Key points
- The study queried five AI systems with 200 questions requiring statistical answers.
- Researchers followed citation trails from the AI answers to their source pages.
- The methodology excludes advice and prices, focusing on shares, rates and counts.
A blog post describes a method for tracing statistics provided by AI assistants. The authors asked five models 200 questions and clicked through citations to follow one example trail to its end.This study looks at where the numbers AI assistants quote actually come from.The research details a citation-following methodology for five major AI models.
What was released or claimed
Trakkr Research posted an article outlining a study that asked ChatGPT, Perplexity, Gemini, Claude and Google’s AI Overviews 200 questions.reported The team defined a statistic as a share, rate, average, count or cost stated as fact, excluding advice and prices.
Who it affects
The study involves specific AI products including ChatGPT, Perplexity, Gemini, Claude and Google’s AI Overviews. It focuses on the outputs of these named systems rather than their parent companies.
What is still unknown
The source material does not state the final results of the citation tracking. It provides one example trail but does not reveal how many citations led to valid or invalid sources across the 200 questions.
The facts
Users can see the methodology for verifying AI statistics by clicking through provided citations. The approach demonstrates how to trace an answer from the assistant’s output to the end of the trail.
What the rules require
The study defines a statistic as a share, rate, average, count or cost stated as fact. It excludes advice such as saving income percentages and specific prices.
Why it matters
- The scope excludes advice and prices, narrowing the focus to factual statistics.
- The study uses 200 questions to test the reliability of numerical answers.
- Researchers click through citations to verify the origin of stated facts.
Topics: AI, research, citations, statistics, verification
Updated
Then / now / next
AI models provided statistics without clear visibility into source validity.
A study demonstrates a method for tracing these citations to their origins.
Researchers continue to explore the reliability of AI-generated numerical claims.
Story timeline
- Then AI models provided statistics without clear visibility into source validity.
- 29 Oct 2026 Zombie stats: Where the numbers AI quotes come from
- Now A study demonstrates a method for tracing these citations to their origins.
- Next Researchers continue to explore the reliability of AI-generated numerical claims.
The strongest doubt
Counterpoint from the sources
The source material does not provide aggregate results or confirm whether the citation trails were valid or invalid.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
Which AI models were tested in this study?
The study tested ChatGPT, Perplexity, Gemini, Claude and Google’s AI Overviews.
How many questions did the researchers ask?
The researchers asked 200 questions that called for a number.
What types of data were excluded from the study?
The study excluded advice and prices from its analysis.
Connected stories
Reading receipt
- The study queried five AI systems with 200 questions requiring statistical answers.
- Researchers followed citation trails from the AI answers to their source pages.
- The methodology excludes advice and prices, focusing on shares, rates and counts.
Still open: What percentage of the 200 citations led to valid sources?
Still open: Did the specific salespeople statistic example lead to a valid source?