Don’t Trust Your Gut (Until You’ve Tested It): Assumptions in Image Analysis and Geolocation

As an OSINT instructor, one of the areas I enjoy teaching most is image analysis and geolocation. I decided to write this blog after noticing a common pattern during geolocation activities in training sessions: students would instinctively assume where particular images were taken.

For some, that assumption would then shape the direction of their investigation. For those who didn't verify their assumptions, students often found themselves on the wrong track, running out of time to complete the task successfully. In contrast, those who took the time to validate their assumptions and were open to being wrong were consistently more successful.

Assumptions focus your attention, direct your lines of inquiry, and can help you decide what to verify first. When investigating an unfamiliar image, you might initially infer a region from its architecture, vegetation, or signage. These assumptions are shaped by our own personal experiences and inherent biases, something we will discuss later.

Assumptions are useful, but put simply, they are a starting point.

As many of us know, generative artificial intelligence (AI) can quickly suggest locations based on a single image. While impressive, it's tempting, but risky, to treat the answer as fact. AI is not infallible. AI may miss contextual clues, misinterpret regional features, carry its own geographical biases from training data or confidently suggest the wrong location. Analysts must know how to evaluate, verify, and challenge AI-generated responses.

That is why we want to focus on testing our assumptions. Noting our biases and AI’s fallibility, we’ll explore how to responsibly make and test assumptions in image geolocation, turning an assumption into something tangible. As this blog will focus on the geolocation of images, we strongly encourage you to read our blog “A Geolocation Walkthrough” prior to this one.

Why Make Assumptions?

When faced with limited or ambiguous information we make assumptions because they help us begin making sense of limited information. They help us generate hypotheses, narrow the field of possibilities, and determine where to focus our efforts.

This concept is directly applicable to OSINT image analysis. Our assumptions if left untested can shape – correctly or incorrectly – the direction of your investigation. But by treating them as working hypotheses, we allow for verification, revision, and accuracy.

We naturally make assumptions because:

• The human brain is wired for pattern recognition. From the shape of rooftops to the type of trees or traffic signage, we instinctively associate visual cues with places we’ve seen before.

• Time is limited. Investigators rarely have the luxury of endless hours or perfect data. Assumptions allow you to start somewhere.

• Assumptions guide verification. Every image tells you a partial story. Making an assumption, such as “this is likely in South Asia” based on vegetation and building materials, gives you somewhere to start.

We can’t avoid assumptions, but make them consciously, state them clearly, and be ready to discard them when the data says otherwise.

Types of Assumptions

When analysing an image, assumptions typically fall into two overlapping categories: broad, ‘gut feel’ impressions and specific, testable object-based assumptions. Each helps you advance your investigation of the image and, ultimately, determine its likely location.

• Broad Assumptions – ‘Your Gut Feel’ - This is the first impressions you get from the overall feel of the image—its atmosphere, colours, light quality, and general layout. These assumptions are typically subconscious and driven by pattern recognition, like “This feels like the southern hemisphere” or “This looks like a European winter.”

• Specific Assumptions - These focus on individual elements you can observe, such as "That license plate looks European." Unlike gut feel, these assumptions are based on specific visual evidence and are easier to verify or disprove systematically.

Let’s take an example: You are tasked with geolocating the image below. You will intuitively have a broad ‘gut feel’ assumption about where this image may be located. Those who are from this region may recognise it immediately. Others with family or friends nearby may also have a sense of where it might be. Some of you may have travelled or known people from the region and could therefore make a solid assumption. Those of us who have no direct connection to the region can still make a gut feel assumption based on other past experiences such as exposure to pop culture through movies or television. What’s important is that you make an initial assumption for your investigation to proceed even if it turns out to be inaccurate. Incorrect assumptions can be just as valuable as a correct one, as they help guide the process of verification.

Where is this image?

Without conducting a full investigation of this image, what is your assumption about where this image was taken? What is your gut feel? What specifically about the image gives you that gut feel (don’t use AI – we want you to practise identifying your bias)?

Before we dive into a systematic process, it's crucial to understand what can go wrong with our assumptions.

Managing Bias

Assumptions are vulnerable to cognitive bias. When left unchecked, bias can mislead analysis, reinforce false leads, or cause you to overlook important details. Being aware of bias and using structured techniques to counter it leads to more accurate, disciplined assessments. So, what are some of the cognitive biases to be aware of?

Just some of the bias we need to be aware of.

Bias isn’t something you can eliminate entirely; however, you can manage its influence with structure. Start by recognising your bias, and also consider:

• Recording your assumptions, revisit and be willing challenge them.

• Considering alternative explanations before narrowing your search.

• Inviting contradictory viewpoints—a second pair of eyes can challenge entrenched thinking.

  
  

A Process for Making and Challenging Assumptions

Whether you're working with AI-generated suggestions or your own gut feel, use this systematic approach.

1. Make and document your assumption explicitly. This is important, and we encourage you to be okay with being wrong!

   
   

2. Identify all object-level clues. Identify those objects/things that would prove you wrong. Once you've taken in the broader image, focus your attention on specific elements that can turn your rough guess into something searchable. Look for objects that could contradict your initial assumption, rather than attempting to confirm what you ‘think’.

   
   

3. Verify systematically. This is your classic OSINT skill set e.g., an identified business is not located in Australia.

   
   

4. Decide. Pursue or abandon your assumption.

Applying the Process: A Worked Example

My assumption is that the image was taken somewhere in Northern Australia - tropical vegetation, informal settlement style, and general feel. It looks familiar.

Now that an initial assumption has been made, naturally we would take one of two approaches:

1. Look for information that verifies my assumption.

2. Look for information that contradicts it. This approach is challenging as we naturally prefer to be correct.

So, following the process, we must now identify and list object-level clues - things to prove myself wrong. In this case, I identified the objects that are easiest to verify: (1) the license plate format and (2) the bollard and tyre chain fence style. You might decide to focus on the shop front, the road signs, traffic, vegetation or something else.

Moving into verification, I want to look at the license plates from the Northern Territory and Queensland. Then I looked to try and find examples of the fence style. The license plates are different and there is no evidence of this fence style occurring in Australia, especially in this setting.

The final step is deciding to pursue or abandon the assumption. In this case the assumption was wrong, but I have been able to move forward and rule out a particular region. I can ask myself where else might be similar. Where might have similar architecture, vehicles, signage or vegetation? Where has Australia had this type of influence? Could it be New Zealand, the Pacific Islands, or parts of Southeast Asia?

When looking into the licence plates, I found information pointing towards the Pacific, so that is where I’ll head next. This is now my new assumption (because it is yet to be confirmed) and using the same things e.g., license plates and fence style, I will try to verify.

At this point, you might be thinking that this is overkill. But what we are trying to do is build a systematic, repeatable methodology that helps prevent bias from leading us astray.

You may note that we haven’t geolocated the image. Attempt to geolocate the image yourself and keep an eye on our socials for the correct location to see if you get this one right!

Final Thoughts

Making an assumption is a key part of any OSINT geolocation investigation. It gives you a starting point. Even if it’s off the mark, it’s still progress. Being wrong helps you rule things out and sharpen your focus—often faster than trying to analyse everything at once. The goal isn’t to be right straight away, it’s to move forward. A good assumption sets the wheels in motion and gives you something to test, challenge, or build on. Without it, you’re stuck looking at an image with no direction.

To support your OSINT collection and analytical capabilities, contact us at info@osintcombine.com to learn more about NexusXplore, our all-in-one, investigation-agnostic software platform, or our various training courses including Image Analysis.