How does Foil test anti-detect and automation browsers?

We run the strongest commercial anti-detect and automation browsers through the same signup flow against a site protected by Foil, then average what Foil saw per engine across many runs collected from May 25, 2026 to July 13, 2026. Each engine is scored on device-integrity issues, network reputation, and TLS authenticity.

Which anti-detect browsers evade Foil?

None. Across all 21 ranked engines, Foil detected every attempt. The leaderboard ranks engines by how few of Foil's detection signals each one managed to avoid, not by whether they got through.

What is the stealth score on the leaderboard?

The stealth score is a composite from 0 to 100 of an engine’s device integrity, network, and TLS results, weighted toward behavior (0.6 / 0.2 / 0.2). Higher is stealthier, and it is the value the leaderboard ranks on.

The stealth-browser leaderboard

The leaderboard

Rank	Engine	Type	Overall score	Network	Device Issues	TLS score	Cost / 100 runs
1	Bright Data	Scraping	89	93	14 ±6	95	$0.40
Bright Data Visit ↗ Web-data platform; cloud Scraping Browser bundled with proxies. Run spread 9–20 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network Residential proxies · 11 networks 93/100 Connection Cloud browser · CDP Profile Default Scraping Browser Pricing Scraping Browser PAYG: $8/GB × 0.05 GB. Residential proxy bundled. ≈ $0.40 / 100 runs
2	Scrapeless	Scraping	88	90	15.3 ±4	100	$0.14
Scrapeless Visit ↗ Scraping platform with a managed cloud browser. Run spread 11–19 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Residential proxies · 8 networks, 6 countries 90/100 Connection Cloud browser · CDP Profile DefaultA higher anti-detect tier wasn’t enabled. Pricing Growth $49/mo + $0.081/hr × 1.667 hr. Residential bundled. ≈ $0.14 / 100 runs
3	Browser Cash	Cloud automation	84	89	16.4 ±6	94	$0.15
Browser Cash Visit ↗ Cloud browser service. Run spread 9–21 signals · 20 runs TLS UA rotates across Chrome 148–149 on one frozen TLS stack 94/100 Network Residential proxies · 7 networks, 4 countries 89/100 Connection Cloud browser · CDP Profile Default Pricing Browser API: $0.09/hr × 1.667 hr per session. Residential bundled. ≈ $0.15 / 100 runs
4	ScrapFly	Scraping	83	95	19.9 ±3	100	$0.90
ScrapFly Visit ↗ Web scraping API with a cloud browser. Run spread 17–23 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Mobile + residential · 4 networks 95/100 Connection Cloud browser · CDP Profile Cloud Browser API · ASP + public residential pool Pricing Discovery $30/mo: $0.15/1k credits × ~30+ credits/req (browser/ASP + residential) × 200 reqs. ≈ $0.90 / 100 runs
5	Scrape.do	Scraping	81	91	18.2 ±7	89	$0.12
Scrape.do Visit ↗ Rotating-proxy web scraping API. Run spread 12–25 signals · 20 runs TLS UA rotates across Chrome 147–148 on one frozen TLS stack 89/100 Network Residential proxies · 9 networks 91/100 Connection REST scraping API Profile Default Pricing Hobby $29/mo: $0.116/1k credits × ~5 credits/req (JS+residential) × 200 reqs. ≈ $0.12 / 100 runs
6	Driver.dev	Cloud automation	81	89	19.6 ±2	95	$0.25
Driver.dev Visit ↗ Hosted & distributed browser automation. Run spread 17–21 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network Residential proxies · 7 networks 89/100 Connection Cloud browser · CDP Profile Hosted profileConsumer-distributed (residential) mode wasn’t enabled. Pricing Advanced $49/mo subscription floor + $5/GB × 0.05 GB. No per-hour meter. ≈ $0.25 / 100 runs
7	ZenRows	Scraping	80	93	19.2 ±8	94	$0.43
ZenRows Visit ↗ Web-scraping API and Scraping Browser. Run spread 14–29 signals · 20 runs TLS UA rotates across Chrome 140–146 on one frozen TLS stack 94/100 Network Residential proxies · 11 networks, 8 countries 93/100 Connection Cloud browser · CDP Profile Default Scraping Browser Pricing Developer $69/mo: ($0.09/hr × 1.667 hr) + ($5.50/GB × 0.05 GB). ≈ $0.43 / 100 runs
8	Notte	Cloud automation	78	95	22.6 ±5	100	$0.58
Notte Visit ↗ Browser-using AI agent platform. Run spread 19–28 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Mobile + residential · 10 networks 95/100 Connection Cloud browser · CDP Profile Stealth profile Pricing $20/mo + ($0.05/hr × 1.667 hr) + ($10/GB × 0.05 GB). Mobile+residential. ≈ $0.58 / 100 runs
9	BotCloud	Cloud automation	73	30	17.3 ±4	95	$1.12
BotCloud Visit ↗ Cloud browser driven over CDP. Run spread 13–21 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network Datacenter · Vultr, 7 countries 30/100 Connection Cloud browser · CDP Profile Default Pricing Starter $20/mo for 30 hr → $0.67/hr × 1.667 hr. Datacenter only. ≈ $1.12 / 100 runs
10	Browser Use	LLM agent	73	93	26.1 ±3	95	$0.35
Browser Use Visit ↗ Cloud platform for LLM browser agents. Run spread 23–28 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network Residential proxies · 14 networks (12/20 residential-proxy exits; 7 hosted/datacenter; 1 unclassified) 93/100 Connection LLM browser agent Profile Cloud default Pricing Dev $29/mo: ($0.06/hr × 1.667 hr) + ($5/GB × 0.05 GB). Excludes LLM tokens (deterministic signup ≈ 0). ≈ $0.35 / 100 runs
11	Rebrowser	Cloud automation	72	55	21.5 ±2	95	$0.33
Rebrowser Visit ↗ Undetectable cloud browsers on real Windows and mobile devices. Run spread 20–23 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network VPN/tunnel exits · 1 network (20/20 same Fast Servers exit) 55/100 Connection Cloud browser · patched Playwright CDP Profile Undetectable Cloud Browser profile 140594 Pricing Shared Windows PC: $0.20/hr × 1.667 hr. Optional proxy data is priced separately. ≈ $0.33 / 100 runs
12	Anchor	Cloud automation	70	100	28.7 ±3	95	$0.48
Anchor Visit ↗ Cloud browser for AI agents. Run spread 26–32 signals · 20 runs TLS Presents two different TLS extension profiles under one identity 95/100 Network Mobile network · T-Mobile 100/100 Connection Cloud browser · CDP Profile Default Pricing Starter $50/mo: ($0.05/hr × 1.667 hr) + ($8/GB mobile × 0.05 GB). ≈ $0.48 / 100 runs
13	Kernel	Cloud automation	62	55	26.8 ±3	88	$0.10
Kernel Visit ↗ Cloud browser infrastructure for agents. Run spread 25–30 signals · 20 runs TLS Injects a non-standard HTTP/2 setting 88/100 Network VPN/tunnel exits · 3 networks (20/20 VPN-associated; consumer ISP ASNs) 55/100 Connection Cloud browser · CDP Profile Default Pricing Hobbyist $30/mo: $0.06/hr headless × 1.667 hr. Datacenter only. ≈ $0.10 / 100 runs
14	Steel	Cloud automation	59	30	27.1 ±3	100	$0.17
Steel Visit ↗ Browser API built for AI agents. Run spread 24–30 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Datacenter · Latitude.sh 30/100 Connection Cloud browser · CDP Profile Default Pricing Starter $29/mo: $0.10/hr × 1.667 hr. Datacenter only (drops to $0.08/100 runs at Pro $499). ≈ $0.17 / 100 runs
15	Browserbase	Cloud automation	59	30	27.5 ±2	100	$0.20
Browserbase Visit ↗ Headless browser infra for agents & automation. Run spread 25–29 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Datacenter · Amazon AWS 30/100 Connection Cloud browser · CDP Profile DefaultAdvanced Stealth (paid) wasn’t enabled. Pricing Developer $20/mo: $0.12/hr overage × 1.667 hr. Datacenter only. Advanced Stealth = sales-only. ≈ $0.20 / 100 runs
16	Oxylabs	Scraping	52	30	24.9 ±10	63	$0.26
Oxylabs Visit ↗ Web Scraper API with Realtime rendering. Run spread 18–38 signals · 20 runs TLS UA claims Chrome 117 — older than its own handshake 63/100 Network Datacenter · 16/20 hosted VPN/tunnel exits, 15 ASNs 30/100 Connection REST scraping API Profile Universal source · JS render Pricing Web Scraper API Micro $49/mo: 37,692 JS-rendered results included → about $0.0013/result × 200 benchmark requests. ≈ $0.26 / 100 runs
17	Hyperbrowser	Cloud automation	51	30	31.2 ±1	88	$0.17
Hyperbrowser Visit ↗ Cloud browsers for agents & scraping. Run spread 30–32 signals · 20 runs TLS Injects a non-standard HTTP/2 setting 88/100 Network Datacenter · Google Cloud 30/100 Connection Cloud browser · CDP Profile DefaultStealth mode (paid) wasn’t enabled. Pricing Startup $30/mo: $0.10/hr × 1.667 hr. Datacenter only. Advanced Stealth gated to Scale $100/mo. ≈ $0.17 / 100 runs
18	Cloudflare	Cloud automation	50	55	35 ±1	82	$0.15
Cloudflare Visit ↗ Remote browser sessions on Cloudflare's network. Run spread 34–36 signals · 20 runs TLS User-Agent self-identifies as HeadlessChrome 82/100 Network VPN/tunnel exits · Cloudflare WARP (17/17 anonymous hosting; Spur: 17/17 VPN) 55/100 Connection Cloud browser · CDP Profile DefaultDefault image self-identified as HeadlessChrome 128. Pricing Normalized marginal rate: Workers Paid is $5/mo with 10 browser-hours included, then $0.09/hr × 1.667 hr = $0.15. The modeled first 600 60-second runs have no browser-hour usage charge; monthly hours are rounded, and average concurrency above 10 is $2/browser. ≈ $0.15 / 100 runs
19	ScrapingBee	Scraping	48	30	34 ±5	100	$0.10
ScrapingBee Visit ↗ Web-scraping API that renders pages. Run spread 32–41 signals · 20 runs TLS Authentic Chrome handshake; UA matches the wire 100/100 Network Datacenter · Datacamp, 10 countries 30/100 Connection REST scraping API Profile Default (JS render) Pricing Startup $99/mo (first tier with JS): $0.099/1k × ~5 credits/JS req × 200 reqs. Datacenter. ≈ $0.10 / 100 runs
20	Browserless	Cloud automation	41	30	36.9 ±2	82	$0.25
Browserless Visit ↗ Headless browser automation service. Run spread 35–38 signals · 20 runs TLS User-Agent self-identifies as HeadlessChrome 82/100 Network Datacenter · DigitalOcean 30/100 Connection Cloud browser · CDP Profile Default Pricing Prototyping $25/mo (annual): $0.15/hr × 1.667 hr. Datacenter only. ≈ $0.25 / 100 runs
21	Apify	Scraping	24	30	47.2 ±5	79	$1.33
Apify Visit ↗ Actor platform; ran the Playwright Scraper. Run spread 42–51 signals · 20 runs TLS UA claims Chrome 101 — older than its own handshake 79/100 Network Datacenter · HostPapa 30/100 Connection REST scraping API Profile Playwright Scraper actor Pricing Starter $29/mo: $0.20/CU × 4 CU/hr (Playwright Scraper) × 1.667 hr. Datacenter only. ≈ $1.33 / 100 runs

Figure 1: Every ranked engine, sorted by overall stealth score. Click a column header to re-sort; click any row for the per-engine breakdown.

Engines are ranked by their overall stealth score1 A 0–100 composite of device-integrity issues, network reputation, and TLS authenticity, weighted 0.6 / 0.2 / 0.2 toward behavior. Higher is stealthier; this is the value the board sorts on.: higher means fewer tells. The table is sortable, and each row expands to the full per-engine breakdown.

The full board, lowest score to highest. The leaders on the right are bunched within a few points of each other, and every engine was still detected on every run.

Overall stealth score by engine

Figure 2: Overall stealth score for all 21 ranked engines, lowest to highest. Higher is stealthier; every engine was still detected on every run.

Disagree with a ranking, or want to see a service on the board? Get in touch, we’ll run it through the harness.

Cost vs. stealth

Stealth is only half the decision; the other half is what it costs to run2 Cost is what you’d pay at each vendor’s entry paid tier to execute 100 benchmark signup runs, each modeled as a ~60-second session, ~0.5 MB of egress, and ~2 billable API calls. Per-hour, per-GB, per-credit, and per-CU pricing is normalized to that fixed workload so the column compares like-for-like. Residential and mobile engines include that bandwidth; datacenter engines don’t, and would cost more once you add a proxy.. Plotting the two against each other, up and to the right is the better corner: cheap and stealthy.

Figure 3: Cost per 100 runs against stealth score (cost axis is log-scaled, cheaper to the right). The dashed line is the Pareto frontier; no engine is both cheaper and stealthier than a point on it.

Frontier: Kernel → Scrape.do → Scrapeless → Bright Data. Engines off the frontier are dominated: you can pay less for equal stealth, or get more stealth for equal spend, by moving to a frontier point.

Where engines trip up

No single tell gives an engine away. Detection is spread across fingerprint inconsistencies, the runtime environment, automation markers, timing, and input behavior, so closing one gap just exposes the next. Here is how the signals that did fire break down by category.

Fingerprint 50%
Automation markers 22%
Environment 16%
Timing 7%
Input behavior 5%

Figure 4: Share of all detection signals fired across the test, by category.

Each category is a different question Foil asks about a session. Here is what they look at, and the kind of tell that lands an engine in each bucket.

Category	What it looks at	Example tells
Fingerprint	The hardware and rendering identity a browser presents: graphics, audio, fonts, screen, and the device it claims to be.	Values that don’t cohere with one another, or that match a known spoofing toolkit instead of a real device.
Environment	The browser’s runtime: which capabilities are present, and how the underlying engine actually behaves.	Surfaces a genuine install would expose that are missing, or behaviour that doesn’t match the browser being claimed.
Automation markers	Traces left by the protocols and drivers used to control a browser remotely.	Side effects of being driven by software that a human-operated browser never produces.
Timing	How long operations take, and the rhythm of events through a session.	Patterns that are too uniform, too fast, or sequenced in an order people don’t manage.
Input behavior	Mouse, keyboard, scroll, and touch dynamics during the flow.	Motion that’s too clean, missing the small corrections and variance of a real hand.

Figure 5: What each detection category looks at.

How we tested

We took the strongest commercial anti-detect and automation browsers and ran each one through the same task: create a new account on a site protected by Foil. Signup is the sharp end of the bot problem, the moment a fake identity is minted, so it is where detection matters most. The numbers above are what Foil saw, averaged per engine across many runs collected from May 25, 2026 to July 13, 20263 Scored end-to-end by the production API at api.usefoil.com, the same path that scores live customer traffic, not a research fork..

Every session is graded end to end by the same production API that scores live customer traffic, not a research fork. The overall stealth score is a 0 to 100 composite of three things: how many device-integrity issues a session trips, weighted most heavily (0.6); the reputation of the network it arrives on, mobile over residential over datacenter (0.2); and whether its TLS handshake is authentic and matches the User-Agent it claims on the wire (0.2). Higher means fewer tells.

Underneath that score, Foil watches dozens of independent signals across five families: fingerprint coherence, the runtime environment, automation-protocol markers, timing, and input behavior. Across the whole test, 150 distinct signals fired. No engine avoided all of them, and not one session ever scored as a genuine human.

Each engine ran at least 20 runs to average out luck; this snapshot spans May 25, 2026 to July 13, 2026 after fresh Browser Use, Kernel, Oxylabs, Rebrowser, and ScrapFly Cloud Browser API runs. We ranked only services we could drive through the complete signup flow; a handful refused the task outright and were left off the board. These tools ship updates constantly, and so do we, so positions will shift over time.

What is Foil?

Foil specializes in catching AI-driven fraud. Agents have popularized tampered browser distributions (AKA stealth browsers) because they help evade bot defenses, but the same browsers are increasingly in the hands of fraudsters. That is exactly why we run this benchmark.

Foil catches them. First it detects when someone is spoofing a device or browser. Then it builds a stable, unique fingerprint for every one of your real users. And finally it analyzes their behavior, every swipe, tap, and form fill, to seal the deal. It is the same engine that scored every run on this page.

You really have to try it yourself to believe it. Or get in touch if you have questions.