Google AI Pro for researchers and postgraduate students

Where a research workload runs into the free limits

Most people meet the ceiling of a free Google account in the same three places, and all three are ordinary research tasks rather than exotic ones.

• Document length. A free account works with a 32,000-token context window. Google AI Pro raises that to 1 million tokens, which Google describes as up to 1,500 pages at a time. A 250-page thesis, a bundle of twelve papers or a long policy document fits in one conversation instead of being fed in fragments.
• Recordings. Uploads are capped at 10 files per prompt, 100 MB per file and 2 GB for a video. Pro raises the total audio you can upload from 10 minutes to 3 hours, and total video from 5 minutes to 1 hour. For anyone working with interview or focus-group recordings, this is the single most useful difference.
• How often you can ask. Google publishes usage as a multiplier rather than a count: Pro gives 4 times the standard limits. There is no published number of prompts per day for any tier, and Google says the limits can change.

Nothing here is Nepal-specific. The Gemini app and Gemini Notebook are both available in Nepal, and the friction is paying Google in US dollars rather than access.

Stage one: mapping a literature you do not know yet

Deep Research is the feature to reach for at the beginning of a project, when you need to know what exists before you can judge it. You give it a brief, it plans an approach, browses and reads a large number of sources, then writes a report with citations. It is available on the free tier too. What Pro adds is report generation with the Gemini Pro model and higher limits, with no numbers published either way. It is 18+, and one run uses considerably more of your allowance than a normal prompt.

Briefs decide the quality. "Remittances in Nepal" produces a survey of the obvious. A usable brief names the question, the period, the disciplines, the kind of source you trust and the shape of the output: "Peer-reviewed and multilateral-agency literature from 2015 to 2025 on how household remittance income affects secondary school enrolment in Nepal. Separate quantitative findings from qualitative ones. Note sample sizes and methods. Flag where findings conflict."

Stage two: a reading room built from your own sources

Gemini Notebook, which Google renamed from NotebookLM, is the opposite tool. It answers only from the sources you upload, and it cites the passage it used, so you can click through to the paragraph rather than trusting a summary. For a literature review this is the workhorse: one notebook per chapter or per theme, every paper in it, then questions across the whole set.

Pro also adds chat customisation, so you can set the response style to something closer to an analyst than a tutor and control output length, plus advanced sharing options including chat-only notebooks, notebook analytics and watermark removal, which Google excludes in India, South Korea and Vietnam. Google changed these limits starting 2 September 2026, so check the live help page before you rely on a number.

The 300-source ceiling is the one that matters. A systematic review that ends with 180 included papers fits in a single notebook, which means you can ask "which of these use a difference-in-differences design" and get an answer grounded in the actual set. Our Gemini Notebook guide covers Audio Overviews and study formats, which are more useful for revision than for analysis.

Stage three: interrogating long documents

The million-token context window changes the kind of question you can ask. Instead of pasting a methods section, you put the whole paper, or several papers, into one conversation and ask across them: reconcile the sample descriptions, list every variable definition and where it differs, find the paragraph where the authors qualify their main claim. Computational researchers can also attach one code folder or GitHub repository, up to 5,000 files and 100 MB, which is enough for most analysis pipelines.

Long context is a reading aid, not a verdict. Ask the model to quote the passage it is basing an answer on, then read the passage. Where it cannot quote, it is guessing. The Gemini Pro model guide explains what the paid tier changes inside the app itself.

Stage four: drafting, data tidying and the parts that stay yours

Google's wording for the subscription is "Gemini in Gmail, Docs, Vids, and more", available to people aged 18 and over and only in select languages, which Google does not enumerate. In practice this is where a lot of small research friction goes: rewriting a paragraph that has drifted, tightening an abstract to a word limit, drafting the covering email to a gatekeeper, turning a messy exported spreadsheet into something with consistent column names.

Analysis is a different matter. Use it to explain a statistical test, to sanity-check the logic of a model specification or to write the boilerplate around your own code. Do not let it produce numbers you cannot reproduce in your own software, and never let generated text describe results you have not run.

Storage for datasets and recordings

Research storage fills up faster than people expect once audio, video and raw instrument data are involved. AI Pro includes 5 TB, and the important detail is that it is one pool shared across Gmail, Google Drive and Google Photos rather than 5 TB in each. Free accounts share 15 GB the same way. If the subscription ends you return to 15 GB: files stay but cannot be edited, uploads and Photos backup stop, and Gmail sending and receiving can be affected, with Google removing content only after two years over quota and at least three months' notice. Plan the archive before the plan lapses. The storage guide works through what fills a pool.

Integrity, disclosure and ethics

Is it worth it for a single thesis?

Honestly, it depends on the shape of the work. If your sources are a few dozen papers you have already read, the free tier plus a careful reading habit is enough. The plan earns its cost when you are handling hundreds of sources, hours of recordings or long documents you must query repeatedly, and when you would otherwise be splitting everything into fragments to fit a 32,000-token window. Undergraduates should also read our student guide, which covers Google's student offer and where Nepal stands in it.

If it is worth it for your project, WebsNP sells Google AI Pro in Nepal for $1.06 a month, billed as one 18-month term of $19.07, paid in rupees and activated on your own Google account. Google's own price is $19.99 a month, billed in US dollars.