Geographical inequalities in mapping, imagery and language resources make open-source verification significantly more difficult and costly in conflict zones such as Sudan, leaving important evidence vulnerable to delays and doubt. Addressing this imbalance requires sustained investment in local archives, proactive mapping, language technologies and editorial transparency.
Open-source verification tools are not geographically neutral. They depend on layers of imagery, maps and linguistic data that are readily available in some places and scarce in others. This disparity becomes most evident where armed conflict coincides with limited visual and linguistic coverage.
In the early days of Russia’s invasion of Ukraine in February 2022, something unusual happened in war reporting: verification sometimes preceded publication rather than following it.
As videos poured in from Lutsk, Kharkiv and the suburbs of Kyiv, a decentralised community of researchers was examining them in real time, exchanging footage, geolocating it and identifying models of armoured vehicles. ABC News described the phenomenon in revealing terms: an “army” of geolocation specialists was publicly posting coordinates, and their findings were largely accurate.
Since then, a particular perception has become firmly established in the professional imagination: open-source verification is a borderless tool available to any journalist with a computer and patience. This perception is accurate, but incomplete, as this article will demonstrate.
The Paradox of Two Countries’ Digital Environments
The central technique in modern visual verification — geolocation — relies on matching. It involves comparing landmarks visible in footage, such as buildings, lampposts, skylines and signs, with an independent source confirming that those features exist in that precise location. Matching always requires a second point of reference.
In Ukraine, that second point of reference was available in several forms: extensive street-level imagery, highly detailed open maps, a substantial visual archive accumulated before the war, and a large volunteer community. ABC highlighted another, often overlooked, human factor: Ukrainians who could recognise a landmark from their home city in a single frame.
Verification skills do not operate in a vacuum. They require material against which footage can be compared, a map that can be searched, an archive that can be consulted, and people familiar with the location. The question, therefore, is not so much whether a journalist knows how to verify something, but whether the reference material needed for verification existed in the first place.
In Sudan, videos have been emerging from Khartoum, Omdurman and Darfur since the outbreak of war in 2023. The editorial standards governing their publication are the same as those applied to Ukraine, but the cost of meeting those standards is different. This is not necessarily because Sudanese footage is less clear, but because the tools needed to obtain independent corroborating evidence are extremely limited.
Here, “cost” does not refer only to time. It encompasses the number of tools that must be tried, the number of alternative sources that must be consulted when the first source fails, the amount of local expertise that must be secured, and the degree of uncertainty that remains after all these efforts have been exhausted. In Sudan, all these costs rise simultaneously.
In Sudan, videos have been emerging from Khartoum, Omdurman and Darfur since the outbreak of war in 2023. The editorial standards governing their publication are the same as those applied to Ukraine, but the cost of meeting those standards is different. This is not necessarily because Sudanese footage is less clear, but because the tools needed to obtain independent corroborating evidence are extremely limited.
The Consequences of Absence: Imagery and Language
According to available data on Google Street View coverage, official coverage is available in Jordan, Lebanon, Qatar, the United Arab Emirates, Türkiye and Israel. Oman was added in October 2024.
In Iraq, coverage extends to no more than a handful of selected tourist sites. Sudan, Syria, Yemen and Libya, meanwhile, lack extensive official street-level coverage that can be relied upon in the same way.
This does not mean that verification is impossible in these countries. Street-level photography is not the only source available. There are satellite images, OpenStreetMap, crowdsourced photography projects such as Mapillary and KartaView, photojournalism archives, and scattered images uploaded by users.
However, the absence of an extensive, openly accessible visual layer means that one of the richest and most easily searchable sources becomes scarce. Investigators must compensate by turning to sources that are less detailed, updated less frequently, more expensive, or all three.
This compensatory approach is visible in published investigations. When BBC Verify examined footage documenting the resumption of civilian flights at Khartoum Airport, it explained that its method involved matching prominent landmarks visible in the footage against satellite imagery, followed by a reverse image search to establish that the footage had not previously been published.
The method was entirely sound. But consider what was unavailable to the team: it had to look down from above because street-level coverage was lacking.
And because BBC Verify is a specialist unit within a major media organisation, it has access to time, subscriptions and staff that a Sudanese newsroom operating under bombardment does not.
One of the most valuable elements of any investigation is also among the hardest to replace: the human element. The existence of a large volunteer community specialising in geolocation, alongside people who recognise their cities through the smallest details, is not simply an additional advantage. These people are themselves part of the infrastructure.
One of the most prominent such communities is GeoConfirmed, a volunteer collective specialising in geolocating footage recorded in conflict zones.
Here, “cost” does not refer only to time. It encompasses the number of tools that must be tried, the number of alternative sources that must be consulted when the first source fails, the amount of local expertise that must be secured, and the degree of uncertainty that remains after all these efforts have been exhausted.
The Second Gap Is Less Visible but More Consequential
Evidence is not confined to images. It also includes what is said in recordings, what is written on signs, and what witnesses publish. Each of these layers requires automated processing tools.
An important distinction is necessary here: Modern Standard Arabic is not a low-resource language. Numerous tools have been developed for it, and their results are reasonably reliable. The problem lies with dialects, particularly those that are less well represented.
An ACM survey covering a decade of research between 2014 and 2024 found that Gulf, Saudi, Sudanese and Emirati dialects appeared only sporadically compared with other varieties.
This disparity can be measured.
A study published in January 2026 established the first benchmark for automatic speech recognition in Sudanese Arabic. When models trained on Modern Standard Arabic were applied to the Sudanese dialect, they recorded word error rates ranging from 73.8% to 123%.
The second figure is not a typographical error. A word error rate exceeding 100% means that the system's output contains more errors than the number of words in the original speech.
Even the best-performing model, following specialised training, recorded a word error rate of 57.1%.
In other words, the best model in that benchmark made errors in more than half the words. This does not make transcription impossible, but it means that the resulting text requires extensive human review before it can be accepted as a reliable transcript.
The problem is further complicated by linguistic diglossia: official documents are written in Modern Standard Arabic, while witnesses write and speak in their local dialects. Many also write these dialects using Latin characters, a practice known as “Arabizi”.
The immediate practical consequence is that it becomes difficult to establish the likely location of a recording from the speaker's dialect. Searches for witness posts are also hindered because keywords in Modern Standard Arabic do not match the words actually used.
An Unequal Burden of Proof
At this point, the issue shifts from the technical to the professional.
On the surface, the editorial standard is the same: nothing should be published unless it has been verified.
Yet applying a single standard to two environments with vastly different resources produces different outcomes. The longer verification takes, the less likely the evidence is to reach the public while a story is at the height of its news value. Footage that takes days to verify may never be published once the news cycle has moved on.
More troubling still, the phrase “we have been unable to verify” does not reach audiences merely as a description of the limitations of available tools. Instead, it is understood as a judgement on the material itself.
In the audience's mind, inadequate infrastructure is transformed into doubt about the account.
This is not a deficiency in the competence of Arab journalists or their commitment to professional standards. It is a disparity in resources that exists before the journalistic work even begins.
In an article entitled “The Political and Digital Roots of the Invisibility of Sudan's War”, published in Al Jazeera Journalism Review, the author examines the marginalisation of the war in Sudan and attributes the silence surrounding it to three factors: algorithmic preferences, geopolitical marginalisation and the destruction of local media infrastructure.
These are real and convincing factors. But a fourth can be added, one that operates before all the others: cost.
Even when footage reaches journalists, and even when someone is willing to publish it, the cost of verifying it remains higher.
Silence, therefore, is not simply a question of attention. It is also a question of proof.
The Ukrainian experience demonstrates that having someone who can recognise their city from a small detail is not an additional advantage, but part of the infrastructure itself.
This particular human element has also been tested in Sudan. The Humanitarian OpenStreetMap Team worked with organisers from Sudan to build a mapping community among Sudanese people in the diaspora and those outside the country.
The essential requirement, therefore, is not physical presence inside the country, but knowledge of the location, organisation and mutual trust.
What Can Be Built?
The purpose of this assessment is not to complain, but to identify shortcomings that can be addressed. This gap is neither natural nor permanent. It is the result of decisions concerning documentation and investment made years before the war.
Archiving: An Established Model, but Who Owns It?
Since 2019, Sudanese Archive, a programme of the Germany-based organisation Mnemonic, has been collecting and archiving material documenting violations in Sudan. Its collection has grown to more than half a million records, and it has submitted evidence files to official bodies for accountability purposes.
Similarly, since April 2023, Sudan Witness, a project run by the UK-based Centre for Information Resilience, has been monitoring, verifying and archiving incidents related to the conflict. It publishes open maps and databases documenting village fires and air strikes.
These two examples carry a dual significance.
First, the model is both feasible and proven. It is not merely a theoretical proposal.
Second, the infrastructure through which the events of this war will be documented and established is being built, funded and managed from outside the region.
This raises questions extending beyond which organisation collects the archive to who will own its knowledge infrastructure in the future: Where will it be located? Who will decide what remains in it? And who will control access to it?
What an Arab newsroom can begin doing is less ambitious but more realistic: archiving the material it publishes, recording publication dates, links and descriptions of locations wherever possible.
Verification does not require an official source. It requires an independent, pre-existing and retrievable source.
There is, however, a limitation: an archive without descriptive information cannot be used effectively. Its value lies in indexing, not merely in storage.
Proactive Mapping
Street-level photography is not the exclusive preserve of any single organisation. Crowdsourced photography projects such as Mapillary and KartaView make imagery available under open licences, with material that can be uploaded using ordinary mobile phones.
As for mapping data, Sudan itself illustrates the problem.
According to research published in Frontiers in Climate, cited by the Humanitarian OpenStreetMap Team (HOT), 83% of the buildings currently recorded in OpenStreetMap within Sudan were edited between August 2021 and August 2024 — a period spanning the outbreak of the war.
In other words, the map was drawn because the war happened.
By contrast, the idea behind Missing Maps, a collaborative initiative established by the Humanitarian OpenStreetMap Team and humanitarian organisations, is to map areas vulnerable to crises before those crises occur, rather than afterwards.
The role available to Arab newsrooms and universities is specific and inexpensive: photograph routes through their cities and upload the imagery, while also contributing data about those cities to open maps before they become front lines.
The limitations are clear. This approach cannot cover areas that are already inaccessible, nor does it help someone seeking to verify footage recorded today.
It is an investment in verification that may take place years later — and that is precisely why it is constantly postponed.
The Human Element
The Ukrainian experience demonstrates that having someone who can recognise their city from a small detail is not an additional advantage, but part of the infrastructure itself.
This particular human element has also been tested in Sudan. The Humanitarian OpenStreetMap Team worked with organisers from Sudan to build a mapping community among Sudanese people in the diaspora and those outside the country.
The essential requirement, therefore, is not physical presence inside the country, but knowledge of the location, organisation and mutual trust.
However, the model cannot simply be transferred unchanged.
Publicly sharing coordinates during an ongoing conflict can carry serious risks for people on the ground. What is appropriate for a volunteer community working remotely may not be appropriate for someone filming from their window.
Organisation and safety in the field must therefore take priority.
Language: Computing Power Is Not the Obstacle
The benchmark for Sudanese Arabic was developed using free and experimental computing resources.
This helps redefine the obstacle: the problem is not computing capacity, but the availability of annotated audio data.
Newsrooms possess something that research laboratories often lack: hours of speech in authentic local dialects, preserved in archives of interviews and field reports.
Making some of this material available to researchers, even in limited samples, could bring about a qualitative improvement in automatic speech-recognition tools.
This must, however, take into account the need for legal consent and guarantees of safety for the people whose voices appear in the recordings.
Editorial Transparency
This requires neither substantial funding nor years of preparation. It requires an editorial decision.
Instead of simply stating, “We have been unable to verify”, newsrooms could publish the steps they have completed, explain what could not be established, and clarify why.
Was an attempt at matching made but unsuccessful? Or was there no reference material available against which the footage could be matched in the first place?
Such a decision would shift attention towards the limitations of the tools rather than casting doubt on the credibility of the event itself.
Who Should Bear the Cost?
A Sudanese newsroom operating under fire cannot reasonably be expected to build an archive of cities being bombarded or collect linguistic data while struggling to publish the news.
The responsibility falls instead on better-equipped regional institutions and Arab universities operating at a safe distance from the war.
This capacity is not hypothetical.
On 31 August 2025, when reports emerged of a landslide in the village of Tersin in Darfur, claiming that the entire village population had been killed, an Al Jazeera report investigated the incident using the same methods described here.
Despite the scale of the reported disaster, witness-generated content was unavailable.
Investigators therefore turned to satellite imagery, which showed that the village covered an area of no more than approximately 2.7 square kilometres and contained relatively few houses.
This evidence challenged the widely circulated death toll, while subsequent findings confirmed that the landslide itself had occurred.
What is missing, therefore, is not proof that such verification is possible, but sustained foundational investment in archives, maps and language resources.
It is often said that open-source verification has given journalism tools it did not previously possess. That is true.
But the question worth asking is not about the tools themselves. It is about the ground on which they stand.
When verification tools are built on data from one part of the world but not another, what is implicitly determined is not merely the level of accuracy, but the distribution of doubt.
Which accounts are accepted quickly? And which will remain unresolved?
The question facing Arab newsrooms is not, “Were they capable of verifying the evidence?”
They have already demonstrated that they are, even with fewer resources.
Rather, the question is:
“How much evidence will remain in the ‘unconfirmed’ category before Arab newsrooms and universities build the foundations on which their verification work depends?”
This article was originally published in Arabic on 08/11/2026