What Caused the "Madness of King George?" Researchers Used AI to Find Out
For decades, it was believed King George III had the rare genetic disorder porphyria. Then researchers used AI to claim differently.

Grim Reminder: A Grim Historian is a reader-supported newsletter and depends on your 5$ donations to keep the most depressing politics and history in your inbox. Please consider becoming a paid subscriber.
If you are a Bridgerton fan, don't expect to get your petticoats in a romantic rustle over the show's new spinoff, Queen Charlotte: A Bridgerton Story. Instead, picture a weightier One Flew Over the Cuckoo's Nest, set in Georgian London with much hotter actors.
The story begins with the usual starry-eyed hijinx. In 1761, the German princess Charlotte of Mecklenburg-Strelitz must marry King George III. (That's the guy who made us so mad we dumped out his tea and proclaimed, "No taxation without representation!" Then a revolution followed, and representation excluded Black people and women.)
History lesson aside, Queen Charlotte and King George's chemistry sizzles. Unfortunately, George refuses to consummate the marriage and hides in his palace in Kew.
What's a horny new bride to do?
Patience is called for as it is revealed George is battling some blue devils. Unfortunately, his mental health struggles prevent him from performing his husbandly duties. Queen Charlotte's solution is shagfests on "even days" only so George can rest and stargaze on odd days. (Note: He really did build an observatory to gaze at stars, but the sex problems are fictional. George and Charlotte were busy enough to have 15 children.)
His doctor's solution is to torture him with arsenic powder, leeches, purgatives, straight jackets, beatings, and ice baths.
Sadly, George's mental illness was not creative nonfiction. George suffered his first bout of hypomania at 50 and slowly declined until he died at 82.
His appalling medical treatment was also accurate. When George got frisky, his doctors pretty much beat the snot out of him.
However, what remains a question today is what caused King George's illness.
The porphyria theory
In 1969, psychiatrists Ida Macalpine and Richard Hunter analyzed George's medical records to compile a list of symptoms — abdominal pain, nausea, vomiting, jaundice, constipation, neuropathies, and psychiatric symptoms. His urine was also bluish and sometimes dark red.
From these symptoms, Macalpine and Hunter concluded that George had the rare metabolic disorder porphyria. Porphyria is a group of disorders caused by abnormalities in heme production. Heme is a key molecule for most living cells. Its primary function is to bind and transport gases, such as oxygen, throughout an organism. (Thus, why your urine turns the color of Fruitloops.)
Then they dug deeper. Since porphyria is a genetic disease, they examined the medical records of George's descendants.
The cases of suspected porphyria in George's family tree were numerous. George III's two greats granddaughters, Victoria Adelaide Mary Louisa (1840–1941) and her daughter Princess Charlotte of Prussia (1860–1919), were suspected of suffering from the condition. In letters to her physicians, Charlotte described debilitating abdominal pain, muscle weakness, blisters all over her face, and dark red urine — all symptoms of porphyria.
In the 1990s, historians got permission to exhume Princess Charlotte and her daughter, Princess Feodora of Saxe-Meiningen (1879–1945). DNA testing on their remains found both had a mutation consistent with porphyria. And although a mutation does not mean they were afflicted with porphyria, it increases their probability.
If there is one piece of evidence that doesn't lie, it is DNA. Although we cannot assume George suffered from the same mutation as Charlotte and Feodora, we know that porphyria was in the genetic mix. (Incidentally, historians asked to dig up King George and conduct genetic testing. You can bet the Queen said nay.)
A more recent case of porphyria in the family was Queen Elizabeth II's first cousin, Prince William of Gloucester (1941–1972). Prince William died at 31 in a plane crash but was reliably diagnosed with variegate porphyria before his death.
Despite the genetic evidence, George’s diagnosis was not a slam dunk. So historians tested a sample of his hair and found the arsenic levels high enough to embalm a horse. Arsenic was part of George's medical treatment and may have also caused his erratic behavior. Arsenic is also a known inhibitor in heme iron, so it would have worsened porphyria.
Of course, diagnosing a patient without the patient gets tricky. Consequently, many historians accused Macalpine and Hunter of cherry-picking George's medical record for data that supported their conclusion.
In 2013, psychiatrists Peter Garrard and Vassiliki Rentoumi from the University of London came to a different diagnosis. And they came to this diagnosis with the help of machine learning.
AI rewrites history
Instead of relying on unreliable medical records, genetic predisposition, and salacious rumors, Garrard and Rentoumi programmed a computer to "read" George's letters from over his 60-year reign (1760–1820). Their artificial history detective found a pattern— George's writing style changed as his illness worsened.
Machine learning diagnosed George by comparing people with mental illnesses with those without. These differences include sentence length and complexity, vocabulary nuances, and the frequency of words.
From AI's language analysis, they concluded that George suffered from "acute mania," — an excitable, hyperactive condition common in the manic phase of bipolar disorder.
If that diagnosis sounds like a straw dog argument, you are not alone. Several historians have balked at the bipolar diagnosis.
I don't doubt that George's letters got whacky as he sickened. (He did shake hands with trees and run around the palace naked.) But since porphyria's symptoms include anxiety, restlessness, insomnia, confusion, paranoia, and hallucinations, it is impossible to separate symptoms from causality. Obviously, many diseases can be the underlying cause of mental illness.
To be fair, Garrard and Rentoumi didn't base their diagnosis solely on AI's analysis of George’s letters. Several eyewitnesses described George’s hypersexualized behavior, extreme mood swings, and speaking without stopping — common symptoms of bipolar disorder.
They also argued that his medical treatments could have caused his odd-colored urine. However, urine color is not the most prominent symptom of porphyria. Stomach pain is. And according to his doctor's medical records, George had his fair share of stomach woes.
Unfortunately, we will never know what caused the "madness" of King George. But as AI advances, historians may wrestle with a thornier problem — how much should we rely on AI to analyze the past?
Already, I have seen the claim that "King George III suffered from bipolar disorder" repeated across the internet as a case-closed conclusion by armchair historians. Not everyone agrees.
Will AI become the new medical detective?
King George isn’t the only example of using AI to play medical detective. In 2004, machine learning analyzed Iris Murdoch’s last novel and found her language revealed the early signs of Alzheimer’s disease. AI was correct.
In 2012, medical historians used AI to diagnose another famous king with mental illness — King James VI and I. King James was the great-great-great-great grandfather of George III and was also suspected of suffering from porphyria. AI disagreed. The computerized diagnostic aid SimulConsult found an even rarer genetic disorder that causes mental illness — Lesch-Nyhan syndrome. The jury is still out on whether we need to rewrite the history books, but it might explain why King James sanctioned so many witch trials.
One thing is for certain. Machine learning can analyze diseases across populations, especially admixtures.
But on an individual level, its generalizations can misdiagnose illnesses. To start, language and even our handwriting change as we age, so establishing a baseline for mental acuity gets dubious.
The same applies to history. While machine learning can synthesize copious historical documents, allowing us to see a broader picture of society, it also can introduce biases and outright falsifications. These falsifications are demurely referred to as "hallucinations." In layperson's terms, a hallucination is when AI makes shit up and tries to convince a human it is true. It's basically the artificial version of gaslighting.
This design flaw also begs a more salient question — if AI hallucinates, should we use it to analyze the hallucinations of others? Perhaps someday we will if our "mad" kings become artificial ones.
As Winston Churchill knew all too well, "History is written by the victors."
Carlyn Beccia is an award-winning author and illustrator of 13 books. Subscribe to Conversations with Carlyn for free content every Wednesday, or become a paid subscriber to get the juicy stuff on Sundays.
I think you are in error in casting James VI and I as George III’s direct forbear.
When Queen Anne died with no surviving children, that was the end of the Stuart line. Anne’s second cousin George, the Elector of Hanover, acceded to the throne as George I.
This discussion is fascinating, especially since I saw “The Madness of King George,” where all the dreadful medical treatments were shown. Poor guy!