AI Battle: ChatGPT vs Claude vs Gemini 🤖

Links and Codes:
Paperpal: https://paperpal.com/?linkId=lp_726731&sourceId=andy&tenantId=paperpal (PAP20 - 20% off)
Thesify: https://www.thesify.ai/?via=andrew
Thesis AI: https://www.thesisai.io/?via=andrew, ANDY20 - 20% off
Elicit: https://elicit.com/?via=andrew
SciSpace: https://bit.ly/AndyStapleton (ANDY20 — offers 20% off on monthly plan)
Jenni AI: https://jenni.ai/?via=andy-stapleton (Use codes: andy30, ANDY20)
Julius AI: https://julius.ai/?via=andrew-stapleton (ANDY20 — offers 20% off)
AnswerThis: https://answerthis.io?ref=andy49 (ANDY20 - 20% off)

I've spent the last few weeks putting the most popular AI models through their paces, and what I discovered about chatgpt, gemini ai, and claude ai honestly surprised me. My team and I didn't just casually test these platforms - we ran systematic experiments to see which ai tools for research actually deliver accurate, trustworthy information versus which ones are essentially making things up.

▼ ▽ Sign up for my FREE newsletter
Join 21,000+ email subscribers receiving the free tools and academic tips directly from me:
https://academiainsider.com/newsletter/

▼ ▽ MY TOP SELLING COURSE ▼ ▽
▶ Become a Master Academic Writer With AI using my course: https://academy.academiainsider.com/courses/ai-writing-course

................................................

▼ ▽ TIMESTAMPS
00:00 Intro
00:09 What we actually did
01:13 Models Tested
01:54 Example Prompts
02:15 1st Order Hallucinations Results
04:57 2nd Order Hallucinations Results
06:56 1st and 2nd Order Hallucinations Means...
07:55 Take Home Message
09:49 Elicit
10:07 Scispace
10:24 Consensus
11:00 Outro

Here's what most people don't realize: when you ask an AI to cite sources for academic work, there are actually two levels where things can go wrong. The first is obvious - does the reference even exist? But the second level is far more insidious and takes real effort to verify. A citation might be completely real, pointing to an actual published paper, but the AI could be citing it for entirely the wrong reasons. The paper might not actually support the claim being made at all.

This matters enormously for anyone doing serious research. I tested chatgpt across multiple versions, including the new thinking models with web search enabled. I put claude ai through its paces with different configurations. And I ran gemini ai through the same rigorous testing protocol. The results were stark - some of these ai tools performed shockingly well, while others failed spectacularly in ways that could genuinely damage your academic credibility.

What fascinated me most was how paying more money didn't automatically mean better results. Some premium features actually performed worse than their free counterparts. I also discovered that these best ai tools have very specific strengths and weaknesses. One might excel at finding real papers but struggle with accuracy. Another might be brilliant at one task but completely unreliable at another.

Throughout this video, I break down the actual data, show you the specific failure modes, and explain which platforms you should trust for different research tasks. I'm not here to tell you to avoid AI entirely - I'm here to help you use it intelligently and safely.