declare their width and height, endodontists reach 64.9% and cosmetic dentists reach 42.4%. The ordering is reversed and we do not know why; we have not measured anything that would explain it, so we are not going to offer a reason.

Six checks at once, sortable, with both remainder rows
Share meeting each check, by Google category. Click a column heading to sort.
Dental implants provider34791.1% 316 / 34793.1% 322 / 34697.7% 339 / 34760.0% 171 / 28543.9% 125 / 285
Cosmetic dentist1,85783.8% 1,557 / 1,85790.3% 1,672 / 1,85198.0% 1,820 / 1,85755.6% 892 / 1,60342.4% 680 / 1,603
Oral and maxillofacial surgeon51980.9% 420 / 51978.5% 405 / 51692.5% 480 / 51941.1% 179 / 43660.3% 263 / 436
Periodontist56280.1% 450 / 56280.5% 451 / 56096.3% 541 / 56243.7% 215 / 49255.5% 273 / 492
Pediatric dentist4,06178.2% 3,175 / 4,06190.2% 3,644 / 4,04194.2% 3,827 / 4,06144.5% 1,492 / 3,35043.9% 1,471 / 3,350
Orthodontist5,62575.9% 4,272 / 5,62583.7% 4,658 / 5,56590.8% 5,108 / 5,62543.3% 2,162 / 4,99849.8% 2,489 / 4,998
Dentist59,72466.8% 39,890 / 59,72475.9% 44,880 / 59,10290.9% 54,261 / 59,72450.4% 25,567 / 50,72846.0% 23,318 / 50,728
Dental clinic4,28466.4% 2,843 / 4,28480.1% 3,390 / 4,23292.8% 3,974 / 4,28437.9% 1,332 / 3,51349.6% 1,744 / 3,513
Oral surgeon1,99763.0% 1,258 / 1,99771.7% 1,406 / 1,96188.0% 1,757 / 1,99739.9% 679 / 1,70258.8% 1,000 / 1,702
Dental implants periodontist1,20759.5% 718 / 1,20770.8% 840 / 1,18790.3% 1,090 / 1,20739.0% 405 / 1,03852.8% 548 / 1,038
Endodontist2,00958.2% 1,170 / 2,00968.5% 1,354 / 1,97689.4% 1,796 / 2,00926.5% 458 / 1,72664.9% 1,120 / 1,726
Prosthodontist47457.0% 270 / 47461.1% 287 / 47087.1% 413 / 47441.6% 169 / 40651.7% 210 / 406
Other category, 114 distinct values1,21367.0% 813 / 1,21375.4% 890 / 1,18082.9% 1,005 / 1,21327.0% 269 / 99850.9% 508 / 998
No category read on this pass1440.0% 0 / 410.0% 0 / 377.3% 3 / 4131.6% 36 / 11455.3% 63 / 114

Each cell reads pass over base. The base is the locations where the check could be run; a location the check could not run on leaves the denominator and is never counted as a failure, which is why the bases differ from the location count. The last two rows are not specialties and do not take part in the ranking: together they hold 1,357 locations, and every row above plus those two sums to 84,023.

05 What ordinary looks like

Six readings, spread from the tenth to the ninetieth percentile

The middle half is shaded and the median is marked. Hover any mark to read it. Each band carries its own record count, because the checks do not all apply to the same locations. Home page load time is deliberately absent and is not coming back: twelve sampled hosts were re-fetched by hand and all twelve read more than three times high. That was our own request queue, not their server.

06 What this will not do

Six things this page refuses, each one having cost us a figure

  • It names nobody. There is not one practice, one domain or one clinician on this page, and there will not be.
  • It ranks nobody. No worst list, no league table, no score. A score would need a weighting we cannot defend.
  • It does not treat a refusal as a fault. A 429 or a 403 means our request was declined. It says nothing about whether the site works for a patient, and those responses are excluded rather than counted against anyone.
  • It does not turn silence into absence. Missing from the copy we fetched is not the same as missing. Those 312,522 readings are counted in the open and phrased as questions, never as claims.
  • It withdraws figures it cannot reproduce. A rate of 95.5% for enforced https was published once and could not be reproduced under any reading; it is withdrawn. A storefront rate of 39.00% mixed soft readings into a hard count and is withdrawn with it.
  • It does not sell an opinion as a measurement. Nothing here says a practice should change anything.

Alex Zhang; Umbra Systems. The figures on this page come from one ledger, each with the file and the denominator it was computed from recorded beside it. A figure with no ledger entry is not printed. Snapshot exported 2026-08-09.