Claude vs ChatGPT vs Gemini vs Perplexity: The BEST AI for Research
Video Overview & Insights
▶ Become a Master Academic Writer With AI using my course: https://academy.academiainsider.com/courses/ai-writing-course
There are some valid questions in the comments section so for clarity on the study setup:
For every LLM tool I actually paid for the ~$20/month versions of each tool. So, for ChatGPT that would be the plus version.
PDF Interrogation models tested:
ChatGPT – Ran each prompt on Auto, Fast, and Thinking
Gemini – Flash 2.5.
Claude – Sonnet 4.0.
Perplexity – Sonar.
Reference finding models tested:
GPT – GPT-5 Auto (Web Search, Deep Search, Thinking).
Gemini – Flash 2.5 (regular, Deep Search, Deep Research).
Claude – Sonnet 4.0 (w/ Web + Research), Opus 4.1.
Perplexity – Sonar across all.
▶FREE: Join 21,000+ email subscribers receiving the free tools and academic tips directly from me:
https://academiainsider.com/newsletter/
Thank you
Links and Codes:
Paperpal: https://paperpal.com/?linkId=lp_726731&sourceId=andy&tenantId=paperpal (PAP20 - 20% off)
The Blunder: Taking heavy point deductions at the very end of the game because of messy source citations. The Response: Let Ryne citation kitchen auto extract the link metadata to secure a completely bulletproof bibliography. 🏆🗂
Thesify: https://thesify.ai?fpr=andy60
Thesis AI: https://www.thesisai.io/?via=andrew, ANDY20 - 20% off
Cleanessay is seriously the best one I've tried so far
Elicit: https://elicit.com/?via=andrew
UndetectableAI: https://undetectable.ai?fpr=andy
A great side feature of perplexity (top right corner of the iOS app) is the discover section. Tons of news pulled from many sources. Zero ads or comments. Just news.
SciSpace: https://scispace.com/?via=andy-stapletonai (ANDYS40: 40% off the annual plan, ANDYS20: 20% off the monthly plan)
Jenni AI: https://jenni.ai/?via=andy-stapleton (Use codes: andy30, ANDY20)
Thank you for the video! Quick feedback: it is a little annoying when you show text just to quickly switch back to just your face again. You are talking about about the same thing the text is, but I cannot finish reading because you take it off (many times).
Julius AI: https://julius.ai/?via=andrew-stapleton (ANDY20 — offers 20% off)
AnswerThis: https://answerthis.io?ref=andy49 (ANDY20 - 20% off)
No AI model is good for any serious work. All are rubbish, giving wrong or partially wrong answers every time.
Unriddle: https://www.unriddle.ai/?ref=andrew (ANDY20 - 20 % off)
Choosing between paid AI tools can feel overwhelming, especially when you are trying to find the best AI tools for research. In this video, I share my personal experience of putting the most popular models side by side in an academic stress test.
Have you tried NotebookLM by Google? Thoughts?
The comparisons include Perplexity vs ChatGPT, Claude vs ChatGPT, and even a closer look at Google Gemini, because so many students and researchers are asking the same question: which AI tool can I actually trust when my work depends on accuracy?
When I say ChatGPT vs the rest, I don’t just mean testing casual questions. I wanted to see how these systems hold up when you push them into the kind of tasks researchers actually face: extracting details from complex PDFs, checking whether citations really exist, and handling deliberately misleading prompts. These are challenges that anyone in a PhD program or an academic career will eventually run into, and it is important to see how different AI tools for research perform under that kind of pressure.
Would be interesting to try the different models available through Perplexity besides their own Sonar. Claude, Gemini, and ChatGPT are all options when using Perplexity Pro, so how does using those through Perplexity compare to their "native" versions?
Rather than treating this like a product review, I take the perspective of someone who is living inside the research process every day. I am interested in whether these tools can cut down wasted time, reduce the risk of error, and provide clarity when the workload feels unmanageable. The goal is not only to compare the interfaces or pricing, but to ask which of these systems can become reliable partners in academic work.
By running these stress tests, I have been able to see both the strengths and limitations of each model. The results matter if you are starting a PhD, if you are deep into data collection, or if you simply want to explore the best AI tools for research before committing to a paid subscription. The decision is less about which one wins overall, and more about which tool matches the specific task you need to accomplish. This is why a head-to-head look at Perplexity vs ChatGPT, Claude vs ChatGPT, and Google Gemini can help you decide where your money and trust should go.
What about the API offerings? They have been consistently better than any of the default/chat offerings when ive A/B tested these things against themselves (chatgpt 4o vs gpt-4o api, gemini 2.5 pro chat vs the same model over the AI, same with anthropics models). It comes down to specificity and tightness of system prompting in my experience. When you control more of the layers between your prompt questions and the models themselves, you get fewer errors, fewer hallucinations, and more specific/tailored responses.
................................................
▼ ▽ TIMESTAMPS
I realized that Chat GPT is good for "earthing" ideas or brainstorming for papers, but for outline design and precise advice y prefer Gemini. So CHATGPT=Creative phase; Gemini=Productive phase
00:00 Intro
00:22 The Presentation
Hi, @DrAndyStapleton.
Thank you for the video
I'd like to know if you set Perplexity's source of info to academic papers for the experiments you discussed in this video. I'm aware it has such a feature.
00:46 What I wanted to achieve
01:37 Interrogating PDFs
Talk too much
02:20 Interrogation Results
03:11 Citations
Switching between yourself and the ppt every other second distracted me so much i couldnt watch this video beyond 2 minutes!!!
04:04 Accurate References Results
04:50 Take Home Message
What's wrong with you? Why do you move around like that? Is it Adderall or Vyvanse>?
06:04 What to use for the moment?
07:08 Outro
What about scispace AI??
................................................
▼ ▽ Socials for shorts and reels
Try GLM-4.6 and thanks me later
Instagram: https://www.instagram.com/drandystapleton/
TikTok: https://www.tiktok.com/@drandystapleton
Awesome video! Exactly what I was looking for. Still, can't pay for both Claude and Chat GPT :(
................................................
Precisely what I've been searching for for a couple of months! Superb combination and juxtaposition.
More User Perspectives
I wonder what using gpt-5 on perplexity instead of sonar would be like
@demigodessWhat about the case you crate chatGPT? Its kind of amazing after you kind of train it 😅
@ransoffer6649Gemini pro works really well for reports. The references, however, are slightly off but it gives direction to go on Google Scholar to citations and references.
@rampee1000finally understand the diff between these AI!
@SANJAY__GAMING__64Olovka's pretty solid for finding relevant sources. Definitely worth trying for those citation challenges.
@RakeshSoni-q8pandy always finds the best tools, love it
@dishanBagwan-mg9imbeen super confused about AI for research, this helped a lot
@aandverma6751didn't know claude was an option, interesting 🤔
@oliviaphilip3988gemini is kinda new for me, anyone tried it?
@ASWINIHAS111didn't know there were that many ai's for research 🤔
@Salomenoklangbro this is just what i needed for my thesis
@SamratBhollaUsing Olovka for my lecture notes has been great. It's been a perfect companion to these AI tools.
@bijoychacko9594claude vs chatgpt was interesting, never heard of them before
@MK_FIRE_OFFICIAL_001claude sounds interesting, gotta check it out 🤔
@Fairy_11223i was confused about these ai but this helps a lot
@MrTrend88wasn't sure which ai to try for my thesis research
@NoorAlam-dm2jwhonestly didnt know about gemini till now
@YashDeshani001This breakdown is super helpful! I use Olovka for notes and writing, it's been a solid study buddy. Anyone else?
@NeetuPandey-tx9dtwhos winning the ai battle here lol
@TttRtt-f1ythis was kinda helpful for my research project
@djshivrajdjannurpsfinally someone breaking this down right
@JatinKangra-b5wnever heard of gemini before, interesting 🤔
@Sumitbabryachatgpt is my fav but gemini seems cool too
@Altmas_vloghonestly this was helpful for my thesis work
@Kuldeepsinghania2468I've been using ChatGPT for citations, but Olovka's feature to pull in sources specific to my field is such a time-saver for essays!
@GoluKumar-q8mAppreciate the head-to-head comparison, but I'm a bit confused by the model choices, which seems to skew the results.
For Gemini, you used Flash 2.5, which is their fastest but least capable model. Why not use Gemini 2.5 Pro? It feels like comparing a high-end Claude model to Gemini's entry-level option. That's not really a fair fight.
Similarly for Perplexity, you used their default Sonar model. But the power of Perplexity Pro is that you can switch the underlying model to GPT-5 or Claude Opus 4.1 for more complex tasks. Not testing that feature misses a huge part of its value proposition.
Because of these choices, the comparison doesn't feel accurate or reflective of what each platform can really do at its best. Tbh, it almost seems like the test was set up for Claude to win. Are they a sponsor?
finally someone breaks down ai tools simply
@Bhabangamer-c4pdidn't know gemini was even a thing till now 🤔
@rankpusher08helped me decide which ai to try, thx
@HappyKidsLandsclaude vs chatgpt is the showdown we needed
@JothiPradishSuper helpful video! Still figuring out the best combo of tools, but Olovka's been solid for notes.
@LOLMomments-p3f