Why the exit address decides your captcha rate

What Google says

Google's help page for the message Our systems have detected unusual traffic from your computer network is unusually plain about the cause. It says the message appears when a network you use, including VPN networks, seems to be sending automated traffic. It says you can get it because others on the same network, at a school or a business, are sending automated searches. And it says some VPN and tunnel services cause all traffic to be blocked because they make it impossible for Google to tell the difference between abusive and non-abusive traffic.

Read that last sentence as an engineer. The decision to challenge is made on the address, before your request is looked at. A perfect request from a bad address gets the challenge; the same request from a clean address doesn't.

Why solvers sell proxies

The captcha vendors know this, and two of the seven reviewed here say so on their own pages. 2Captcha's navigation lists a proxy product beside the solver, along with a browser API described as an anti-bot headless browser. CapMonster Cloud's homepage carries the line proxy included in the price above its table. A vendor whose revenue is solves telling you that the address matters is about as strong a signal as the market gives.

The bill, as a formula

Captcha spend is requests, times the share of requests that get challenged, times the price per solve. A solver only touches the third term. The second term is set by the address, the request pattern and the fingerprint, in roughly that order for the search and shopping sites that challenge hardest.

That's why the cheapest solver rarely produces the cheapest bill. Halving the challenge rate saves as much as halving the price, and the price side of this market is already close together: $0.60 to $0.95 per 1,000 for reCAPTCHA v2 across CapMonster Cloud, CapSolver and Anti-Captcha's low band. The rate side can move by multiples.

What moves the rate

Shared datacenter ranges carry the history of everyone who used them before you, which is what Google's page means by others on your network. Addresses that belong to a consumer ISP or that were assigned to a single customer carry only that customer's history. That's the reason residential and ISP proxy providers such as HProxy exist as a separate market from datacenter ranges, and why their per-gigabyte prices are higher: the address is the product.

Whether the trade pays is a measurement, not a rule. Run the same request set through each exit you're considering, count challenges per thousand requests, and put that number into the formula above with the solver price you'd pay. Sometimes the datacenter range plus a cheap solver wins. Often it doesn't, and the solver was treating a symptom the whole time.

• Measure challenge rate per exit before choosing a solver, not after.
• Keep the request pattern constant while you compare, or you're measuring two things.
• Re-measure when the rate moves; Google's page says blocks are network-wide and can be caused by other users of the same range.