unfortunately I had issues with hallucinations in paid apps, but try Google NotebookLM instead. it's free and basically perfect. TL;DR: best budget tool for thesis PDFs.

hey, saw this earlier but just now responding... in my experience, the thesis struggle is real. over the years ive tried so many tools cuz i was terrified of getting a methodology wrong and looking like an idiot during my defense lol. highkey, accuracy is everything when youre drowning in 40-page PDFs.

heres what i recommend for staying safe on a budget:

- Google NotebookLM - honestly, this is the gold standard for students. it's currently $0/free and stays grounded only in the files you upload, which reallyyy limits those wierd hallucinations.
- Claude 3.5 Sonnet - the free tier is actually insane for technical nuance. if you need more volume, the Claude Pro sub is about $20/month, but the free version is usually enough for a few papers a day.

basically, the lesson i learned is to always use the citation features to click back to the original text. dont trust any AI 100% blindly!! gl with the mountain of work.

Seconding the recommendation above for sure!! Been thinking about your mountain of PDFs and honestly, Google NotebookLM is a total lifesaver for thesis work because it's free and actually built for this stuff. Most AIs fail because their "memory" is too small, but this one handles huge documents without breaking a sweat.

Here is why it's a solid budget pick:
- It uses your specific PDFs as the ONLY source, which highkey cuts down on hallucinations.
- You get citations for every claim. If it says something about the methodology, you can click it and it shows you the exact paragraph on page 30.
- It's free, unlike paying $20/month for ChatGPT Plus or Claude Pro.

If you need a second opinion on a tricky paper, maybe try the free version of Claude 3.5 Sonnet. It’s super smart with technical language, but you'll hit a message limit pretty fast. Stick with NotebookLM for the heavy lifting tho. gl with the thesis!!

Wow ok that changes things. Gonna have to rethink my approach now.

Noted!

.

Totally agree that those free tools are a solid starting point, but if you're really trying to go the DIY route for a full thesis, you might want to build a more 'pro' research stack. Tbh, managing a mountain of PDFs is basically a full-time job and just dumping them into a chat box can get messy fast. Here’s what I’d recommend if you want a more self-service, systematic approach: - Scholarcy - This is basically the gold standard for 'professional' DIYers. It creates summary flashcards and extracts tables or figures into Excel, which is a lifesaver for checking methodology without reading the whole 40 pages.
- Zotero with AI plugins - If you aren't using a reference manager yet, you gotta start. There are plugins that let you run LLMs directly on your library so your 'brain' is always organized and you aren't just uploading files one by one.
- Elicit - Instead of just summarizing, it’s great for extracting specific things like sample sizes or specific results across multiple papers at once. I think maybe these options feel a bit more robust because you're building a database rather than just chatting with a file, you know? It’s definitely helped me avoid those weird hallucinations when things get too technical.

Ok so I have been using Elicit for over half a year now and it has basically saved my life with my lit review. Most people just want a quick summary, but for a thesis you really need to keep track of the same details across every paper you read over several months. Anyway, the reason I like this one is the table feature. You can upload like 30 papers and ask it to extract the methodology or specific results for all of them at once. It creates this big grid that stays in your account so you can come back to it weeks later. I am still not 100 percent sure how it handles the really dense math sections, but for general methodology and findings, it feels way more reliable than just a regular chatbot because it is designed for researchers.

• It lets you create persistent tables to compare methodology across dozens of PDFs at the same time.
• You can save your library and keep adding to it as you find more sources for your thesis.
• It uses a specialized search that feels much more grounded in the actual text than a basic LLM. TL;DR: Go with Elicit if you want to stay organized for the long haul. It is less about a one-time summary and more about building a searchable database for your whole project.

Finally someone says it. Ive been thinking this for a while but wasnt sure.

Re: "Finally someone says it. Ive been thinking this..." - seeing the consensus land on grounded models is good. From a technical perspective, managing 40-page papers requires a high context window and specific source-grounding to avoid hallucinations. Heres a summary of the community insights so far for those on a budget:

• Google NotebookLM is the top recommendation for free users. It handles massive token counts and keeps the model focused strictly on your uploaded files.
• Elicit works best if you need to compare methodologies across multiple papers in a grid, though it uses a credit-based system.
• Scholarcy is decent for generating flashcards of papers but involves a subscription for the full suite. Tbh, if accuracy is the priority and you dont want to spend money, sticking to the first option is the move. It basically functions as a specialized RAG (Retrieval-Augmented Generation) tool without the setup complexity. Just make sure your PDFs have clean metadata so the citations stay accurate.