Best Web Scraping APIs to Bypass PerimeterX: June 2026 Benchmark

We benchmarked 8 scraping APIs against PerimeterX. Scrapfly came out on top at 100%, while most of the field struggled. Only 1 of the 8 cleared above 90%. The rest either hovered around the 50% mark or failed outright. This benchmark covers live PerimeterX-protected sites tested in June 2026, open source on GitHub, refreshed twice a month, with no sponsors and no affiliate links.

PerimeterX (now HUMAN Security) is not like most anti-bots. It builds a separate machine learning model for each site it protects, trained on that site's own traffic patterns. That means there's no fixed signature to reverse-engineer. What works on one PerimeterX deployment may not transfer to another. The APIs that score well here do so because they get the full stack right: clean IPs, browser fingerprints, JavaScript execution, and behavioral signals all at once.

Quick verdict: the best web scraping API for PerimeterX

Scrapfly is the pick for PerimeterX. It leads the table at a 100% success rate, which makes it the default for production work. Of the 8 APIs tested, only 1 Scrapfly stayed above 90%, a sharper split than most anti-bots, and it reflects how PerimeterX's per-site ML scoring punishes anything that looks even slightly off.

#	Service	Success	Speed	Cost/1k
1 🥇	Scrapfly	100%	4.3s	$3.35
2 🥈	Scrapingant	90%	34.7s	$1.9
3 🥉	Zenrows	90%	12.7s	$6.9

The web scraping APIs for PerimeterX, ranked

These are the best scraping APIs against a PerimeterX-protected target, Walmart.com.

#	Service	Success	Speed	Cost/1k		Code
1 🥇	Scrapfly	100%	4.3s	$3.35	(237) ★ 4.9	code
2 🥈	Scrapingant	90%	34.7s	$1.9	—	code
3 🥉	Zenrows	90%	12.7s	$6.9	(103) ★ 4.8	code
4	Scrapingdog	87%	6.8s	$1.0	—	code
5	Firecrawl	85%	8.8s	$10.9	—	code
6	WebScrapingAPI	66%	12.3s	$2.71	—	code
7	Scraperapi	51%	9.2s	$2.45	(62) ★ 4.6	code

Data range Jun 05 – Jun 19

Service specifications, explained

The ranking above sorts providers on success, speed, and cost. The specs below cover the capabilities that decide how each one fits your stack, language SDKs, JavaScript rendering, custom JS execution, sessions, and geotargeting, so you can match a provider to your setup without opening every feature page.

How the PerimeterX rankings have changed over time

Every benchmark re-ranks a service against the same live PerimeterX target, so a position change reflects a real shift in success, speed, and cost rather than a one-off run. Watch how the standings have moved across past benchmarks.

The 7 web scraping APIs for PerimeterX, reviewed

1. Scrapfly: 100% success on PerimeterX

Scrapfly cleared 100% of PerimeterX requests at $3.35 per 1,000 successful requests and averaged 4.3s per request. That is an uncommon combination on PerimeterX, where clearing the per-site ML scoring usually costs latency.

PerimeterX's trust-score engine is designed to degrade over time, flagging sessions that look mechanical as they repeat. Scrapfly didn't give it that opening, requests came back clean on the first attempt, which is why its cost per successful request stays low: few failed requests mean you rarely pay twice for the same page.

Pros:

• Highest success rate in the benchmark at 100%
• Cheapest cost per successful request among reliable providers at $3.35/1k
• Fastest reliable provider at 4.3s

Cons:

• Credits do not carry over month-to-month.
• No annual plans are available.

2. Scrapingant: 90% success on PerimeterX

Scrapingant cleared 90% of PerimeterX requests at $1.90 per 1,000 successful requests and averaged 34.7s per request.

The low headline price doesn't reflect what you actually pay per usable result once most requests are discarded. Not recommended for PerimeterX targets.

Cons:

• Slow at 34.7s, better suited to batch than real-time work
• Clears only 90%, below the 90% bar and too inconsistent for production

3. Zenrows: 90% success on PerimeterX

Zenrows cleared 90% of PerimeterX requests at $6.90 per 1,000 successful requests and averaged 12.7s per request.

It works, clearing PerimeterX well above the 50% threshold, but there's no single reason to choose it over the providers above it: one is cheaper, another is faster, and the top three are all more reliable.

Cons:

• Slow at 12.7s, better suited to batch than real-time work
• Clears only 89.8%, below the 90% bar and too inconsistent for production

4. Scrapingdog: 87% success on PerimeterX

Scrapingdog cleared 87% of PerimeterX requests at $1.00 per 1,000 successful requests and averaged 6.8s per request.

Pros:

• Reasonable speed at 6.8s
• Among the cheapest cost per successful request in the test at $1.00/1k

Cons:

• Clears only 87%, below the 90% bar and too inconsistent for production

5. Firecrawl: 85% success on PerimeterX

Firecrawl cleared 85% of PerimeterX requests at $10.90 per 1,000 successful requests and averaged 8.8s per request.

What you get for the cost is clean, structured markdown rather than raw HTML. If your scraping feeds an LLM or RAG pipeline, that's a real time saving. If you just need the HTML, it's hard to justify.

Pros:

• Reasonable speed at 8.8s

Cons:

• Clears only 84.7%, below the 90% bar and too inconsistent for production

6. WebScrapingAPI: 66% success on PerimeterX

WebScrapingAPI cleared 66% of PerimeterX requests at $2.71 per 1,000 successful requests and averaged 12.3s per request.

The catch is speed, which rules out anything time-sensitive. But for overnight batch jobs or pipelines that process at their own pace the latency is irrelevant, and the cost advantage is genuine.

Pros:

• Does not charge for blocked requests.

Cons:

• Slow at 12.3s, better suited to batch than real-time work
• Clears only 66.3%, below the 90% bar and too inconsistent for production

7. Scraperapi: 51% success on PerimeterX

Scraperapi cleared 51% of PerimeterX requests at $2.45 per 1,000 successful requests and averaged 9.2s per request.

Speed is its story here. The one-in-ten failure rate means retries in production, but for pipelines where wall-clock time matters and a small retry overhead is acceptable, it's the practical pick.

Pros:

• Reasonable speed at 9.2s

Cons:

• Clears only 51.5%, below the 90% bar and too inconsistent for production

What makes PerimeterX hard to scrape

Most anti-bots have a fixed ruleset: learn the rules, build something that satisfies them, and you can scrape at scale. PerimeterX doesn't work that way. It builds a machine-learning model for each site it protects, trained on that specific site's traffic, so it learns what normal users look like on that domain, their request patterns, timing, navigation paths, and device characteristics. When something deviates from that learned baseline, it gets flagged, and because the model is trained per-site there's no single bypass that transfers reliably across deployments.

That's the deeper reason the ranking splits so sharply. It's not just about having the right proxy type or rendering JavaScript, a scraping session has to look plausible on every signal PerimeterX is watching at once.

The first layer is IP reputation. Datacenter IPs are flagged immediately. Residential and mobile addresses are the baseline, not a bonus feature.

The second is the TLS and HTTP/2 fingerprint. PerimeterX inspects cipher order, extension order, and header structure at the handshake level. A standard client like requests or axios produces a fingerprint no browser would, and that alone is enough for a block regardless of IP quality.

Then comes the JavaScript environment. Browser-side scripts check hardware concurrency, WebGL renderer, canvas behavior, and other signals that a headless client either can't produce or produces incorrectly. Failing these triggers a challenge or a block before the page content loads.

Finally there is behavioral scoring. Even if a session passes the static checks, PerimeterX watches how it behaves over time: scroll patterns, click timing, navigation sequences. Sessions that don't accumulate the right behavioral history get flagged even when the initial checks passed.

Getting one layer wrong is enough. That's why so many APIs score well on simpler anti-bots but fall apart here, and why the providers that do clear PerimeterX tend to be the ones running a full real-browser stack, which costs compute and shows up in both latency and cost per successful request.

How to choose a web scraping API for PerimeterX

Decide which failure mode you can tolerate first, because every option here involves a trade-off somewhere. If you can't afford failed requests, use Scrapfly: at 100% it has the highest success rate in the test, the consistency across sessions is what keeps it there, and the cost per successful request is lower than you'd expect for an API that clears PerimeterX this reliably.

If you need results fast, Scrapfly is the fastest reliable option at 4.3s and still clears PerimeterX above 90%. Build retries into your pipeline for the small failure rate and you'll get the best speed-reliability balance in the test.

One thing to avoid: choosing on headline price. The cheapest APIs in this ranking fail the majority of PerimeterX requests, so their real cost per usable result is far higher than the rate card suggests. The only number that matters here is cost per successful request.

How we benchmark web scraping APIs against PerimeterX

Every API is tested against the same live PerimeterX-protected URLs at the same time, 1,000+ requests per provider, twice a month. We pay for the plans ourselves. No affiliate links, no sponsors, no providers with early access to results.

Cost is calculated per 1,000 successful requests using entry-plan pricing, because a cheap API with a 30% success rate costs far more per usable result than a pricier API that clears 97%. PerimeterX protects one clean benchmark target, Walmart, so the numbers reflect that anti-bot and nothing else. These results cover Jun 05 – Jun 19, 2026.

Frequently asked questions about scraping PerimeterX

Can web scraping APIs reliably bypass PerimeterX in 2026?

Some can. In this benchmark, 1 (Scrapfly) cleared above 90% and 7 cleared above 50%. The rest failed more requests than they delivered. The difference comes down to whether the API can pass PerimeterX's per-site ML scoring consistently, not just on the first request but across a full session.

What's the cheapest web scraping API that actually works on PerimeterX?

Scrapfly, at $3.35 per 1,000 successful requests with a 100% success rate, the lowest cost-per-success among providers that clear PerimeterX reliably. The cheaper options in the ranking fail too many requests to make their sticker price meaningful.

Why do results on one PerimeterX site not predict results on another?

Because PerimeterX trains a separate ML model per customer. A provider that clears PerimeterX on one site has learned to match that site's behavioral baseline; a different site has a different baseline, trained on different traffic. There's no single bypass that transfers cleanly, which is why we benchmark against a consistent target rather than averaging across uncontrolled domains.

Does bypassing PerimeterX require JavaScript rendering?

Yes, in almost every case. PerimeterX's device fingerprinting runs in the browser via JavaScript. Without a real execution environment the checks either don't run or produce signals that flag the session. The providers at the bottom of this ranking are the ones sending plain HTTP requests; the ones at the top all operate real browser environments.

What is the difference between PerimeterX and HUMAN Security?

PerimeterX rebranded as HUMAN Security in 2024. The platform itself (Bot Defender, the trust-score engine, the per-site ML models) is unchanged. Either name may appear on block pages depending on which version of the product the site deployed.

How often is this benchmark updated?

Twice a month against the same live targets, 1,000+ requests per API each run. Results are published after each run with no adjustments. The code is on GitHub if you want to verify any figure.

Conclusion

PerimeterX produces a cleaner split than most anti-bots: you either get into the reliable tier above 90% or you don't, and the providers that don't tend to fail badly rather than just inconsistently. That makes the choice simpler, with Scrapfly the API to beat at 100% and no other provider keeping pace above 90%.

The per-site ML models also mean these numbers are specific to the benchmark target. If you're scraping a different PerimeterX-protected site, run a small test before committing at scale. This page, refreshed twice a month, is the place to check the latest numbers first.

Other anti-bots: Cloudflare · DataDome · Kasada · Imperva  ·  PerimeterX targets: Walmart  ·  Hub: All anti-bot benchmarks