Updates / lab re-run
Bot protection on real sites, four weeks later
We loaded the same 16 real sites behind Cloudflare, DataDome, Akamai, HUMAN, Kasada and AWS WAF with the same twelve browsers, four weeks after the first run. Every site still used the same vendor; what they let through changed on most of them.
2026-11-20
In late October we loaded 16 real sites behind six bot-protection vendors once with each of twelve browsers and a returning Camoufox profile, five runs each, and published the outcome per cell. On 2026-11-20 we did it again: the same sites, the same setups, one sticky residential exit per site, one session at a time per site, nothing clicked. A day earlier we re-checked which vendor each site uses; all 16 were unchanged. These are dated measurements, not an evasion guide: a site's rules move from week to week, and the point is how much.
What stayed the same
- Headless was still the tell. Headless Chrome and headless Chromium were stopped on the protected sites of Cloudflare, DataDome, Akamai and HUMAN in every run on both dates, and refused by one of the two AWS WAF sites (case 77).
- The easy sites stayed easy. The front pages of the easy Cloudflare, DataDome and HUMAN sites let (nearly) every setup in on both dates.
- AWS WAF checked everyone and let nearly everyone through, with its automatic check clearing by itself in a median 4.5 and 5.0 s on its two sites (4.4 and 5.2 s four weeks earlier), and no CAPTCHA on either date.
- Akamai's
_abckcookie never validated on any load on either date, and HUMAN never blocked outright: every stop was its Press & Hold challenge, on both dates.
What changed
- A Kasada site stopped checking its home page. On 2026-10-30 it refused eleven of twelve browsers in every run; on 2026-11-20 every setup got the page straight away, while Kasada's script was still loading on it (case 77).
- The hardest DataDome site closed further: every setup but headful Chrome met the CAPTCHA in all five runs; one run in five had let most headful browsers in before (case 71).
- Winners swapped on Akamai's hardest site: Brave, which got in every run before, was blocked in every run; Opera stayed the one browser let in every time (case 73).
- Cloudflare's hardest site got stricter for Brave and a fresh Camoufox, while Patchright now got in every run (2 of 5 before) and a returning Camoufox profile still got every site (case 70).
- Camoufox did better on Kasada and worse on HUMAN: it cleared Kasada's check on the hard Kasada site in all five runs (once before), but got HUMAN's pages in 8 of 15 loads (13 of 15 before) (case 75).
What to take from it
- A single load is a sample, and a month-old result is history: re-check the sites you depend on, and plan for the browser that works today to stop working.
- A headful branded browser and Camoufox, ideally with a returning profile, remained the setups to try first on most sites; which one wins depends on the vendor and the site, and changes.
- Each case keeps both dated tables, so you can see every cell that moved: Cloudflare 70, DataDome 71, Akamai 73, HUMAN 75, Kasada and AWS WAF 77. The overview is /bot-protection.