About the author & the site
Ian Bann is an AI-driven SEO strategist focused on diagnosing how authority and visibility break inside AI-generated search results.
As an SEO specialist, I chose to focus on ai visibiity research because the shift from rankings to generated answers changes how visibility is decided. I kept seeing strong pages missing from answers while weaker sources were being used, and that gap was not explained by traffic or traditional visibility.
That is what led me to start comparing answers against original content and analysing what actually gets included, reused, or left out.
Process
Every answer is built through a series of steps. Changes at any stage affect what gets shown, what gets removed, and which sources are used in ai search
01
This is where everything starts. The system pulls from multiple pages, articles, and references, which means the clarity and consistency of the original source matter before the answer is even built.
Breakdown
When I look at an AI-generated answer, I'm not just reading it to see if it's right or wrong. I'm trying to understand how it fits into the AI visibility framework, because that process tells me far more than the answer itself.
I usually start with the final output and then work backwards from there. I look at what information made it into the answer, what was left out, and how those pieces were combined into something that reads clean and simple on the surface.
Once you do that a few times, you start to realise that the system is not just pulling information, it is reshaping it. The next step is comparing that answer directly to the original content it likely came from, because that is where the gaps start to open up.
You can see where sections have been shortened, where context has been removed, and where the meaning has shifted slightly because of how the answer was written. These are not always big changes, but they add up quickly, and over time they explain why some sources stay visible while others disappear.
Deconstruction
I follow the answer from the original page to the final output and check where the source starts losing detail, connection, or visibility.
01
I start with the original page and read it properly before I look at the answer. I need to see the exact wording, the full explanation, and the context around the point being made. If I skip that step, I can't tell whether the answer stayed true to the source or just pulled out the easiest parts.
Background
This didn't start from a theory or a framework. It started from seeing things that didn't line up. I kept noticing that strong websites with clear expertise were missing from AI-generated answers, while weaker or less reliable sources were being included.
At first, it looked like something random, but once it kept happening across different queries, it stopped feeling random. That is when I started checking the answers against the actual content behind them.
Instead of looking at rankings or traffic, I started looking at what the system was actually using and how it was using it. What stood out very quickly was that the system was not always choosing the best source. It was choosing the source that was easiest to extract from and rebuild into an answer.
That changes everything, because it means visibility is not just about how strong a page is, but how usable it is for the system. Once that clicked, the focus shifted completely toward the Research view of how attribution patterns appear across real AI-generated answers and why certain sources make it through that process.
Approach
I focus on how answers are built, because that is where visibility is actually decided. A lot of SEO work still focuses on rankings and traffic, which makes sense in a traditional search setup, but that does not explain what happens inside AI-generated answers.
To understand that properly, I look at what happens before a user clicks anything. I look at how the system picks information in About, and how it puts it together into something that looks simple on the surface.
Each one of those steps can change whether a source is included or ignored. If the content is hard to break down, it is less likely to be used. If the system cannot clearly connect it to a topic or an entity, it becomes less reliable from the system's point of view.
When you look at it this way, you start to see that visibility is not just about being the best source. It is about being the easiest source for the system to understand and reuse without breaking the answer, and that is where most of the real problems begin.
Signals
I don't look at visibility as one signal. I look at the set of things that make a source easier, harder, stronger, or weaker for the system to use.
01
I look at whether the content is laid out in a way that is easy for the system to break down and reuse. If the structure is weak, the system has to work harder, and visibility usually drops.
I run the same type of queries through ai tools in slightly different ways to see what changes in the output.
Even small wording changes can affect which sources are selected and how the answer is written, which tells me how sensitive the system is to input.
I take the generated answer and compare it directly to the original content.
This makes it clear what has been carried through, what has been dropped, and where the meaning has shifted.
I look at how the system connects ideas, how it handles names and brands, and how it decides what is important enough to include.
This is usually where the real issues start to show up.
I run the same checks across different prompts and different systems.
If the same behaviour keeps appearing, it means there is a pattern behind it and not just a one-off result.
What I'm really tracking is how information moves from a page into an answer, because that movement is what decides whether something is visible or not.
I look at how content is broken down and reused inside the answer.
Important details often get lost at this stage, which changes how the information comes across.
I check how clearly the system recognises people, brands, and topics.
If this is not clear, the system struggles to use the content properly.
I look at how the system compares different sources to decide what it trusts.
It rarely relies on one source, so consistency across multiple places becomes important.
I track when a source is clearly cited, when it is mentioned without being credited properly, and when it does not appear at all.
This has a direct impact on visibility.
Perspective
The biggest difference is where visibility is decided and what the system is actually evaluating before the user sees anything.
Traditional SEO focuses on where a page appears in search results. I am looking at whether a source makes it into the answer at all, because that is where visibility is decided much earlier.
Process
I break the work into parts first, because it is much easier to see what is actually causing the issue when each part is checked on its own before everything is connected back together.
01
I start by separating the problem into smaller parts so I can see what is actually causing the issue. Instead of treating visibility as one big problem, I break it into content use, entity clarity, attribution, and trust signals. That makes it much easier to see where things start to go wrong.
Ian Bann
If you have read this far, you already know how I think about answers, what I compare them to, and why I do not treat AI visibility as the same problem as chasing rankings on a results page.
I publish research articles on this site in public because repeating the same checks in different places is often what turns a one-off observation into something you can trust. If something here matches what you are seeing in your own testing, that is usually a good sign we are looking at the same underlying behaviour. If you want more information, visit the contact page.