cdpfleet / Bot protection / Scoreboard
Measured, dated
Which browser got in, per vendor.
Every setup against every vendor on real sites, from this lab week's measurements: one sticky residential exit per site, one load per setup, five runs. Use it to pick the browser to start with — then measure your own site, because each site tunes its vendor differently.
- 6vendors
- 13setups
- 1030measured loads
The scoreboard
| Setup | Cloudflare (2026-10-26) | DataDome (2026-10-27) | Akamai (2026-10-28) | HUMAN (2026-10-29) | Kasada (2026-10-30) | AWS WAF (2026-10-30) |
|---|---|---|---|---|---|---|
| Chrome, headful | 14/14 | 15/15 | 9/14 | 9/14 | 1/10 | 9/9 |
| Chrome, headless | 5/15 | 4/14 | 5/15 | 5/15 | 0/10 | 5/10 |
| Edge, headful | 15/15 | 11/15 | 10/15 | 7/14 | 0/10 | 10/10 |
| Brave, headful | 15/15 | 10/15 | 15/15 | 6/15 | 1/10 | 10/10 |
| Opera, headful | 15/15 | 11/15 | 15/15 | 8/15 | 0/10 | 10/10 |
| Yandex, headful | 15/15 | 6/15 | 5/15 | 13/15 | 0/10 | 10/10 |
| Whale, headful | 15/15 | 11/15 | 10/15 | 14/15 | 0/10 | 10/10 |
| Chromium, headless | 5/15 | 5/15 | 5/15 | 5/15 | 0/10 | 5/10 |
| Patchright, headful | 10/15 | 10/15 | 10/15 | 6/15 | 0/10 | 10/10 |
| Firefox, headful | 10/15 | 11/15 | 10/15 | 5/15 | 0/10 | 9/10 |
| WebKit, headful | 5/15 | 8/15 | 10/15 | 12/15 | 0/10 | 8/10 |
| Camoufox, headful | 15/15 | 7/15 | 12/15 | 13/15 | 6/10 | 10/10 |
| Camoufox, returning profile (2nd visit) | 15/15 | 7/15 | 15/15 | 1/15 | 7/7 | 9/9 |
Each cell: loads that got the real page (directly, or after an automatic check that cleared by itself) out of loads measured, pooled over the case's sites and its five runs (one per language); network errors are left out. Sites are real and unnamed; the date is when the case was measured. Results are snapshots: sites retune their vendor settings and vendors change weekly.
Where to start, per vendor
- Cloudflare — case 70, 3 sites, 2026-10-26: start with Edge, headful (15/15), Brave, headful (15/15), Opera, headful (15/15), Yandex, headful (15/15).
- DataDome — case 71, 3 sites, 2026-10-27: start with Chrome, headful (15/15).
- Akamai — case 73, 3 sites, 2026-10-28: start with Brave, headful (15/15), Opera, headful (15/15), Camoufox, returning profile (2nd visit) (15/15).
- HUMAN — case 75, 3 sites, 2026-10-29: start with Whale, headful (14/15).
- Kasada — case 77, 2 sites, 2026-10-30: start with Camoufox, returning profile (2nd visit) (7/7). Never got in: Chrome, headless, Edge, headful, Opera, headful, Yandex, headful, Whale, headful, Chromium, headless, Patchright, headful, Firefox, headful, WebKit, headful.
- AWS WAF — case 77, 2 sites, 2026-10-30: start with Edge, headful (10/10), Brave, headful (10/10), Opera, headful (10/10), Yandex, headful (10/10).
How to read it
- Per vendor, not per site: a vendor's column pools easy and hard sites, so a setup that got in everywhere but one site shows that site's refusals. The cases show each site separately.
- A returning profile is a separate row: the same Camoufox loading the site a second time with a saved profile. It helped on some vendors and hurt on others.
- On the vendors' own pages (2026-10-20 to 10-23, four setups and two networks): Cloudflare, DataDome, Akamai and HUMAN, and which vendor a site uses: case 62.
- This is measurement, not a bypass recipe: public pages only, one session at a time per site, nothing clicked and no CAPTCHA solved. When a site says no, the right move is usually to slow down, honour
Retry-Afterand ask the site (web scraping).