Data snapshot: October 4, 2026, 20:17 UTC. Atlas source data as of 15:21:09 UTC. The country score covers a rolling 30-day OONI measurement window; the July evidence record was read separately at 20:17 UTC.
Iran's row in the Voidly Atlas country index showed a score of 16 from 6,196 OONI samples when we read it on October 4. Atlas labeled that measured anomaly signal “low.” A separate July 1 evidence record linked 16 OONI entries concerning messaging apps. Neither number measures how many people in Iran could reach those apps. The first describes a rolling sample; the second counts links in a dated record. Their limits are part of the result.
What the score can say
Atlas calculates its score from the share of available OONI measurements marked anomalous during the rolling 30-day window, with a small-sample adjustment before rounding. The public response exposes the score, sample count, method and source-data timestamp together. Iran's score of 16 is therefore a statement about that response and its method. It is not “16% of the country is censored,” a measurement of app users or a ranking of people's lived access.
The denominator matters because OONI data comes from people running tests on particular networks, at particular times, against particular services. OONI's interpretation guide says coverage changes with participation, networks and test choices. More samples can improve the view of a tested signal, but a national total alone does not show whether those samples span the networks or times needed for a specific claim. The index supplies a useful lead; an app-level claim requires the relevant tests.
“Anomalous” also needs care. OONI uses that label for a result that may indicate interference and needs closer inspection. A transient network problem, a test bug or a change in an app provider's infrastructure can produce a false positive. For messaging apps, OONI recommends looking at repeated results and at the network on which the test ran. A country score aggregates those flags; it does not automatically confirm every flagged app as blocked.
What the July links add
The Iran evidence response contains 16 entries from OONI and identifies one source family. Its corroboration field calls that level low. The entries are arranged across four named tests—Signal, WhatsApp, Telegram and Facebook Messenger—with four listed observation times on July 1. That structure gives a reader routes into the Signal, WhatsApp, Telegram and Facebook Messenger measurements.
Sixteen links do not mean 16 independent investigators, 16 separate network outages or complete coverage of Iran. The upstream search pages can be filtered further by network and time. A stronger service-specific finding would compare anomalous and successful tests on the same networks, inspect their failure modes and see whether the pattern persists. The evidence response gives that investigation a starting point. Its count cannot establish who caused any interference or how many people encountered it.
Nor should the July links be folded into the October rolling score. The record's observations and the index's 30-day window answer different questions and are separated in time. A reader who wants to know what was measured in July should open the linked tests. A reader who wants the latest Atlas score should re-fetch the country index and check its dataAsOf field rather than treating its per-request serve time as a new source observation.
A prior case shows why coverage can shrink
A separate OONI investigation published in August 2025 examined WhatsApp testing in Iran during June and July 2025. OONI reported a spike in anomalous WhatsApp tests on some networks, especially MCI and TCI. It also found a sharp loss of measurement coverage during the June connectivity disruption and compared that gap with IODA and Cloudflare Radar signals. Those are findings about a 2025 period; they are not a description of the July 2026 record or October 2026 access.
The case illustrates two distinct risks in reading a measurement chart. First, an anomaly needs interpretation: OONI looked at test behavior and the networks involved, rather than counting orange results alone. Second, a period with few or no uploaded tests can reflect a connectivity problem that prevents probes from publishing. A blank part of a chart is not evidence that access was open. It may mean the observer could not report.
That lesson does not authorize a numeric correction to Iran's October score. We have no basis here to estimate how many additional tests would have arrived under different conditions. It does tell us to carry the sample count, source timestamp and network coverage alongside the score, and to keep older case studies clearly dated.
A defensible reading
For this October snapshot, the narrow claim is clear: Atlas recorded 6,196 OONI samples for Iran in its rolling window and assigned a score of 16. Its separate July record exposed 16 OONI links across four messaging tests and one counted source family. OONI's guide explains why these are leads for network-level review, not a census of nationwide access.
The next useful check is concrete: choose one linked service, inspect repeated measurements and failures by network, compare the period before and after the observation, and seek an independent connectivity source if the question concerns a broader disruption. Publish what those checks support. Leave the reach and cause of the event open where the data do not answer them.