Check my name
What AI assistants say

Why ChatGPT and Gemini give different answers about the same person

The four assistants are not four windows onto one answer. They are four different habits about what counts as worth relying on, and the spread between them is wide.

The measurement

Same person. Same three questions. Same day. Four assistants.

DISTINCT SOURCES BEHIND THE ANSWERS, ONE PERSON, ONE DAY Perplexity 22 ChatGPT 4 Claude 4 Gemini 1 and it was an internal redirect, not a publication all four named him on all three questions
Distinct sources behind the answers, by engine. All four named him every time, so this gap is not about whether they knew who he was.

One engine drew on twenty two distinct hosts. One drew on four. One drew on four. One drew on a single source, and that source was an internal redirect belonging to the company that makes the assistant rather than a publication anybody could read.

Why this matters more than the wording

People tend to judge these answers on tone: does it sound right, is it generous enough. That is the least useful thing on the screen, because the wording changes between runs and the sources are the part you can act on.

An answer built on twenty two sources and an answer built on one can read almost identically. One of them is resting on a great deal and the other is resting on very little, and only the citation list tells you which.

Two answers can sound equally confident while one is built on twenty two sources and the other on one.

So checking one assistant proves almost nothing

If you ask one assistant about yourself and the answer is fine, you have learned that one engine, on one day, was content. You have not learned what the other three say, and on our own measurement they differ enormously.

The same applies in reverse. An assistant that declines to answer, or answers thinly, is not evidence that nothing about you is available. It is evidence about that engine.

Our own check records an engine that could not be reached as a separate result from an engine that answered and did not name you. Those are different findings and collapsing them is how people get told the wrong thing. How to check it properly sets out the method.

What the gap does not tell you

We are not going to rank the four engines, and a single person's reading is not a basis for doing so. Which engine leans on most sources will vary by subject, by question and over time.

What is stable enough to act on is the shape: the four do not agree, so a reading worth having covers all of them, and the part worth fixing is the sources rather than the sentences.

What this page is based on

24 September 2026, our own founder, three questions to each of four engines. All four named him on all three questions. Distinct sources behind the answers: 22, 4, 4 and 1.

Keep reading

What AI assistants say

See what the four assistants say about you

We read your first page of results and put real questions to ChatGPT, Gemini, Perplexity and Claude, then send you what each one actually said and which sources it leaned on. Free, under five minutes.

Run my free AI audit