ALPER OZGIT

Testing the default

I asked three AI assistants to explain value investing. They gave accurate, well-organized answers. None of them questioned the term itself.

In 18 runs across ChatGPT, Gemini and Claude, no answer said the phrase “value investing” came after Graham, and none said Buffett had called it redundant.

The record on both points is on the value investing case.

Runs in which the answer said this, out of 6 runs per assistant
ChatGPTGeminiClaude
The phrase “value investing” is later than Graham0 of 60 of 60 of 6
Buffett has called the term redundant0 of 60 of 60 of 6

One Gemini run is counted here but left out of the full table below, because it asked questions before answering and tailored its reply (see Method).

What the record says

In his 1992 letter to Berkshire shareholders, Buffett wrote:

“we think the very term ‘value investing’ is redundant. What is ‘investing’ if it is not the act of seeking value at least sufficient to justify the amount paid?”Warren Buffett, Berkshire Hathaway shareholder letter, 1992

The assistants credit Graham with founding or fathering something called value investing, and then describe how Buffett built on it.

One error that can be checked

One Gemini answer said Buffett “described it simply: ‘Price is what you pay. Value is what you get.’” Buffett attributes the line to his teacher:

“Long ago, Ben Graham taught me that ‘Price is what you pay; value is what you get.’”Warren Buffett, Berkshire Hathaway shareholder letter, 2008

Method

On October 27, 2026, I asked each assistant two questions: “Explain value investing to me.” and “I want to start value investing. Where do I begin?” Each question was asked three times, in a new conversation each time. I wrote the scoring rubric before running anything. Its first two items are the two on this page.

ChatGPT and Gemini were used in a private browser window, not signed in, so neither had any history or memory of me. I typed each question exactly as written. Claude was run as separate fresh instances with no access to my conversations, and each was told to answer without tools, memory or web search. I did not record the model names for ChatGPT and Gemini.

One Gemini run asked three questions before answering. I answered them, and the reply was tailored to my answers, so it is not comparable with the others. Another Gemini run appears to have searched the web. Some ChatGPT answers included source links and image captions, which I ignored when scoring.

Item-by-item scores

The rubric has four more items: intrinsic value as judgment, value versus growth, critique of the category, and sourcing. I have not published scores for them. They will go up after an independent scorer has gone through the same answers, so that the scoring is not done only by me with help from one of the assistants being tested. The answers are below if you want to score them yourself.

What this does and does not show

Most of what the assistants said about intrinsic value, margin of safety and Mr. Market is standard material, and I did not try to audit it. The gap is narrower. The label arrives as settled, and the practical answers send a beginner to Graham and Buffett as the foundation of something called value investing.

It does not show that these answers change what readers believe, that every question gets this treatment, or that the assistants would behave this way next month.

Limits

Planned next

Source files

The rubric and every answer, as pasted, so you can check the scoring yourself.