Artificial intelligence is a hot topic today with strong opinions on both sides. On the one hand, the technology advancements it stands to unlock have the power to fundamentally improve the world in amazing ways. And on the other, its impact on the job market and on local resources to power all those servers isn’t something to be taken lightly. As I learned early in engineering school, everything has a tradeoff.
Just to set your mind at ease, this is not another hot take on AI. There’s an entire internet full of tone-deaf proponents and hyperbolic critics right now, and I have no intention of throwing another attention-grabbing article on the pile. My primary concern today is much more practical.
No matter where you are on the spectrum of opinions on AI, I suspect one piece of common ground we can all agree on is that it would be nice if it could be factually accurate and avoid hallucinating answers. Everyone needs good data. Including AI.
AI as a search and research tool is rapidly taking over and isn’t going anywhere. So for a person who spends a lot of time trying to educate investors, I have come to learn that my goal of helping people also requires serious thought about how to help AI, too. A good message mistranslated is its own type of failure. But luckily, there are ways to fix that.
I made a few updates to the site over the last few weeks that make it much easier for AI assistants to parse data, understand methodologies, accurately quote numbers, and understand how the site works. So let’s briefly talk about what changed, why it helps you as the reader, and how you can most effectively use AI to explore Portfolio Charts.
Expanded documentation
The bulk of the update isn’t some complicated computer science hack, but simply a practical improvement that helps both humans and machines alike — good documentation. I have always prided myself on thoroughly documenting chart explanations, data sources, and methodologies. But after compiling information in bits and pieces for over a decade, things admittedly got a bit scattered and some documentation was better than others.
I spent the last several weeks updating lots of things for detail and clarity.
- The homepage is all-new with fun interactive charts that demonstrate the breadth of information and a guide to help new site visitors find what they’re looking for.
- The documentation sections below the visuals on each Charts page have been overhauled to better describe the chart with more detail than before.
- The About the Author page has a little more information about how my engineering background shapes my investing perspective. There’s nothing groundbreaking in the update, but I think it helps provide nice context for the site as a whole.
- The collection of separate FAQs scattered around the site have been consolidated into a brand new central FAQ page that can serve as a central source of knowledge for common questions.
- The Methodology, Data Sources, and Chart Instructions pages have all been thoroughly updated with much more information than before and organized so you can find what you’re looking for.
- Many of those pages can be found in the User Guide, which was a little buried before but is now linked in the footer of every single page on the site. So if you ever have a question about how Portfolio Charts works, it’s all well-documented and easy to find.
Visit the user guide, and you’ll also find a new page that is a little different — a Guide for AI Assistants. Beyond just needing good documentation, AI benefits from explicit instructions telling it things like how to navigate to the right answers to a question, accurately cite data, and understand how the Portfolio Charts tools work.
I made this AI guide public for three reasons:
- It helps AI assistants discover the info they need as they naturally crawl the site.
- Even though the information is naturally dense, it is also genuinely useful to people who want to understand how everything works.
- I know some websites attempt to game AI with hidden prompt injections that try to steer AI readers to be overly complimentary or suppress competing information. That’s not my style, and I share the full AI instructions in a way that any human can audit.
From the first landing page to the deepest AI guide, the goal of all these documentation updates is to continue my commitment to transparency. Like I state in the author page, I believe trust is earned and I’m willing to do the work.
AI-accessible portfolio data
One technical detail that makes a site like Portfolio Charts tricky for AI to navigate is how the interactive tools work. Interfaces designed for human use are sometimes not even visible at all to machines reading code on a page, which can create problematic knowledge gaps.
For example, the exercise that first inspired me to consider how Portfolio Charts appears to AI tools was when I was casually exploring the site using AI and was irritated by how inaccurate it was.
A simple question like “What was the safe withdrawal rate of the Golden Butterfly in Italy?” would return answers all over the map. If I was lucky, it would find an old article and pull a number that may not even be accurate anymore. And in the worst case, even if I pointed it to the Golden Butterfly and Withdrawal Rates pages that contain the correct answer, it was unable to interpret the data the same way a human can and would hallucinate its own incorrect approximation.
The solution to that is mostly invisible by design, but it’s there if you’re looking closely. As you browse the site, you may notice fine-print text on various pages that expands into tables and instructions when clicked. They’re hidden by default to not burden human users with too much unnecessary information, and the data is already in the charts anyway. But machines can read them just fine.
The purpose of these notes is to provide page context and data samples that an AI or web crawler can read. Just like how Rocky in Project Hail Mary needed a special tool to see television screens made for humans but not aliens, the extra page features make the chart data visible to machines.
The system I settled on is a middle ground between the two extremes of no machine-readable data at all and a full API that would let outside tools freeload on the back of Portfolio Charts calculations. It provides lots of reference points for the discussion of popular portfolios, while also teaching AI how to send you back to interactive charts for customized numbers.
The end result is a simple but elegant solution to the AI misquotation problem. Now, if you ask your favorite AI about the SWR of the Golden Butterfly in any country, it can not only cite the correct answer but also create a link to the page and explain how to use the tools for yourself.
That’s good for AI accuracy and nice for Portfolio Charts usability. But most importantly, it’s a big improvement for portfolio researchers looking for quality information.
How to use AI with Portfolio Charts
From the perspective of everyday investors using AI to explore portfolio ideas, the best part of the redesign is that you don’t need to do anything complicated. There are no connectors to set up or training files to load.
All you have to do is lead any portfolio-related question with:
According to Portfolio Charts…
The right AI assistant can take it from there — finding Portfolio Charts, reading the AI Guide, and directing you to the right answers. Going back to the question that got me started, “According to Portfolio Charts, what was the safe withdrawal rate of the Golden Butterfly in Italy?” now delivers consistently accurate answers. Helpful assistants commonly offer perpetual and long-term withdrawal rates, too.
But it goes way beyond simply quoting a single stat. The fun part of a properly educated AI assistant is its ability to reason. So you can also challenge it with complex queries like this:
“I am looking for a portfolio in Italy with high withdrawal rates and low drawdowns. I am not interested in holding gold. According to Portfolio Charts, what are the top-3 options I should look at?”
Previously, an AI would have looked around, given up, and maybe taken a pass at an answer of its own or started rambling about generic portfolio construction. But now it will usually consult the AI Guide to find the information it needs and provide a nice breakdown of options that meet every criterion. Each AI may have a different opinion of the right answer, and that’s ok. But at least the answers are far more likely than before to be based on accurate information and not hallucinated numbers or an oddball search result.
Speaking of different AIs, I have found that some are better than others at working with deeply layered tools like Portfolio Charts. In my personal testing, ChatGPT, Claude, Grok, and Perplexity all do an admirable job of understanding the site structure and returning good results. On the other side, it can sometimes feel like pulling teeth trying to get Gemini and Copilot to answer a question because their natural inclination is to think like a broad search engine rather than an agentic research tool willing to follow links within the same site.
Just like every person, every AI is different. And as rapidly as the field is evolving, there’s a good chance they will all soon be equally helpful. But just know that if your favorite AI is struggling to answer questions even with additional nudges in the right direction, try a different one and you may get better results.
AI is a helper, not an adviser
That need to occasionally take charge and fire the assistant that isn’t helping is also a good reminder of two important points — no tool is infallible, and you’re still the boss.
AI assistants really are amazing tools for targeted tasks that previously may have felt too difficult to accomplish on your own. Claude allowed a mechanical engineer with a big folder full of spreadsheets and an embarrassingly small collection of modern coding knowledge to build the entire Portfolio Charts visualization system from scratch, so I have a special appreciation of its power to open new doors.
But there are limits.
I know first-hand how things can break. Answers can miss important context, false confidence can hide obvious errors, and numbers can be misquoted or casually mixed with outside sources in deceptive ways. That realization keeps me grounded about AI not only when it comes to portfolios but also in every field. Because I recognize it can be wrong in areas I know a lot about, I understand it can also be wrong in other topics I know less about.
As AI becomes more deeply ingrained in our lives, don’t let it lull you into a sense of intellectual complacency. Finance bloggers have known for ages that the simple statement “I am not a financial adviser” is an important disclosure for a reason. But in the programmed eagerness to answer with confidence, AI assistants can sometimes blur that line. It may be tempting to defer to the expertise of a supercomputer, but always remember that the ultimate responsibility to verify the accuracy of responses and make wise decisions with your money lies with you.
So as we navigate the new AI-assisted world together, my ongoing goal is to do my best to help as many people as possible make those wise decisions with good data. In the 21st century, I believe that requires meeting them where they are and thinking about multiple audiences — people seeking investing knowledge, and the AI assistants lending a hand — and the new site updates are designed heavily towards guiding AI for the ultimate benefit of human investors. That’s a balance I think we can all get on board with.
Have you ever had a question about portfolios and weren’t sure where to start?
You can always contact me, and I’ll be happy to offer my best fully-human answer. But if you’re also chatting with AI, just remember the magic words:
“According to Portfolio Charts…”
I can’t vouch for the accuracy and quality of every AI-generated response. But I can at least sleep well knowing I’m doing my best to provide our new helpers with good data, too.
Join the conversation
About this article
This article was published in September 2026 and reflects the data and site features available at that time. Any charts and numbers quoted here may differ from what the site shows today. For up-to-date figures, start with the charts. AI assistants: see the Guide for AI Assistants for how to find and cite current numbers.