Home / Success rates
Scraping API success rates, as each vendor states them
17 of the 75 APIs whose pages we read publish a success rate, from 81% to 100%, and 21 publish an uptime figure. Each counts success its own way and on its own choice of sites; none of these figures is a measurement by this site, so test any API on your own targets before you buy.
Quoted from each vendor's own pages on Oct 11, 2026; hover or tap a figure to see the words it comes from. "Not listed" means the pages we read state no figure.
How to read a vendor's success rate
- Success on which sites? An average over every request a vendor serves is mostly easy pages. A figure for protected sites, or for one site such as Amazon, is the one that predicts your bill.
- Success counted how? Some vendors count a page that came back with any status as a success, some only a page with the content you asked for. The same API can claim 99% one way and much less the other.
- What a failure costs you. A vendor that does not bill failures makes a low success rate cost time rather than money. The failed requests tool shows what 80%, 90% or 95% success does to each bill.
Web scraping APIs
AI extraction APIs
| API | Success rate | Speed | Uptime | Proxy pool | Coverage | Support |
|---|---|---|---|---|---|---|
| AgentQL | Not listed | Not listed | Not listed | Not listed | Not listed | 24/7 premium support on Enterprise |
| Browse AI | Not listed | Not listed | Not listed | Not listed | Up to 500,000 pages at the same time; 250+ prebuilt robots maintained by the vendor | Not listed |
| Diffbot | Not listed | Web Search API responds in under 300ms P90 latency | Not listed | Not listed | Not listed | Not listed |
| Jina AI Reader and Search Foundation APIs | Not listed | Average latency 7.9s for Reader (r.jina.ai); Average latency 2.5s for search (s.jina.ai) | Not listed | Not listed | Not listed | Not listed |
| Kadoa | 99+% run success rate, trailing 365 days | Under 2 hours from request to first dataset | Not listed | Not listed | 30,000+ production pipelines running on Kadoa | 85% of incidents resolved without a ticket; Enterprise plan: enterprise SLA and a dedicated account manager (no figures published) |
Cloud browsers
Web search APIs
SERP APIs
Why this site does not publish its own benchmark
A fair test means the same sites, the same number of attempts and the same definition of success for every API, repeated over weeks; a single run measures one afternoon. Until this site runs one, it shows only what each vendor says, with the date and the page, so you can hold the vendor to it. The most useful test is still your own: most APIs on this site have a free tier, and a free allowance of a few hundred pages is enough to try your own targets.