[ Methodology ]

A score you canrecompute yourself

Six criteria, public weights, one line of reasoning per rating. The number on every review is arithmetic on the ratings below.

The rubric, 100 points
CriterionWeightThe question each rating answers
Price25What do 1,000 solves of the common challenge types cost, as published on the vendor's own price list?
Coverage20Which challenge types does the vendor list as supported, and are the current ones (Turnstile, reCAPTCHA Enterprise, Arkose, GeeTest v4) among them?
Stated speed15Does the vendor publish a solving time per challenge type, and how does it compare to the token lifetime of that challenge?
Integration15REST API, official SDKs, browser extension, and compatibility with other vendors' request formats.
Transparency15Prices, speeds, success rates, limits and error codes published in the open, without an account.
Terms10Minimum deposit, payment methods, free trial, and what happens to unused balance.

Scale

85 to 100Excellent
70 to 84Good
55 to 69Fair
0 to 54Weak

What a review is built on

Every review on this site is built from pages the vendor publishes: the price list, the FAQ, the documentation, the homepage. We read each page, quote what it says, and record the date. The quote and the date sit in the sources list at the foot of the review, numbered, so a reader can check any figure against the page it came from.

We didn't run traffic through any vendor, and no review claims we did. A success rate on this site is always the vendor's own statement and is labelled as such. A speed or latency figure is either a figure the vendor commits to on its pricing page, a ceiling it advertises, or a live counter we read on the day, and the review says which.

How the score is computed

Six criteria, each rated 1 to 5 with one line of reasoning that cites the facts. The score out of 100 is the weighted sum of the ratings (price 25, coverage 20, stated speed 15, integration 15, transparency 15, terms 10), with 1 of 5 on every criterion scoring 0 and 5 of 5 on every criterion scoring 100. Nothing else enters the number: no editorial adjustment, no rounding toward a vendor, no bonus for being a partner. The weights are in the table above and are the same for every vendor.

Grades follow the score: 85 and above is excellent, 70 to 84 good, 55 to 69 fair, below 55 weak. A rating of 2 on stated speed, for instance, means the vendor publishes no solving time at all; a 5 means a figure is committed on every row of the price table.

What we don't do

No affiliate links. No paid placements, sponsored reviews or score changes for money. No ranking by traffic or by who asked. A vendor that publishes more gets a higher transparency rating, which is the only way to move a score from the vendor's side: publish the number.

  • No affiliate or tracking links to any vendor.
  • No payment from any vendor reviewed, in any form.
  • No estimates where a vendor publishes nothing; the absence is rated instead.

Updates and corrections

Prices in this market change without notice. Each review carries two dates: the day it was first published and the day its facts were last re-read. When a page changes, the review shows the new number and the new date, and the old figure goes. If a figure here doesn't match the vendor's page today, tell us at [email protected] with the page URL and we'll re-read it.

Run a captcha service?Get it on the record.

Send us the pages that state your prices, speeds and limits. We read them, score them on the public rubric, and publish the review with the date.