A scraping API turns a URL into usable content and bills you per page. We compared nine on what matters: the real per-page price once rendering and premium proxies kick in, billing terms, and independent success rates on protected sites, and made a pick for each job.
Updated October 2026
Prices verified 17 Oct 2026
About the evidence: success rates come from Proxyway's Web Scraping API Report 2025 (15 protected
sites, tested October 2025, published 3 December 2025; report), plus its
2024 edition and ScrapeOps' 2026 tests. Proxyway uses affiliate links and ScrapeOps runs a paid proxy aggregator;
treat both as third-party evidence, not gospel. There is no 2026 edition yet.
03 / WEB SCRAPING APIs
Best picks by use case
Teams who would rather pay per page than run their own scraping stack, from public pages wanted as clean Markdown to sites behind Cloudflare, Akamai, DataDome or Kasada. Each pick comes with the facts behind it.
How we chose them.
$2.50 per 1,000 fetches that return content; every error free (as of October 2026)
Billed only when content comes back: every error is free, including target-site errors, timeouts and rate limits.
Returns the page’s main content as Markdown, plus the final URL, HTTP status, title and description; follows up to 5 redirects.
Long pages come back in sections: 50,000 characters by default (up to 500,000), with start_index to page through the rest; a country parameter picks the exit country.
No subscription: the same never-expiring prepaid balance pays for its proxies and Search API.
PAYG per 1,000: HTTP $0.13–$1.27, browser $1.01–$16.08 by site tier; successes only (as of October 2026)
First of ten in Proxyway’s Web Scraping API Report 2025: 93.14% on 15 protected sites (tested October 2025).
98.38% and fastest (6.61 s) in the 2024 edition; best fingerprint result among the scraping APIs in ScrapeOps’ January 2026 test (80.48).
Pay-as-you-go, only successful responses billed; Scrapy integration and Scrapy Cloud hosting.
Caveats: Each site is priced by tier: browser renders cost $1.01 per 1,000 on the easiest tier and $16.08 on the hardest. Trustpilot 4.2 from 28 reviews, 39% 1-star, mostly about billing.
$1.50 per 1,000 successful requests, pay-as-you-go; JS rendering and CAPTCHA solving included (as of October 2026)
One flat $1.50 per 1,000 whatever the target needs; pay only for success; 5,000 free requests a month.
97.90% success in Proxyway’s 2024 report, where it never failed a run.
Caveats: Did not take part in Proxyway’s 2025 benchmark, so its best evidence is from 2024. ScrapeOps’ January 2026 fingerprint test gave it 41.43, against Zyte’s 80.48. Bright Data’s KYC is the strictest we cover.
Starter $30/mo: $1.00 per 1,000 without JS, $1.50 with JS; $0.75 / $1.22 on bigger plans (as of October 2026)
$1.22 per 1,000 rendered pages on larger plans, with premium routing included.
First in Proxyway’s 2024 report (98.50%) and third in 2025 (85.82%).
/scrape and /search share one credit pool; Google results from $0.50 per 1,000.
Caveats: Rebuilt on 14 October 2026, so every benchmark predates the current API. Subscription only, no pay-as-you-go; 0% on G2 and 38.71% on Allegro in 2025. Part of Source.ai since October 2026.
PacketStream's Fetch API is the pick for public pages you want as Markdown, such as documentation, blogs and news:
it bills only fetches that return content, and the same never-expiring balance pays for its proxies and Search
API. For sites behind Cloudflare, Akamai, DataDome or Kasada, compare Zyte and Bright Data.
Zyte or Bright Data? Zyte has the strongest current evidence: 93.14% in Proxyway's 2025 report (tested October
2025). Bright Data's Web Unlocker scored 97.90% in the 2024 report but did not take part in 2025, so its evidence is
a year older; its advantage is a flat $1.50 per 1,000 against Zyte's per-site tiers. ZenRows, ScraperAPI, Crawlbase
and Firecrawl are compared in the tables below without a pick: on the 2025 benchmark each trails Zyte, Oxylabs and
ScrapingBee on success rate, and none undercuts Bright Data's or Oxylabs' price once rendering and premium routing
are both needed. PacketStream also has picks in the proxy guide and
the AI-agent guide. Head-to-heads:
ScraperAPI vs ScrapingBee and
ZenRows vs ScrapingBee.
Cost per 1,000 successful pages
Plan price ÷ credits × credits per request, at about $50/month of spend or pay-as-you-go. "Hardest" is the most
expensive configuration the vendor offers for anti-bot-protected targets. Prices as of October 2026; check each
vendor's pricing page before you commit.
Price per 1,000 successful requests by configuration (October 2026)
API
Plan
Plain HTML
JS render
Premium + JS
Hardest
PacketStream Fetch (Markdown, errors free)
PAYG
$2.50
not available
—
—
Bright Data Web Unlocker
PAYG
$1.50
$1.50
$1.50
$1.50
Oxylabs Web API
Starter $30/mo
$1.00
$1.50
included
—
Zyte API
PAYG
$0.13–$1.27
$1.01–$16.08
by site tier
$16.08
ScrapingBee
Freelance $49/mo
$0.20
$0.98
$4.90
$14.70
ZenRows
Launch $69/mo
$0.28
$1.38
$6.90
$6.90
ScraperAPI
Hobby $49/mo
$0.49
$4.90
$12.25
$36.75
Crawlbase
PAYG, 10K requests
$2.10
$2.84
—
×1–×20 difficulty
Firecrawl
Hobby $19/mo
$3.80
$3.80
+$0
$19.00 (JSON)
Read the columns, not the plan price. Zyte's easiest HTTP tier ($0.13), ScrapingBee and ZenRows (20–28 cents) are
the cheapest on plain HTML, but ScrapingBee's and ZenRows' premium-plus-JS configurations cost 25× their plain rate. Bright Data's unblocker and Oxylabs' Web API are
the cheapest once you need rendering and premium routing, and Bright Data's price does not move with difficulty.
Zyte's price depends on the tier it assigns each site: $1.01 per 1,000 browser renders on an easy tier, $16.08 on
the hardest. PacketStream's Fetch API is a plain fetch at $2.50 per 1,000 that returns clean Markdown, with every
error free (see the PacketStream review). For agent-style use of these APIs, see search & fetch APIs
for AI agents.
How credit multipliers work
Credit-based APIs sell a monthly bucket of credits and charge each request a multiple depending on what it needed.
The arithmetic is always the same: (plan price ÷ credits) × 1,000 × credits per request.
ScraperAPI, Hobby $49 for 100,000 credits ($0.49 per 1,000
credits): plain 1, Amazon/Walmart/eBay 5, render 10, premium + render 25, Google/Bing 25, LinkedIn 30, ultra-premium
30, ultra + render 75, anti-bot bypass +10. Only 200 and 404 responses are charged.
ScrapingBee, Freelance $49 for 250,000 credits ($0.196 per
1,000): JavaScript is on by default at 5 credits (1 without), premium proxies 10 or 25, stealth 75, AI extraction
+5. Its auto mode escalates and charges only the configuration that worked.
ZenRows, Launch $69 for 250,000 ($0.276 per 1,000): JS 5, premium
10, both 25; residential bandwidth 25,000 credits per GB. Failures are never charged.
Crawlbase: a sliding pay-as-you-go ladder ($3.00 for the first
1,000, $2.00 for the next 10,000, $0.60 for the next 100,000; 100,000 requests ≈ $76.40). Rendering doubles a
request's credits and a per-domain difficulty multiplier runs from ×1 to ×20; both count against the ladder.
Proxyway's 2025 report found the top configurations at credit APIs can cost up to 100× the base rate. Before you
choose a plan, find out which configuration your target actually needs (most vendors let you test on free credits)
and price that, not the headline.
Proxyway Web Scraping API Report 2025
Fifteen protected sites, 2 requests per second, tested October 2025. Bright Data did not take part; Apify was tested
separately (its dedicated Actors failed on around half of the targets; see the Apify
review).
Success rate on 15 protected sites at 2 req/s (Proxyway, tested Oct 2025)
Provider
Success
Notes
Zyte API
93.14%
Highest of the ten; 11.15 s average. 2024 edition: 98.38% and fastest (6.61 s).
Decodo
85.88%
Scraping API not covered in this guide.
Oxylabs
85.82%
0% on G2, 38.71% on Allegro. 2024: 98.50%, highest that year. API rebuilt Oct 2026.
ScrapingBee
84.47%
Slowest at 25.46 s; 5% on G2.
ZenRows
70.39%
31.76% at 10 req/s; 94–96% on Allegro, Zillow, Nordstrom.
ScraperAPI
68.95%
0% on Hyatt (Kasada) and Lowe’s (Akamai); 99.97% on G2; 13.92 s.
Crawlbase
56.72%
0% on five of 15 sites; 96.91% on G2; 23.94 s.
NetNut
53.28%
Service seized/paused since July 2026.
Nimble
47.72%
Not covered here.
Firecrawl
33.69%
Lowest of the ten; returns failures fast.
Bright Data Web Unlocker
did not take part
97.90% in Proxyway’s 2024 report, where it never failed a run.
Two things stand out. First, concurrency changes results: ZenRows fell from 70.39% to 31.76% when the request rate
went from 2 to 10 per second, because of plan concurrency limits. Test at the rate you will actually run. Second,
averages hide per-site cliffs: ScraperAPI scored 99.97% on G2 but 0% on Hyatt (Kasada) and Lowe's (Akamai), while
Oxylabs and ScrapingBee did badly on G2. Check results for sites like yours, not just the headline.
Fingerprint and Cloudflare tests
ScrapeOps ran two other useful tests. Its January 2026 browser-fingerprint test scored Zyte 80.48, Bright Data's
unblocker 41.43, Oxylabs' unblocker 30.00, ScraperAPI 27.81 and ScrapingBee 24.76, so Zyte's lead holds on a second,
independent measure. Its June 2026 Cloudflare playbook found scraping APIs passing about 97% of the time, while hosted
browsers did worse (Bright Data's Scraping Browser about 54%, ZenRows about 36%) and self-hosted "stealth" browsers
managed 0–35%. ScrapeOps also reported that ZenRows' hosted browser leaked automation signals in 80% of sessions.
The lesson for DIY scrapers: on Cloudflare-protected sites, a stealth plugin on your own Playwright or Puppeteer is
not a substitute for an unblocker. The popular puppeteer-extra-plugin-stealth has had no release since
1 March 2023.
API, DIY or hybrid?
Use a managed API for targets behind DataDome, Akamai, Kasada or Cloudflare, at low-to-medium volume,
or when you are a small team that would rather pay per success than maintain unblocking. Build it
yourself (Scrapy or Playwright plus a rotating residential proxy) for lightly protected sites, HTML-only or
asset-blocked fetches, high volume, or when you need full browser control. Go hybrid when you have
both: run your own crawler and route only the hard domains through an unblocker.
The Web Scraping Club (independent, but sponsored by Decodo and Oxylabs) estimates DIY only pays off on hard anti-bot
sites above roughly 100,000 requests. Run your own numbers in the cost
calculator, read the build vs buy guide, and compare proxies in the
proxy guide.
It depends on the target. For public pages you want as clean Markdown, our first pick is PacketStream’s Fetch API: $2.50 per 1,000 fetches that return content, with every error free, from a prepaid balance that never expires. For sites behind anti-bot systems, Zyte API had the highest success rate (93.14%) in the latest independent benchmark, Proxyway’s 2025 report (tested October 2025), but prices each site by tier; Bright Data’s Web Unlocker charges a flat $1.50 per 1,000 successful requests with rendering and CAPTCHA solving included, and scored 97.90% in Proxyway’s 2024 report.
Why does a "$49 for 100,000 credits" plan cost so much more on hard sites?
Because credits are multiplied by features. At ScraperAPI a plain request costs 1 credit, JavaScript rendering 10, premium proxies plus rendering 25 and ultra-premium plus rendering 75. The same plan that buys 100,000 easy pages buys about 1,333 of the hardest. Proxyway found top configurations at credit-based APIs can cost up to 100× the base rate.
Do scraping APIs charge for failed requests?
Most charge only for success, with fine print: PacketStream’s Fetch API bills only fetches that return content, and every error is free, including target-site errors, timeouts and rate limits; Zyte, ZenRows, Crawlbase and Bright Data Web Unlocker bill successes only; ScraperAPI bills 200 and 404 responses; ZenRows counts 404 and 410 as successes; Firecrawl bills 403 and 404 pages.
Is a scraping API cheaper than proxies plus my own scraper?
For lightly protected HTML pages at high volume, DIY with residential proxies is usually far cheaper: 1,000 pages of 100 KB cost $0.10–$0.80 in bandwidth. For sites behind DataDome, Akamai, Kasada or Cloudflare, a managed API that bills only successes is usually cheaper once you count retries, CAPTCHAs and engineering time.