How Bodhi works

Where does the data come from?

Bodhi connects to 10 academic databases simultaneously — covering published journals, theses, preprints, open access papers, and biomedical literature. This includes a dedicated index of 6 lakh+ Indian theses that no other platform offers.

When you search, all databases are queried in parallel. Results are deduplicated — if the same paper appears in multiple databases, you see it once with the richest metadata (DOI, journal, volume, pages, citation count).

We currently access 250M+ unique academic works spanning all disciplines — STEM, humanities, social sciences, medicine, and more.

Is the gap finder actually finding gaps — or just summarizing?

It's finding real gaps. Here's the process:

1. Quantitative search: Your topic is searched across all databases. The system counts how many papers exist — in India and globally.
2. Comparison: AI compares what EXISTS against what COULD exist — identifying specific angles, geographies, methodologies, or populations that haven't been studied.
3. Ranking: Gaps are ranked by severity — critical (zero papers), moderate (few papers), partial (partially covered).

When Bodhi says "zero studies exist on groundwater microplastics in Maharashtra" — that's based on actual search results across all 10 databases, not an AI guess. The actual papers found are shown below the analysis so you can verify.

Are the citations real and verifiable?

Yes. Every citation Bodhi suggests is a real paper that exists in our databases. We do not generate or hallucinate references.

The process:

• You select text in your document
• The system searches academic databases for papers semantically related to your text
• Only real, indexed papers are returned — with title, author, year, journal, and DOI where available
• DOI links let you open and verify any citation directly

What we're transparent about: Our evidence classification (supports/contradicts) is based on paper abstracts, not full text. The AI reads the abstract and provides reasoning — but you should always verify by reading the actual paper.

How do Compare Sources and Peer Review work?

Compare Sources: Select a claim or paragraph in your document. The system finds related papers, then AI compares each pair to determine if they agree, conflict, or are unrelated. You get a conflict matrix showing consensus points, disagreements, and gaps that no paper addresses.

Simulated Peer Review: Click one button and AI reads your entire document as a strict academic reviewer would. It scores your work across 6 dimensions (clarity, argument, evidence, methodology, originality, completeness), flags issues by severity (critical/major/minor), and gives a verdict (accept, revise, reject) with specific fix suggestions.

Why this matters: Many Indian PhD students don't get timely feedback from supervisors. A simulated review gives instant, structured feedback so you can improve your draft before submission.

Is this just a ChatGPT wrapper?

No. A ChatGPT wrapper sends your text to an AI and returns whatever it generates — including made-up citations and unverifiable claims.

Bodhi works differently:

• Database-first approach: Real papers are found FIRST from academic databases, THEN AI analyses them
• AI cannot hallucinate sources: It can only reference papers that were actually retrieved from the database
• Structured workspace: Full document editor with formatting, references section, export — not a chat interface
• You write, AI assists: We never generate research content for you. AI only finds evidence, improves language, and validates claims

The AI layer is used for language tasks (improve English, classify evidence) — not for creating research content. Your thinking, your words, your thesis.

How does the Proposal Builder work?

The Proposal Builder creates a structured research proposal by combining real search data with AI generation. It follows the standard Indian university proposal format:

1. Topic analysis: Your topic is searched across all databases. The system identifies how many theses and papers exist, which universities have studied it, and the year distribution.
2. Gap identification: AI analyses the existing research landscape and identifies what's missing.
3. Problem statement & objectives: Based on the gaps found, AI generates a formal problem statement and 4-5 measurable research objectives.
4. Full proposal (Pro): Hypotheses, literature review outline (themed sections), methodology suggestion, and expected outcomes — all grounded in the research data found.

Key principle: The proposal is generated based on REAL gaps found in the database — not generic AI text. When it suggests "mixed-methods approach," it's because the data shows that methodology hasn't been applied to your specific topic.

You can copy individual sections, copy the full proposal, or send it directly to the Writing Workspace as a new document to continue editing.

How does overlap checking work?

The overlap checker scores your proposed topic against all existing research:

1. Your title/abstract is searched across all databases
2. The closest existing works are identified
3. AI scores overlap across 5 dimensions: subject, geography, methodology, research questions, and population
4. A weighted score (0-100%) tells you how original your topic is
5. Clear verdict: safe to proceed, proceed with caution, or high overlap

Important distinction: This is overlap checking (is my topic original?) — not plagiarism checking (is my text copied?). They serve different purposes at different stages of research.

What Bodhi doesn't do (and why)

• We don't write your thesis for you. AI-generated academic text is increasingly detected and penalized by universities. We help you write better — not write for you.
• We don't provide page numbers in citations. Page numbers require full-text PDF access. We work with abstracts and metadata for speed and scale.
• We don't do plagiarism checking. That's a separate tool for a separate purpose. We check topic originality, not text similarity.
• We don't guarantee completeness. No tool can search every paper ever published. We cover 250M+ works across 10 databases — but niche, non-indexed works may be missed.