Two views of the signal.
Traffic trends and source-linked interference evidence. Align the time and network before comparing them.
Different observation systems can add context. Their agreement does not by itself prove intentional censorship or nationwide scope.
Same question. Both perspectives.
Capabilities in context, with source links. No color-coded winners or unmatched accuracy scores.
Documentation reviewed . The dates and scope of the underlying evidence remain separate. Unknown coverage or unverified interface absence is not scored as “no.”
Cloudflare Radar
Radar exposes traffic, DNS, attack and outage views from Cloudflare’s network and other documented inputs. It publishes outage material, BGP analysis tools and API access. These are useful for traffic/routing questions without requiring a claim that one provider sees a fixed share of the whole internet.
Explore Cloudflare Radar ↗Voidly
Voidly organizes evidence from multiple sources into incident and country records, with protocol or domain attribution where available. Source lists are not automatically independent corroboration: Citizen Lab lists are context, and two views may share observations. Missing evidence remains missing.
How we measure ↗API docs ↗Follow the work
Choose by the task, then check the actual record. Each source has a different role.
Inspect Radar’s outage/traffic views and other observations for the same time and network. A traffic decline is evidence of disruption, not by itself proof of deliberate shutdown or national scope.
Use evidence that actually names the domain, network, protocol and time. Voidly can link such evidence where available; a country outage record without domain attribution is insufficient.
Inspect the model’s target and validation first. Country-ranking risk and next-seven-day OONI events differ from precise shutdown onset. The historical AUC 0.98 must not be used as current forecasting accuracy.
Start with Radar’s documented BGP and attack views. Keep detected routing anomalies, attack telemetry and confirmed incident causes distinct.
Use the documented accessibility/API/MCP response as a scoped signal. Separate provider geography restrictions, measured reachability, stale or missing evidence and the user’s actual network.
Use both where relevant, aligning geography and observation windows. Check source overlap and alternative causes; coincident traffic drops and blockpages do not automatically establish one shared cause.
Use the overlap
Compare the location, network, timestamps and uncertainty first. A Radar traffic decline and a Voidly record may concern different populations or causes. Use agreement as context, then inspect the source evidence; neither observation alone proves a shutdown’s intent.
The source trail
This page’s conclusions are editorial comparisons of the linked documentation, not a controlled comparative evaluation.
Voidly-original material is CC BY 4.0. Upstream data keeps its source license. Public access and licensing are different questions; current account-key allowances and endpoint limits are documented in Pricing. Current task/settlement release details are on Voidpay.
Historical text captured from the original source on 2026-09-06. That is a capture date, not an observation date. These earlier claims are preserved for review, may be stale or incorrect, and should not be copied as current facts. The corrected comparison above takes precedence.
How Voidly compares to Cloudflare Radar. Cloudflare sees a vast slice of the internet. Voidly is censorship-specific with multi-source aggregation and citable incidents.
Cloudflare Radar shows what their network sees. Voidly shows what censors do, across many networks.
| Feature | Voidly | Cloudflare Radar |
|---|---|---|
| Vantage point | Yes4 independent sources + own probes | PartialCloudflare network only |
| Scale of underlying data | Partial3.5B+ measurements (via OONI) | Yes~20% of internet traffic |
| Censorship-specific focus | YesSole focus | PartialOne section of broader product |
| Citable incident IDs | YesIR-2026-0142, 7,500+ incidents | PartialOutage Center events |
| Per-domain blocking detection | YesDNS, TLS, HTTP, blockpages per domain | PartialAS-level traffic anomalies |
| Free public API | YesCC BY 4.0 | YesFree with API token |
| AI/MCP integration | Yes@voidly/mcp-server, 84 tools | NoNo MCP server |
| ML classification | YesGradientBoosting v3.3, LOCO mean F1 0.71 (0.63 on the countries with n≥30) | PartialAnomaly detection |
| BGP / route data | NoOut of scope | YesBest-in-class |
| DDoS / attack telemetry | NoOut of scope | YesBest-in-class |
| Independent of vendor self-reporting | YesExternal probes + 3rd-party feeds | NoCloudflare-network view |
| Predictive forecasts | YesSentinel 7-day forecast, AUC 0.98 | NoReal-time only |
Cloudflare Radar is unmatched on scale. Cloudflare's network sees roughly 20% of internet traffic, which means Radar can show real-time DNS query patterns, BGP route changes, attack traffic, AS-level outages, and HTTP request distribution at a granularity nobody else has. Their internet quality and certificate transparency dashboards are excellent. The team publishes year-in-review reports and Outage Center alerts that have become a primary citation for journalists covering internet shutdowns. The tooling is free, the data updates near-real-time, and the visualizations are best-in-class. For "is the internet up in country X right now," Radar is often the fastest answer.
Cloudflare Radar's strength — its singular vantage point — is also its constraint. It sees what Cloudflare sees: requests to Cloudflare-fronted properties, queries to 1.1.1.1, BGP from their peers. That is enormous, but it is one perspective. Voidly aggregates four independent measurement sources (OONI, CensoredPlanet, IODA, Citizen Lab) plus our own probe fleet (live size at api.voidly.ai/v1/probe/network), specifically to triangulate censorship events from outside any single vendor's view. We focus narrowly on censorship — DNS poisoning, TLS interference, ASN-specific blocking, blockpages — and produce tracked incidents with structured evidence (7,500+ to date). We also publish the data under CC BY 4.0 and expose it via MCP for AI agents. Radar is broader; Voidly is deeper on this one slice.
Cloudflare Radar — they see traffic drop in real time. Voidly is 6h-batched for evidence quality.
Voidly. Radar shows aggregate AS-level traffic; Voidly shows per-domain DNS poisoning, TLS resets, and blockpages with evidence permalinks.
Voidly Sentinel (7-day forecast, AUC 0.98). Radar is real-time-only.
Cloudflare Radar. Voidly does not cover routing or attack telemetry.
Voidly /v1/accessibility API or MCP server. Radar does not expose per-domain accessibility queries this way.
Use both. If Radar shows a traffic drop and Voidly shows blockpages on the same date, the evidence triangulates.
Use both. Radar is the right tool for "what is happening on the internet right now at a global scale." Voidly is the right tool for "what was blocked in country X last Tuesday, and where is the evidence." A serious analyst writing about a shutdown event should check Radar for traffic-level confirmation and Voidly for protocol-level evidence and citable incidents. Cloudflare publishes their own outage reports openly; we link to them when they overlap with our incidents.
The old page’s final CTA described all data/ML/forecast reads as free without a key, applied one CC BY license, and described a handful of metered Voidpay endpoints. Current access, source licensing and Pay releases must be checked separately; that text was not a universal capability or pricing guarantee.
Complete historical comparison ↗Public data access
Start with the public dataset. Check its source terms, dates and endpoint limits.
Keep exploring
Country research and source-linked incident records. Two different ways to understand internet freedom.
OONI publishes network measurements. Voidly connects selected observations to source-linked records.
