Since US and Israeli strikes began on the last day of February, millions of Iranians have been living under attack, an internet blackout and tight restrictions on journalists and humanitarian agencies.
But many people are on the move, trying to get away from dangerous places or to be reunited with family at a time of conflict. In an information blackout, with internet access almost completely shut down across Iran, it’s hard to build a detailed picture of this population movement. But in the absence of conventional data on internal population displacement, we have been piecing together where people are moving by looking at faint but persistent signals of internet activity.
Our latest analysis and situation report covering the war since its outbreak, shows a clear geographic pattern and timeline of movement.
This is one of the first near real-time pictures of displacement within Iran. It complements cross-border figures from the UN’s International Organization for Migration, which recorded roughly 40,000 departures from Iran between March 3 and 10, mainly to Afghanistan, Pakistan, Turkey and Azerbaijan.
Our data offers a partial view of movement inside the country, where conventional methods of counting displaced people have largely broken down.
What the data show
In the first days of the war, our estimates indicate relative increases in population presence in provinces near the borders with Turkey, Armenia and Azerbaijan. As the conflict evolved, the pattern shifted eastward and towards the capital. By the third week, provinces bordering Afghanistan and Tehran showed the strongest signs of population concentration.
Tehran stands out. Despite being repeatedly struck by Israeli and US missiles, the Iranian capital shows what appears to be a modest rise in population compared with its pre-war baseline. That is consistent with research on other conflicts, where capital cities often absorb displaced people. This is because, even under bombardment, they usually offer better access to services and infrastructure.
Central and southwestern provinces, such as Qom, Isfahan, Fars and Zanjan/Qazvin – several of which host nuclear, military and defence production sites – show signs of sustained declines in estimated presence. These are also the areas with the highest concentration of recorded strikes on the Iran Strike Map, an open-source intelligence website which plots strikes on and by Iran in this conflict based on verified reports. The alignment between strikes and population declines is one of the strongest validation points in our analysis.
How we know
Weeks of active hostilities and Iran’s tight information controls have closed off most of the usual population statistics we might rely on to track population movements. Instead, we use what researchers call digital trace data – the everyday digital footprints people leave when they use connected devices.
GPS-based mobile data and Meta’s population maps have been useful in other crises, but for Iran, they are unavailable. So our main source is Cloudflare Radar, a US-based content delivery network which publishes aggregated, anonymised counts of encrypted web requests passing through its network, broken down by province.
Despite the widespread internet shutdowns, some weak internet signal remains and we were able use it, translate it to population numbers and compare these numbers with a baseline control set in December 2025 to assess increases and decreases in population. More requests than usual is a tentative signal that more people are present and online. Fewer requests may suggest fewer people or less activity.
We built a baseline model for December 2025 translating provincial internet traffic to population numbers, using WorldPop population estimates. We then applied that baseline to each day of the war, adjusting for network shocks and coverage, and cross-checked the patterns against Farsi Wikipedia pageviews for border regions and against recorded strike locations. Obviously at a time of internet restriction pageviews tend to be very few, so this information serves as validation only for our other evidence. A full account of the methods we used, with interactive maps, are on the project website.
Why it matters
The UN’s International Organization for Migration has already reported rapidly evolving displacement across more than 20 Iranian provinces. But with the internet cut, journalists barred and little official information available, even a rough picture of internal movement matters. Our findings point humanitarian agencies to three pressure points: the northwestern border corridor, the provinces adjoining Afghanistan and Tehran’s hinterland.
These patterns also matter politically. US and Israeli officials have framed the campaign as a targeted operation against Iran’s nuclear, missile and leadership infrastructure. Our data indicate whether strikes hit their intended targets. But they do show that the civilian response extends well beyond the struck sites. Estimated population is falling across several provinces and rising in others, including areas without major military infrastructure. However precise the targeting, the human footprint of this war is broad and spatially uneven.
What the data cannot show
These are proxy estimates, not head counts – they capture relative population change, not absolute numbers. There are three main caveats to consider.
First, Iran’s near-total internet blackout has kept national connectivity at 1–4% of normal levels for much of this period. A drop in requests from a province could reflect people leaving. It could also mean a cut cable or a shutdown order. We adjust for these effects, but uncertainty remains high.
Second, the data only capture people with internet-connected devices. Although we adjusted our estimates to mitigate biases, children, the elderly and poorer households may be underrepresented. Ethnic minorities who read primarily in Azerbaijani Turkish or Kurdish are less visible in our Farsi Wikipedia cross-check, which covers roughly half the population.
Third, we analyse movements that correlate with or follow attacks, not movements caused by them. People also flee ahead of strikes, return between them or move for reasons unrelated to the war. The alignment with strike data strengthens the case, but it does not prove it.
In past crises, from Ukraine to Sudan, researchers and humanitarian agencies have increasingly turned to digital trace data when the usual sources are unavailable. Iran is a hard case. Since the war began, the state has imposed a near-total internet blackout, keeping connectivity for officials and state media but cutting off most of the population, using control of the network as an instrument of wartime information control.
Even so, the digital traces still carry information about where life goes on, and where it has stopped. Used carefully – and with clear caveats – they can help the outside world maintain some visibility of a population that is otherwise hard to see.
Francisco Rowe receives funding from the Economic Social Research Council for supporting DEBIAS (ES/Y010787/1).
Carmen Cabrera receives funding from UK's Economic and Social Research Council.
Elisabetta Pietrostefani does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.
This article was originally published on The Conversation. Read the original article.
