String is the better fit when protected pages must return consistently without choosing a stronger mode first. ScraperAPI is the better fit for high-volume, cooperative pages where its one-credit request stays one credit. In the September 16, 2026 Web Data Frontier Benchmark, String returned verified content on 97.0% of 500 requests. ScraperAPI returned it on 84.0%.
ScraperAPI still wins meaningful cases. It returned Idealista on all five attempts where String returned nothing, and posted 92.0% against Akamai, the largest anti-bot group in the suite. Its Startup plan supplies 1,000,000 credits for $149, making cooperative pages cheaper at volume than String's subscription-plus-usage price. The choice depends on the domains and ScraperAPI modes the workload actually needs.
Last updated September 22, 2026. Benchmark figures come from the September 16, 2026 run; see /benchmark. ScraperAPI product, pricing, credit, parameter, failure-billing, cost-control, and MCP claims were checked against its published pricing page, credit documentation, parameter documentation, status-code documentation, and MCP documentation on August 31, 2026.
| String | ScraperAPI | |
|---|---|---|
| Product shape | One fetch API, plus search, sitemap crawling, remote Chrome, and MCP | Scraping API with proxy rotation, retries, rendering, geotargeting, data pipelines, a crawler, and MCP |
| Benchmark path | String selects the request or browser path | Its strongest published configuration as of the September run |
| Benchmark success | 97.0% (485/500) | 84.0% (420/500) |
| Failure-aware latency score | 7.06s, first of 16 | 12.97s, third of 16 |
| Sites passed on no attempts | 2 of 100 | 9 of 100 |
| Entry offer | 5,000 free requests; Starter $20 plus usage | 1,000 free monthly credits; 7-day trial with 5,000 |
| Roughly $100 tier | Growth $100 plus $0.20 to $4.00 per 1,000 verified requests | No $100 plan; Hobby $49 for 100,000 credits, Startup $149 for 1,000,000 |
| Cost control | Published rates by path; blocked requests free | max_cost limit and domain-cost estimator |
| Best fit | Protected pages that must return consistently | Cooperative, predictable pages at high volume |
The Web Data Frontier Benchmark sends every provider the same 100 URLs five times. A response passes only when it contains a marker from the real page. A CAPTCHA or block page returning HTTP 200 still fails. The September run made 8,000 live requests across 16 providers.
String built the benchmark and sells the product that ranked first. The test is public so that conflict is inspectable: the target list and pass criteria, ScraperAPI adapter, and complete results JSON are available for review.
| Provider | Successful requests | Success rate | Latency score |
|---|---|---|---|
| String | 485/500 | 97.0% | 7.06s |
| ScraperAPI | 420/500 | 84.0% | 12.97s |
The latency score is failure-aware. For each target it takes the p75 of successful attempts; a provider that never returns the page inherits the p75 of providers that did. That prevents a fast block page from looking like good performance.
The August run sent ScraperAPI a plain URL and API key, with no premium, render or ultra_premium flag. That was its default one-credit request rather than its strongest published configuration, and it scored 64.2%. The September run corrected the adapter, and ScraperAPI scored 84.0%, a gain of 19.8 points and the second-largest move in the field.
Read the earlier figure as a measurement of our adapter and this one as a measurement of the product. The stronger paths carry their own economics, which the pricing section below sets out: ScraperAPI's parameter reference charges 10 credits for premium proxies or rendering, 25 for premium plus rendering, 30 for ultra-premium and 75 for ultra-premium plus rendering.
String makes a different product choice: one endpoint selects the request, browser, standard-proxy, or premium-proxy path automatically and bills the path used.
ScraperAPI returned nothing on all five attempts for 9 sites: Temu, Realtor.com, AutoZone, G2, Allegro, Instagram, Facebook, X, and GoodRx.
String returned content on 8 of those 9. It passed all five attempts on 7 and four of five on Allegro. Temu shut out every provider in the run.
ScraperAPI won one target outright. It cleared Idealista on all five attempts where String returned nothing, which is one of only two sites String failed entirely. On Coupang and Skyscanner, where the August run split the two products, they are now level: both returned Coupang four times out of five and Skyscanner five out of five.
| Anti-bot system | String | ScraperAPI |
|---|---|---|
| Cloudflare | 100% | 89.3% |
| Akamai | 98.0% | 92.0% |
| DataDome | 92.6% | 82.1% |
| PerimeterX | 100% | 88.9% |
| Kasada | 93.3% | 66.7% |
Akamai is ScraperAPI's strongest named anti-bot category. It passed 92.0% of requests across 20 Akamai targets, second only to String's 98.0% and ahead of every other provider in the run. Its weakest is Kasada at 66.7%, which covers three sites; in August it failed all fifteen Kasada attempts, so that is a real improvement.
Across eight social platforms, ScraperAPI passed 25 of 40 attempts. LinkedIn, Reddit, TikTok, YouTube, and Pinterest were perfect; Instagram, Facebook, and X returned nothing. String passed all 40.
ScraperAPI wraps proxy rotation, CAPTCHA and ban detection, retries, geotargeting, rendering, and domain-specific logic behind an HTTP API and proxy endpoint. Its current plans include full crawler access, and its DataPipeline schedules and exports larger jobs.
The knobs are part of the product. ScraperAPI lets a buyer choose rendering, premium or ultra-premium routing, sessions, country, device, and a per-request cost ceiling. The max_cost parameter rejects a request before it exceeds the specified credit limit, and the URL-cost endpoint shows a domain and configuration's expected weight.
String is narrower. Send a URL to the Web Access API and get Markdown, HTML, or JSON. Proxy rotation, fingerprinting, CAPTCHA solving, and browser escalation happen behind the endpoint. String also provides search, an asynchronous sitemap crawl, remote Chrome over CDP, and a hosted MCP server.
Both providers support agent integrations. String is in the official registry as ai.usestring/web-access. ScraperAPI publishes a hosted MCP server and documents render, premium, ultra_premium, country, and device controls in its MCP quickstart. ScraperAPI offers more explicit controls; String gives the agent fewer choices to get wrong.
ScraperAPI pricing starts at Hobby for $49 per month with 100,000 credits, Startup for $149 with 1,000,000, and Business for $299 with 3,000,000. Unused credits reset at renewal.
The per-request weight depends on the domain and parameters:
| ScraperAPI request | Credits | Cost per 1,000 on Hobby | Cost per 1,000 on Startup |
|---|---|---|---|
| Normal page | 1 | $0.49 | $0.15 |
| Amazon | 5 | $2.45 | $0.75 |
| Cloudflare, DataDome, or PerimeterX bypass | 11 | $5.39 | $1.64 |
| Google or Bing | 25 | $12.25 | $3.73 |
| 30 | $14.70 | $4.47 | |
| Ultra-premium with rendering | 75 | $36.75 | $11.18 |
The 11-credit row combines the normal one-credit request with the 10-credit bypass addition described in ScraperAPI's pricing FAQ. Forty-three of the benchmark's 100 targets sit behind Cloudflare, DataDome, or PerimeterX, so the domain multiplier is material for this workload.
ScraperAPI charges status 200 and 404 responses. A direct 500 failure after its retry window is free. A CAPTCHA response can still arrive with status 200, consume credits, and fail the benchmark because the requested content is absent. ScraperAPI's docs ask customers to report those pages so its detection database can be updated.
String charges a subscription plus requests that return verified content. The first 5,000 requests are free. Starter is $20 per month and Growth is $100. Growth rates are:
| Fetch path | Price per 1,000 verified requests |
|---|---|
| Request, standard proxy | $0.20 |
| Browser, standard proxy | $1.00 |
| Request, premium proxy | $2.00 |
| Browser, premium proxy | $4.00 |
The honest price comparison runs both ways. ScraperAPI Startup is cheaper for one-credit pages at volume, and its 11-credit protected-page rate of about $1.64 per 1,000 is below String's $2.00 premium-request usage rate before String's subscription. ScraperAPI's Google, LinkedIn, and rendered ultra-premium weights are more expensive than String's highest Growth rate. String rolls credits over and never bills a blocked request; ScraperAPI credits reset monthly and a 200 block page may bill.
Do not divide String's success-billed rate by benchmark success. A blocked request costs String nothing. ScraperAPI's cost per usable page does depend on whether an unusable body came back with billable status 200, which the published benchmark results do not classify by response status. The broader best web scraping APIs, 16 compared keeps known billing facts separate from unknowns.
Keep ScraperAPI for cooperative, one-credit pages and send the domains that require repeated flags or still miss through String. The reverse can work too: use String as the protected-page path and ScraperAPI's crawler or DataPipeline for scheduled bulk work.
Both products offer a free evaluation. Test the default and stronger ScraperAPI modes against the same representative URLs before changing production traffic. Use max_cost so the test cannot silently move from one credit to 75.
If ScraperAPI is one option among several, the ScraperAPI alternatives comparison places five substitutes next to the same benchmark.
On the September 2026 Web Data Frontier Benchmark, yes, but by less than the August run suggested. String returned verified content on 97.0% of 500 requests and ScraperAPI on 84.0%, with 2 total shutouts against 9. ScraperAPI is strong against Akamai at 92.0% and beat String on Idealista, which is one of only two sites String failed entirely.
Not always. ScraperAPI Startup's one-credit pages cost about $0.15 per 1,000 when its allowance is fully used, below String Growth's $0.20 usage rate and $100 subscription. ScraperAPI's cost rises with domain and mode: about $1.64 for an 11-credit page, $4.47 for LinkedIn, and $11.18 for ultra-premium rendering. String Growth charges $0.20 to $4.00 per 1,000 verified requests and rolls credits over.
It charges responses with status 200 or 404 and does not charge a 500 failure after its retry window. A CAPTCHA or block page can return 200 and consume credits, which is why the benchmark validates page content rather than status alone. String bills only when verified content returns.
No. The public adapter sends only the URL and API key. ScraperAPI's stronger modes cost 10 to 75 credits and may clear pages the one-credit default missed. Test them separately and compare their different unit costs.
String on the failure-aware benchmark score: 7.06 seconds against ScraperAPI's 12.97. On the 68 sites both products returned on all five attempts, the median per-site time was 2.50 seconds for String and 4.80 seconds for ScraperAPI, and ScraperAPI took twice as long or worse on 33 of them. The failure-aware score includes the cost of targets a provider fails and should not be read as a raw response-time promise for every successful page.
Yes. String publishes ai.usestring/web-access in the official MCP registry. ScraperAPI publishes a hosted MCP integration whose scrape tool exposes its premium, rendering, geotargeting, and device controls.
Stay when most targets are cooperative one-credit pages, you use the monthly allowance fully, or Akamai-heavy coverage and explicit request controls matter more than the overall pass rate. ScraperAPI also returned Idealista on all five attempts where String returned none.
Yes. The benchmark repository, ScraperAPI adapter, and August results file are public. A full run makes thousands of billable requests, so begin with a subset.
