Every month at DriftProxy we publish the raw output of our continuous benchmark system as a public report. This is the June 2026 edition — 30 days of latency data, success rates, and IP-quality scores across every provider in our index, current as of June 3, 2026.
I write this report by hand each month rather than auto-generating it, because the numbers alone don’t tell the full story. What changed, what we investigated, and what we think is driving the patterns — that analysis is what turns interesting data into useful data.
How we measure
Every figure below comes from automated requests run continuously against live targets — Google SERP, a Cloudflare-protected e-commerce site, and mobile-specific endpoints — across each provider’s residential and mobile pools. We report median latency, 30-day variance, and success rate, alongside our own IP-cleanliness score.
What changed in the market this month
Two developments from May 2026 directly shaped this month’s data, and you need both to read the tables correctly.
The first is Cloudflare’s updated bot-management rules, deployed in the second week of May. The impact showed up almost immediately: providers with lower residential IP-cleanliness scores saw the steepest success-rate drops against Cloudflare-protected targets from around May 14. The update appears to specifically tighten detection of datacenter-origin IPs registered inside residential ranges — the “fake residential” problem we covered in our 30-day testing article — and it made the gap between clean and dirty pools far more visible.
The second is the continued surge in AI data-collection operations, which is consuming more of the available residential pool capacity. Several providers confirmed to us directly that concurrent session counts from AI training-data customers are far higher than six months ago. That pressure shows up in our data as higher latency variance during peak hours.
Google SERP latency leaderboard
Google SERP is our most important benchmark target because it maps to the largest group of proxy buyers — SEO monitoring, rank tracking, and search-results analysis.
| Provider | Avg. latency | 30-day variance | Change vs May |
|---|---|---|---|
| Bright Data | 41 ms | ±5 ms | −2 ms |
| Oxylabs | 44 ms | ±6 ms | No change |
| NodeMaven | 52 ms | ±3 ms | −1 ms |
| SOAX | 58 ms | ±9 ms | +3 ms |
| Smartproxy | 63 ms | ±11 ms | +6 ms |
| IPRoyal | 69 ms | ±10 ms | No change |
| NetNut | 74 ms | ±12 ms | +4 ms |
| Webshare | 78 ms | ±14 ms | +8 ms |
The degradation in positions four through eight traces back to the Cloudflare update indirectly: when more requests get blocked on protected targets, providers reroute traffic through their Google SERP infrastructure, increasing contention and latency there too.
NodeMaven’s ±3 ms variance is worth singling out. Most providers swing far more widely depending on time of day and pool load; NodeMaven’s ASN-filtering approach produces a noticeably more consistent profile than larger pools with more variable IP quality.
Cloudflare-protected success-rate leaderboard
This is where May’s Cloudflare update is most visible. Success rates against our Cloudflare-protected e-commerce target fell across almost every provider — the question is which fell hardest.
| Provider | Success rate | Change vs May | IP cleanliness |
|---|---|---|---|
| Bright Data | 97.8% | −0.4 pp | 96 / 100 |
| Oxylabs | 96.9% | −0.7 pp | 94 / 100 |
| NodeMaven | 95.4% | −1.1 pp | 93 / 100 |
| SOAX | 93.1% | −2.3 pp | 88 / 100 |
| NetNut | 91.7% | −2.8 pp | 85 / 100 |
| Smartproxy | 90.2% | −3.6 pp | 82 / 100 |
| IPRoyal | 88.5% | −4.4 pp | 79 / 100 |
| Proxy-Seller | 84.0% | −6.3 pp | 71 / 100 |
The pattern is clear: the providers with the highest IP-cleanliness scores showed the smallest drops. Bright Data lost less than half a point; Proxy-Seller lost more than six. IP cleanliness is the leading indicator — cleaner pools are more resilient to bot-detection improvements because their IPs genuinely look like residential traffic rather than masked datacenter infrastructure.
The cleanliness gap is widening
If you scrape Cloudflare-protected targets, a provider’s cleanliness score now predicts resilience better than its headline success rate. Pools that lean on datacenter IPs dressed up as residential are the first to crack when detection tightens.
Mobile proxy results
Mobile deserves its own section this month because the data moved in an unexpected direction. Several providers that had been underperforming on mobile quality showed measurable improvements — and the timing lined up with the AI-demand surge.
Our working theory: AI data-collection operations are heavy consumers of residential proxies but rarely use mobile, because large-scale crawling doesn’t match the behaviour of real mobile users closely enough to be worth it. So the AI surge is concentrating in residential pools, leaving mobile pools relatively less congested — and that lighter load is improving mobile performance for everyone else in those pools.
| Provider | Mobile latency | Mobile success | Trend |
|---|---|---|---|
| Bright Data | 0.7 s | 98.1% | Best overall |
| SOAX | 0.8 s | 97.4% | Strong mobile specialist |
| IPRoyal | 0.9 s | 95.8% | Improved May→June |
| Smartproxy | 1.0 s | 95.1% | Consistent performer |
| Infatica | 1.1 s | 94.3% | Improved significantly |
What we’re watching in July
Three things are on my radar going into July.
- WWDC 2026 (June 8). A major iOS announcement usually arrives with updated platform protections, and Apple has been steadily improving its detection of non-human traffic. Any acceleration of Private Cloud Compute could change how iOS-related targets respond to proxy traffic.
- Pool expansions. Several providers have privately signalled IP-pool expansion announcements for June and July. Fresh IPs usually start with high cleanliness — they haven’t been flagged yet — then decline as they accumulate history. We’ll watch whether the expansions translate into measurable gains.
- AI demand. If consumption keeps climbing at current rates, we may see residential pool quality decline industry-wide in the second half of 2026 as genuinely clean residential IPs become scarce. Providers with the best sourcing — ongoing, consent-based relationships with real residential users — will be the most resilient.
How to use this report
The most practical approach is to ignore the overall rankings and go straight to your use case: find the target category that matches your work and read the latency and success rate for that target specifically.
- Google SERP monitoring → the SERP latency leaderboard.
- Cloudflare-protected retail scraping → the success-rate and cleanliness table.
- Mobile operations for social media → the mobile table.
No provider is universally best. Bright Data leads on Cloudflare success rates but carries a price premium you don’t need for lighter targets. Webshare’s economics work well for Google SERP monitoring at scale, where 78 ms is perfectly acceptable and the savings are real. NodeMaven’s consistency makes it the right call when predictable performance matters more than peak performance.
If you’d rather not read the tables and decide yourself, browse our proxy use-case guides or compare every provider side by side — both are updated continuously, not just monthly, so they always reflect the current best performer for your scenario.