2025-06-02 15:00:53
Fentanyl, a powerful opioid 50 to 100 times more potent than morphine, has also flooded illegal markets since 2013. The effects have been devastating: synthetic opioid overdoses, largely driven by fentanyl, have claimed tens of thousands of lives each year, peaking at 110,000 deaths in the US in 2022. The drug is relatively easy to produce and can be made entirely in a lab using a few essential chemicals known as “precursors.” But it’s difficult for US regulators to prevent the creation of fentanyl simply by banning precursors, because new or different combinations of these chemicals are created faster than regulations can take effect.
To show the business of fentanyl production in the US today, a team of Reuters investigators led by Maurice Tamman, the editor-in-charge of computational and forensic journalism, pitched an audacious idea: the team would legally purchase the precursor chemicals needed to fully create fentanyl and have them shipped to New York and Mexico City using international carries such as FedEx. Over the course of their reporting, they succeeded in purchasing enough precursor materials to create up to $3 million worth of fentanyl tablets for little more than US$3,600 in bitcoin, mostly from international sellers based in China.
Their resulting seven-part investigative series, Fentanyl Express, revealed just how easily chemicals used to create fentanyl can be acquired, and was recently awarded the 2025 Pulitzer Prize for Investigative Reporting.
GIJN spoke to Tamman, Reuters correspondent Laura Gottesdiener, and Stephen Eisenhammer, Reuters’ Bureau Chief for Mexico and Central America, to walk through how they reported the story.
Image: Screenshot, Pulitzer Prize Committee
Coming off an investigation that tracked the smuggling of electronics into Russia to make weapons for its invasion of Ukraine, Tamman considered using a similar approach to track import-export data in Mexico — a country at the heart of the international fentanyl trade due to the role of Mexican organized crime in producing and trafficking illegal fentanyl.
But that strategy hit a dead end, so the team pivoted to contact the overseas sellers directly. “What I was working on at the time was trying to understand, essentially, the formula. How are these things advertised? What are the kinds of chemicals we really wanted to get? Which ones could we legally get? Which was a lot of understanding the different regulations in different countries,” Gottesdiener explains.
They decided to try legally purchasing chemicals originating from Chinese companies that had been indicted by the US Department of Justice for their role in the fentanyl trade. “We thought, if we can test what impact these indictments make by contacting the companies that had been indicted and seeing if they would still be willing to sell fentanyl precursor chemicals, either to the United States or Mexico,” he adds.
The team quickly found that these chemicals could easily be found on the public internet, on social media, Google Image Search, and even Soundcloud. From there, it was a matter of contacting the sellers, which was done through encrypted messaging platforms like Signal.
“These relationships we developed with these sellers were pretty dynamic and kind of curious,” Tamman says. “They are at once flirty and friendly, but also completely agnostic as to why we were getting this stuff. They didn’t ask questions. They had no curiosity about what the purpose was, although they were clearly advertising these chemicals for the purposes of making fentanyl.”
Ultimately, the team was able to acquire 11 of the 12 precursor chemicals needed to create fentanyl from seven separate Chinese suppliers (the 12th chemical was acquired through an order from a US chemical company via Amazon), purchasing them through transactions made through the difficult-to-track blockchain currency bitcoin.
By forming consistent relationships over the course of a year with sellers — including landing multiple buys from the same sellers — the team was able to see how dynamic the fentanyl industry was.
“At the very beginning of the project, they [the sellers] would offer you a very straight pre-precursor for [the key fentanyl chemical compound] piperidine, for example. Six, nine months later, they were not hawking those anymore. They were hawking variations of those because they were trying to get around regulation,” Gottesdiener explains. “So by continuing to talk with these people, buy from these people, we were able to sort of see just this tiny snapshot about how their market was evolving.”
The fentanyl trade is embedded in the global supply chain. As many precursor chemicals are legal and used for legitimate means, such as in pharmaceuticals and perfumes, the team found it easy to import the chemicals into the US, including through carrier services like FedEx. In total, 14 chemical deliveries were made, including six to Mexico City, seven to a mailbox in New Jersey, and one to an apartment in New York City.
The diversity in delivery locations was intentional. “There has been — and continues to be, to some extent — a fairly simplistic narrative from some areas of government and law enforcement, that this stuff all just comes in and is produced in Mexico and is then consumed in the US, a very kind of binary narrative,” Eisenhammer said. “We really wanted to show that it’s not as simple as that, that these supply chains are incredibly complicated and actually the US plays a pretty important role.”
The Reuters team’s reporting showed that the US plays a key link in the supply chain of providing fentanyl precursors to illegal drug manufacturers in Mexico. Image: Screenshot, Reuters
Tamman adds: “By making that choice to try and ship it into the states as well as into Mexico, we opened up an avenue of reporting that we probably hadn’t anticipated.”
In at least two situations, the team found gaps in US regulations and customs enforcement. In one case, a package containing a precursor chemical mislabeled by the seller as “ink” was held by FedEx until an updated form could be provided by the team. On advice of their legal team, the team filled out an updated form that correctly identified the chemical being shipped, and the package was subsequently waved through. In another situation, a package containing a precursor chemical was searched by US customs, but was still permitted to enter the country without issue.
“Clearly they looked at it and they said — this chemical, this package is suspect, but they opened it up, they resealed it, and they sent it along its way, so it arrived. No problems, right? And it’s remarkable to me that that was allowed to happen,” Tamman says.
After acquiring the chemicals, there was one more step: verifying that the chemicals they received could, in fact, be used for fentanyl. To do so, they approached a chemist at the Center for Forensic Science Research and Education, who identified the chemicals as fentanyl precursors, as well as one unidentified chemical that later proved to be a little-known “designer precursor” not on the list of US-regulated chemicals.
The team did not make fentanyl with the chemicals, and ensured the chemicals were safely destroyed after identification.
Fentanyl precursors were shipped from China in a variety of container sizes, and some were mislabeled to avoid detection. But other suspicious packages were flagged by customs officials and searched yet still allowed to proceed for delivery. Image: Screenshot, Reuters
The team wanted to ensure that every transaction, and every chemical imported, was legal according to US and Mexican law — not just for their own ethical and legal reasons, but to show how easily fentanyl ingredients could be obtained.
“We thought about where we have the chemical sent to, then the legal concerns were also really significant, and involved a lot of research here in Mexico… around what things were banned, what wasn’t banned, what chemicals we could buy, what we couldn’t buy,” Eisenhammer explains. “We also have in-house lawyers who went through all of this with a pretty fine-toothed comb.”
During their reporting, laws in the US and Mexico were updated to target different precursor chemicals, so the team had to be continually aware of what was currently legal to buy and import.
“One of the things that the lawyers stressed when we were doing the reporting was that you could not ask the sellers to do anything illegal,” he says. “If there was anything around mislabeling, or how they were going to get the stuff into Mexico, they [the sellers] had to do that entirely of their own accord.”
The team also maintained strict security protocols throughout the investigation. To contact sellers, journalists used burner phones, burner SIM cards, and even a burner laptop, and used end-to-end encrypted platforms like Signal and WhatsApp to communicate with sellers.
The team was also careful about personal security, and took precautions against inadvertently provoking cartels or drug traffickers during their transactions. “Mexico has a terrible reputation for attacks on journalists, murders of journalists, and so that was something we took incredibly seriously,” Eisenhammer points out.
The Reuters team wasn’t able to uncover every aspect of the transnational fentanyl trade. “Reuters couldn’t determine whether any of the Chinese suppliers were the actual manufacturers of the chemicals received or simply middlemen. Nor could the news organization determine where the operations were located. Reporters could dig up nothing more than phone numbers for two of the sellers. For the others, corporate websites and Chinese business-registry documents yielded addresses,” the article states.
Reuters did send reporters to one of the locations listed — an office tower in Wuhan, China — where a seller, Hubei Amarvel Biotec, was registered, but was told by the building’s management that the company did not rent a space there. Nevertheless, the team found it important to follow up on the lead: “Just because they’re in China doesn’t change how we go about doing our reporting. We had an address for these outfits, and we certainly wanted to see whether or not they were legitimate, if there was somebody there who could talk to us,” Tamman says.
While the investigation helped illuminate the nature of the global fentanyl trade, many questions remain unanswered that could prompt future investigations. “I think we did break new ground with it, and did get closer to understanding the supply chain than anyone has done before, but we didn’t get all the way,” Eisenhammer says. “Where were they [the sellers] getting these chemicals from, and where were they producing them? Were they getting them from someone else? Those were questions that we weren’t really able to answer.”
For investigative reporters looking to dig into big, front-page stories like fentanyl, looking for unique angles is key, Tamman says. “You have to think about reporting in ways or approaches to stories that cannot be replicated,” he adds.
But Gottesdiener also notes that even criminal networks or drug rings share similarities with legitimate businesses: “Remember always that it is a market governed by logical financial interest, and there’s always going to be a reason, a logical reason, that crime groups are doing the things they’re doing.”
Devin Windelspecht is a Washington DC-based writer and editor with a passion for solving pressing global problems through journalism. His writing highlights the work of independent journalists covering some of the most pressing topics of today, including conflict, human rights, climate change and democracy. His reporting has highlighted the work of pro-democracy reporters in Russia, environmental reporters in Brazil, journalists covering reproductive rights in the US, war correspondents in Ukraine, and exiled media outlets covering authoritarian countries around the world.
2025-05-30 15:00:04
The last few weeks have been quieter for data journalism. Instead of one or two major news items dominating visualizations, a wide variety of topics were covered, such as an analysis of the impact of measles vaccination since its introduction in the 1960s and an investigation into the narco-submarines that transport cocaine across the Atlantic and Pacific Oceans. This edition of our Top 10 in Data Journalism, which considered stories between May 10 and May 23, also highlights a special report on the fate of Jews from the Polish city of Lubartów during World War II; an examination of the number of plane crashes in the US; an investigation into an ambitious reforestation program in the Philippines; and the list of Sigma Awards winners.
Image: Screenshot, Our World in Data
Measles has made a comeback in doctors’ offices and in headlines, with countries around the world seeing a rapid increase in cases. In this report, Our World in Data discussed the history and impact of vaccines against measles — a highly contagious disease that was once widespread, affecting more than 90% of children just 60 years ago. The first effective vaccine was developed in 1963, and vaccination rates have increased significantly since then — first in wealthier countries and then worldwide in the 1970s and 1980s. According to the report, the impact of measles vaccination was rapid and substantial, with an estimated two to three million deaths each year prevented over the past half century, making measles vaccines among the most life-saving vaccines currently in use.
Image: Screenshot, InSight Crime
In 2024, authorities intercepted a near-record number of so-called narco-submarines crossing the Atlantic and Pacific Oceans. To understand what the increasing use of these vessels says about drug trafficking routes and criminal dynamics, InSight Crime analyzed hundreds of these seizures. According to the report, these vessels have been around for decades and are traditionally used to transport drugs from the Pacific coast of Colombia to Central America or Mexico, but are now expanding to new areas, such as the Azores, Sierra Leone, and even Australia. Graphs show, for example, the increase in the number of narco-submarine seizures between 1988 and 2025 and a map of them, and illustrations of the different types of vessels used on trafficking routes. According to the report, traffickers are constantly adapting their vessels to avoid interception.
Image: Screenshot, WeDoData
Designed by French dataviz agency WeDoData, the scrollytelling special, Visualizing Lubartworld, traced the “collective biography” of the Jewish population in the Polish village of Lubartów throughout the 20th century. It presented the results of the Lubartworld project, led by the French National Centre for Scientific Research (CNRS) and the School for Advanced Studies in the Social Sciences (EHESS). The report us divided into three chapters and throughout the narrative, it is possible to see historical documents and photographs that illustrate and corroborate the story: The first presents a portrait of the population of Lubartów in the 1930s — where less than 8,000 Jewish and Catholic inhabitants lived — through maps, network graphs, photographs and animations; the second chapter tells the story of the various displacements of the village’s Jewish population — around 3,400 people — before and during World War II until only 18 Jews remained in the region; and the third and final chapter features an animation based on a 3D globe with a timeline, which tells the story of the migration flows of the population of Lubartów to the rest of the world during the 20th century.
A Tagesanzeiger analysis found that Eastern European countries awards performers from their own region the most points in the Eurovision Song Contest. Image: Screenshot, Tagesanzeiger
Held annually since 1956, the Eurovision Song Contest (ESC) is one of the most eagerly awaited music events for European audiences each year. This year alone, 166 million people watched the competition — held in Basel, Switzerland — on television. Shortly before the final on May 17, the Swiss daily Tagesanzeiger ran a special report analyzing the event’s numbers. A series of graphs showed, for example, that the number of participants is decreasing due to finances, politics, and war; “block voting” occurs between countries that are historically, politically, culturally, or linguistically linked; and also how the competition’s points system works. The team also reflects on what makes a song a winner, through an analysis of the attributes used by the streaming service Spotify to characterize songs. The report also says that Ireland and Sweden are the two nations that have won the ESC most often — seven times each.
Image: Screenshot, CNN
In this special report, CNN examined commercial aviation safety in the United States, citing a series of recent incidents that have raised concerns among travelers — notably the January 2024 crash in Washington, DC, in which 67 people died in a collision between an American Airlines plane and a Black Hawk helicopter. The team analyzed all aviation investigations conducted by the National Transportation Safety Board (NTSB) from January 2000 to March 2025, which included accidents with injuries, fatalities, or major damage, and asserts that flying has never been safer. According to the report, between January 2024 and March 2025, there were only 60 incidents involving commercial carriers, with two resulting in fatalities — including the accident in the US capital. Before that, the last major collision was in 2009. It also featured data on near-misses involving high-profile aircraft and data from the Department of Transportation on the share of fatalities in the US by mode of transportation from 1990 to 2022 — which showed how much safer flying is than driving.
Image: Screenshot, USA Today
Dragonfly is the name of a NASA mission planned to study Titan, one of Saturn’s 274 moons. Titan has long fascinated scientists because, within our solar system, it has the atmosphere most similar to Earth’s. To reach the moon’s surface, NASA plans to use an eight-rotor probe, similar to a drone, that recently completed a crucial test. Through a series of illustrations, USA Today explained in detail the mission planned for launch in 2028, which has faced several delays and budget overruns. According to NASA, the probe has a total life cycle cost of US$ 3.35 billion.
Reforestation Failure in the Philippines
Image: Screenshot, Davao Today
The Philippines’ National Greening Program (NGP) is considered the country’s most ambitious reforestation program to date, hoping to preserve the rainforest and promote sustainable agriculture by planting 1.8 billion trees across more than 2 million hectares of public lands devastated by decades of rampant logging. But 14 years after its launch, an investigation by DavaoToday, Thibi, and Lighthouse Reports showed how the program has failed on both counts. The team used machine learning tools to analyze millions of satellite images and detect deforestation at more than 100,000 NGP sites, analyzed official government data on program spending and the status of reforestation sites, and conducted interviews with whistleblowers. Among the report’s findings: one in every 25 hectares of NGP land has suffered major deforestation rather than reforestation, with many of the areas designated as protected having no trees at all, let alone thriving rainforests of native species.
Image: Screenshot, The New York Times
On January 5, New York City implemented a congestion zone system for the lower third of Manhattan, similar to those already been adopted by major cities such as London, Milan, and Gothenburg. This system charges additional tolls for vehicles entering the area at busier times of day to reduce traffic and fund public transportation. According to The New York Times, almost immediately after it was implemented, the policy began to alter traffic patterns, passenger behavior, public transportation, and city noise. Using maps and animations, the Times showed that, compared to a year earlier, 80,000 fewer cars per day were entering lower Manhattan; 15% fewer car accidents with injuries occurred; local bus lines ran about 3.5% faster, with some routes seeing an increase of up to 28%; and complaints about vehicle noise and fines have fallen by nearly half.
Image: Screenshot, Público
In this special report, the Portuguese newspaper Público analogized electoral voting trends to shifting winds to analyze Portugal’s political map after its legislative elections on May 18. The scrollytelling report presented a series of interactive maps with arrows indicating the growth that political parties have had in relation to the last election. According to the data, the 2024 trend continued this year too, as Portugal’s voters leaned even further to the right, with the populist right-wing party Chega — described in the report as a hurricane — becoming the second largest political force in the country. The newspaper also described the center-right Democratic Alliance as a current: it was the most-voted-for party, electing 88 deputies. According to the report, the biggest losers on election night were the Socialist Party (PS), which came in third, and the Left Bloc (BE), which won only one seat in the Assembly of the Republic.
Image: Screenshot, Moody’s
In the months right before wildfires consumed large swaths of the Los Angeles metropolitan area, many insurers had canceled large numbers of local policies. In this report, Moody’s presented 10 potential major risks that are changing the insurance industry. Charts show the age of condominium units in Florida to note that “paradise” is now located in ever more risky places; a map of major floods since 2011 that have significantly disrupted supply chains; a network diagram of the transmission of H5N1 (avian flu); and a heatmap of wildfire risks in the US and active fire locations around the world monthly since May 2024. In addition to these potential risks, the report also lists flash flooding caused by increasing heavy rainfall, IT outages, building collapses, and contamination by forever chemicals (PFAS) and microplastics.
Image: GIJN
The Sigma Awards — a competition celebrating the world’s best data journalism now hosted by GIJN — has announced its 2025 winners. Ten data-driven journalism projects were selected from 498 entries from 80 countries by a Prize Committee of 17 judges. Check out the award winners, which include stories that have exposed the impact of violence against authorities on public life in Mexico; the recruitment of prisoners for a mercenary army in Russia’s war in Ukraine; and a land trafficking scheme by organized crime syndicates in Ecuador — and be inspired!
Ana Beatriz Assam is GIJN’s Portuguese editor and a Brazilian journalist. She has worked as a freelance reporter for the newspaper O Estado de São Paulo, mainly covering stories featuring data journalism. She has also worked for the Brazilian Association of Investigative Journalism (Abraji) as an assistant coordinator of journalism courses.
2025-05-29 15:00:41
You know that feeling when you’re staring at a crucial piece of evidence — a blurry license plate, a pixelated document with names you can’t make out, or a grainy screenshot where you know the information you need is right there, taunting you from behind a wall of pixels?
You don’t? Lucky you. I encounter this problem constantly in my investigations — whether it’s extracting names from footage, decoding partial numbers from social media posts, or reading distorted text in documents. While everyone else is out there living their lives, I’m playing “Guess That Pixel.” I don’t mind. The ability to transform the unreadable into readable intelligence is awesome.
Time for a manual on making blurry nonsense make sense. The tools and techniques in this article aren’t theoretical. They’re practical methods you can apply to your own difficult-to-read evidence. Because in OSINT, the difference between a dead end and a breakthrough often comes down to making those last few pixels count.
The real work happens between your ears and behind your eyes — knowing which tools to combine, how to verify outputs, and when to trust (or distrust) your results. Because at the end of the day, the difference between amateur hour and professional investigation is having a system that consistently works, even when the pixels are fighting back.
People in open source intelligence love their tools. Ask any specialist for recommendations and they’ll fire off names like “just use Topaz Gigapixel Pro” or “try any Gyro-based Neural Single Image Deblurring tool — here is a list of all these tools. But the real solution isn’t always in your favorite software; it’s often about admitting you don’t know everything.
I showed this blurred license plate to 50 people from BBC Verify during a session in London. Most of them could easily name three tools that would help deblur it. But here’s the thing — they all didn’t work. What are your options now?
Image: Courtesy of Henk van Ess
My favorite technique in 2025: I put all my failed attempts into an AI chatbot like some kind of digital therapy session — “I tried Topaz, Remini, DeblurGAN v2, ImageJ+ DeconvolutionLab2” — and then watch as it suggests “BeFunky Image Editor.” Seriously? I’d never heard of BeFunky before, but it turns out this free tool with a name that sounds like a rejected Netflix original actually worked better in this case than the $200 fancy software. It’s peak “maybe I don’t know everything” energy, and honestly, that’s when the real breakthroughs happen:
Images: Courtesy of Henk van Ess
That tool never worked that well again, but hey — when I listed it among the tools I’d already tried, I got fresh suggestions. Sometimes the best advice comes from sharing your failures. When you actually can read the text, you need to find context. While researching a red Chevrolet Camaro’s license plate (used by a Dutch criminal), I didn’t have trouble reading the digits — my problem was with reverse image searching. Sometimes Google simply doesn’t recognize a cropped detail of a photo.
Image: Courtesy of Henk van Ess
You can work around this by typing the visible text into Google Images instead of reverse searching the photo itself. This led me to images of tourists in Iran driving the same red Chevrolet Camaro with the same license plate. Ah, the criminal had rented a car (full story here).
Images: Courtesy of Henk van Ess
Here’s a pro tip that sounds almost too stupid to work: If your text isn’t completely potato-quality, just literally ask AI to transcribe it directly. No fancy tools, no image processing wizardry — just upload the thing and say: “What does this say?” My current favorite is Gemini Pro 2.5, which apparently has decided to become the world’s most overqualified proofreader.
Image: Courtesy of Henk van Ess
While you’re still squinting to figure out if that’s an “a” or a sad face, the chatbot has already transcribed AND translated the unreadable text for you:
Image: Courtesy of Henk van Ess
Take a look at this photo. I travel a lot, so I can’t lug monitors around with me. Instead, I use virtual reality glasses to get work done. How many words can you make out in this screenshot that I intentionally blurred as much as possible?
Image: Courtesy of Henk van Ess
While you’re still squinting at the image, I uploaded it to Gemini 2.5 Pro. It managed to read about 170 words from the photo and gave me an accurate summary of what I was actually doing.
Images: Courtesy of Henk van Ess
I was recently working for two weeks in Berlin, and I love giving these little OSINT intro sessions to students about how terrifyingly effective investigative techniques can be. It’s educational, it’s slightly horrifying, and it definitely makes everyone immediately check their privacy settings.
Let’s dissect this photo. The question is : Where is this, and when?
Image: Courtesy of Henk van Ess
The “no bikes” sign probably rules out Malta, Cyprus, Spain, Luxembourg, and the UK and makes Netherlands, Denmark, and Finland highly likely candidates. The reason is simple: “no bikes” signs are most common in countries where cycling is actually a thing. These signs probably don’t exist where nobody bikes in the first place — they exist where so many people bike that you need to actively tell them not to park here. It’s like finding “no swimming” signs at the beach versus finding one in the middle of the Sahara Desert — one is practical public safety, the other is just a mirage. You can see text appearing twice — a word that says “essen” or ends with “Essen” — plus a greenish logo with three, four words. This time, AI doesn’t outsmart us:
Image: Courtesy of Henk van Ess
Can BeFunky come to the rescue once more? It improved the text quality enough that I could read the words “samen redden.” This means the text is in Dutch and says “samen redden [something]” — which translates to “save together [something].”
Image: Courtesy of Henk van Ess
So what’s worth saving? It’s probably a sticker on a window of a shop or restaurant, so it probably won’t say “Save together… communism.” Maybe it says: “Save together capitalism”?
Nah, scratch that — that’s way too far-fetched. It’s probably something uplifting like “Save together… energy” or “Save together… the whales” or just “Save together… on parking.” Or maybe… wait, no, I’m doing that thing again where I overthink everything. Don’t think. Stop guessing. Start searching.
We’re pretty sure “save together” is followed by one or two more words — probably not more than seven characters if the font size matches the first line. Now here’s the fun part: how exactly do you explain this incredibly specific font-analysis-based word count estimate to Google without sounding like a conspiracy theorist who’s had way too much coffee? This is the point where normal search queries meet forensic typography (which we will discuss in part two of this article), and everyone starts questioning your life choices.
How do you tell Google that you don’t know the right words?
Replace the unknowns with a star:
Image: Courtesy of Henk van Ess
Why use quotation marks? Because without it, Google assumes you meant “samen” and “redden” anywhere will do when you really need the words together in that exact order. The quotation marks force Google to find these words as an actual phrase, not scattered across different paragraphs like linguistic confetti. And critically, when you add that asterisks at the end, you’re telling Google:
“These words definitely continue — there’s more text after this, don’t show me results where the sentence just ends here.”
It’s basically saying: “No, I really do mean these two words next to each other, AND I know there’s more coming.” Here is the result:
Image: Courtesy of Henk van Ess
It turns out it’s “together we save food” — which is from the Too Good to Go app. Before we study that new fact, why did we use Google Images? While regular Google is trying to be all sophisticated with its natural language processing, Google Images is like “Oh, you want text in images? I’ve got millions of them, and I’ll show you exactly where these words appear together, including ones you didn’t even know existed.”
It’s the difference between asking a librarian for books about “saving food together” and asking them to show you every photo that has those exact words on it. One gives you articles and think pieces, the other shows you actual store windows and campaign posters. Like this great new tool that shows you every text in StreetView in New York:
Image: Courtesy of Henk van Ess
Back to our case. Too Good to Go is basically the dating app for vegetables and day-old pastries, where restaurants, supermarkets and cafes put up their unsold food for rescue at the last minute.
Image: Courtesy of Henk van Ess
Time to investigate those “SSEN” letters. Next stop: Claude, the semantic analyzer.
I gave it the specifics — Dutch words ending in “essen” on restaurant or shop windows — and it immediately fired back “Delicatessen.” Of course. While I’d been playing detective the actual solution was just asking an AI that specializes in language patterns.
Image: Courtesy of Henk van Ess
Sometimes the simplest approach is just admitting you need a better brain than your own, especially one that actually knows Dutch vocabulary.
I downloaded the Too Good to Go app and searched for delicatessen shops. I wasn’t impressed. Claude explained why:
Image: Courtesy of Henk van Ess
Most of the listings are in Amsterdam, so I started there. Behind the windows there appear to be white and blue boxes.
Image: Courtesy of Henk van Ess
Here’s the fun little trick nobody tells you about — if your investigation hinges on specific colors, just tack them onto the end of your Google Images search like some kind of digital afterthought.
Image: Courtesy of Henk van Ess
It’s like adding sprinkles to your ice cream, except the sprinkles are forensic evidence and the ice cream is a Google search. Delicatessen Amsterdam white blue suddenly becomes this incredibly specific query that cuts through all the generic restaurant listings and gets you straight to the shops with those weirdly specific colored boxes in their windows. Who knew colors could be a search parameter? It’s both genius and obvious at the same time.
Image: Courtesy of Henk van Ess
Our first candidate is Flo’s Deli in Amsterdam.
Image: Courtesy of Henk van Ess
And wouldn’t you know it, it’s a perfect match right off the bat.
What we’ve got here is basically the OSINT equivalent of winning the lottery on your first ticket. See, on the left you’ve got our starting point — some mystery person in a red jacket, sitting outside what could be literally any shop in Europe, with their face helpfully blacked out because apparently privacy still matters. All we had to work with was a “no bikes” sign and some vague color patterns that could’ve been anything from a barber pole to a really ambitious tic-tac-toe board.
The right side shows us the money shot: Flo’s Deli, complete with that “samen redden we eten” we’ve been obsessing over, the exact same blue and white striped window situation, and that “no bikes” sign sitting there like a proud little beacon saying: “Yep, this is definitely the place!”
After bouncing around between different AI analysis tools like some kind of digital pinball — with them helpfully determining that yes, this photo was probably taken in spring or fall (thanks for that groundbreaking seasonal detective work, robots) — we’re wrapping up this first installment of “Turning Unreadable Text into Evidence” with the most beautiful, time-tested classic tool in the OSINT arsenal: the Time Machine in Google Maps.
Image: Courtesy of Henk van Ess
In May 2022, the shop wasn’t open yet. So the time frame must be somewhere between now and May 2022, right?
Image: Courtesy of Henk van Ess
Well, if you have plenty of time left — and apparently nothing better to do with your life — you can even pinpoint it to a more specific period by looking at thousands of tourists’ photos of the shop. Because obviously that’s a completely normal way to spend your afternoon.
Those white and blue boxes? They contain bagels.
Image: Courtesy of Henk van Ess
Which, in the grand scheme of investigative breakthroughs, is about as earth-shattering as discovering that ice is cold. But here’s where it gets delightfully ridiculous: by looking at other contextual photos and tracking the normal seasonal rise and fall of bagel box displays, you could actually pinpoint this photo to autumn 2024.
So what have we learned? That advanced forensic techniques sometimes mean asking an AI chatbot for help, that a $200 tool might be beaten by something called ‘BeFunky,’ and that analyzing Amsterdam bagel patterns is somehow legitimate detective work now.
But here’s the real takeaway: the secret to making unreadable text readable isn’t in any single tool. It’s about having a system — scrutinize each detail, combine multiple approaches, admit when you’re stuck, and ask for help (even from AI).
Next time: we’ll share more of Bellingcat member Timmi Allen’s techniques from our Chinese credit card investigation, including how to recover some financial insights from impossibly blurry video footage. Because apparently, that’s just a typical Sunday in the world of digital forensics.
Editor’s Note: This article first appeared in Digital Digging, Henk van Ess’s newsletter on the Substack platform. It has been lightly edited and is reprinted here with permission.
Dutch-born Henk van Ess teaches, talks, and writes about open source intelligence with the help of the web and AI. The veteran guest lecturer and trainer travels around the world doing internet research workshops. His projects include Digital Digging (AI & research), Fact-Checking the Web, Handbook Datajournalism (free download), and speaking as a social media and web research specialist.
2025-05-28 22:27:34
In 2020, Liberty launched an exciting new project using the tools of investigative journalism to expose and challenge abuses of power and violations of human rights: Liberty Investigates. You can find the investigative unit’s work at libertyinvestigates.org.uk. Our two current key topics are policing and protest rights, while we have previously also specialised in migrant and asylum seeker rights.
We believe rigorously pursued, collaborative investigative journalism can be instrumental in challenging abuses of power, and we believe it’s a time when it’s needed more than ever. The team undertakes investigations designed to hold power accountable, change narratives and spark positive change in human rights.
Our small team has worked with publishing partners including the Times, the Guardian, the Observer, Independent, Channel 4 News, ITV News, and local press. We’ve been shortlisted for the Private Eye Paul Foot Award twice, were finalists two years running for the Orwell Prize for Exposing Britain’s Social Evils, and were recently shortlisted for the European Press Prize.
We are looking for an Investigations Editor with the investigative, strategic and management skills to lead this exciting work into the future.
You will need to understand the power of telling stories and the role that journalism plays in mobilising action in the public interest. You will need a passion for upholding high factual and ethical standards. You will be excited about working in a multi-disciplinary campaigning organization, and you will inspire the team to carry out investigations to spark positive change.
At Liberty we are striving to build a team that is truly inclusive – we understand that as an organisation we can only work at our best when we have a diverse workforce sharing a wealth of ideas and experiences. We therefore encourage applications from marginalised groups, particularly people of colour, trans and non-binary people and disabled people. Liberty supports hybrid working, with a minimum of two days per week in the Westminster office.
2025-05-28 22:22:53
The International Fund for Public Interest Media (the “International Fund”) is seeking a dynamic and strategic individual to join its ‘What Works’ Unit in a part time position, equivalent to 80% FTE (four working days per week). The ideal candidate will have a solid background in monitoring, evaluation, and learning (MEL), with strong project management skills essential for leading MEL processes across complex programs. They will be responsible for ensuring rigorous implementation of MEL frameworks by grantees, tracking impact, and communicating results clearly to internal and external stakeholders. Exceptional writing skills and meticulous attention to detail are critical, as the role requires drafting high-quality reports, donor communications, and strategic insights. Experience working in media, media development, or within multilateral organizations is a plus. The candidate should ideally also bring an understanding of media markets in Eastern Partnership countries (Armenia, Azerbaijan, Belarus, Georgia, Moldova, and Ukraine).
2025-05-28 22:14:33
ProPublica is looking for a motivated early-career investigative data journalist to join our editorial team as the David Burnham-TRAC Data Reporting Fellow, specifically focused on investigating the federal bureaucracy. This is a new fellowship, meant to provide up-and-coming data journalists the opportunity to learn from some of the best data and investigative reporters in the business and to give them the space to shed light on both the inner workings of the government and impacts of federal policy.
The fellowship honors Burnham, an investigative journalist who reported on local, state and federal enforcement corruption for 50 years. He also helped found the Transactional Records Access Clearinghouse, TRAC, a data gathering, research and distribution organization at Syracuse University.
ProPublica’s investigative journalism does more than expose wrongdoing and injustice; we intend for it to spark real-world change. You’ll be part of ProPublica’s award-winning data team, a group of nearly a dozen journalists who use data analysis to uncover scoops, stories and deep dives.
The David Burnham-TRAC Data Reporting Fellow will report to ProPublica’s senior editor for data and news applications and work with members of our data team to produce or support at least one major deep-dive investigative project related to the federal government for each year of the fellowship.
