AI for Research
AI that assists your research. Never replaces it.
Dissertatio's research copilot verifies citations, repairs references, checks claims against sources, and reviews consistency — under guardrails you can read, inspect, and export.
Guardrails first
Three promises, before any feature.
01
Never fabricates
The copilot cannot cite a source that does not resolve. Every suggested reference is verified against Crossref before it reaches you — a hallucinated citation is structurally impossible to insert.
02
Always inspectable
Every AI action shows what it read, what it concluded, and why. Suggestions arrive as reviewable diffs, and each one is logged to your project's audit history — nothing happens silently.
03
Disclosure-ready
Dissertatio keeps a per-manuscript record of every AI interaction. If your journal asks what role AI played, you export the report — a complete, accurate answer in one click.
Citation help
Find the right source — never a fake one.
Ask for supporting literature on a claim and the copilot searches real, resolvable records. Every candidate arrives with verified metadata, and you decide what enters the manuscript. It cannot invent a paper, an author, or a page number.
- Candidates verified against Crossref before you ever see them
- Suggestions ranked by relevance to the exact sentence
- You insert; the copilot never cites on your behalf
Reference repair
A reference list that fixes itself — with your sign-off.
Dead DOIs, preprints with published successors, duplicate entries, and malformed metadata are found automatically. Each repair is proposed as a diff with the evidence attached; nothing changes until you accept it.
- Broken DOIs re-matched by title, author, and year
- Preprint-to-published upgrades suggested with both records shown
- Duplicates and citation-style drift merged cleanly
Claim checks
Does the source actually say that?
The copilot compares each claim with the source it cites and shows you the relevant passage. When the citation over-reaches, you get the exact wording mismatch — with the evidence in front of you, not a black-box verdict.
- Sentence-level comparison against cited sources
- The supporting (or contradicting) passage is always displayed
- Warnings suggest precise language, never rewrite silently
Consistency review
The whole manuscript, cross-checked.
Numbers, abbreviations, terminology, and cross-references are checked across every section. The reviewer who catches your abstract disagreeing with your results table should be a machine — before it is a human.
- Sample sizes, statistics, and units reconciled across sections
- Abbreviations and terminology checked for first-use definitions
- Figure, table, and equation references validated
Journal policies
Assistive, not generative.
Most journal AI policies distinguish tools that verify and check from tools that generate prose. Dissertatio's checks are built to sit firmly on the assistive side of that line — closer to a spell-checker or a statistics linter than to a ghost-writer.
Verification of references, statistics, and internal consistency typically falls outside the AI-use declarations journals require, in the same way grammar and plagiarism checkers do. And for anything beyond checks, every AI interaction in your project is logged — so your disclosure, when one is needed, is generated from the record rather than from memory.
Always follow your target journal’s policy. Dissertatio makes that easy: export the AI usage report for any manuscript, any time.
Limitations
What our AI will not do — on purpose or otherwise.
Trust requires knowing the edges. These are the boundaries of the copilot, stated as plainly as its capabilities.
It does not judge scientific merit.
Novelty, significance, and interpretation are yours and your reviewers'. The copilot checks verifiable properties of the manuscript — it has no opinion on whether the science is good.
A passing check is a signal, not a proof.
Validation reduces the risk of slips; it does not replace careful reading by you and your coauthors. Treat green ticks the way you treat a passing test suite.
It can only read what it can access.
Claim checks work against open-access texts, records you attach, and resolvable metadata. Paywalled sources it cannot reach are marked as unverified — never assumed correct.
It will not write your paper.
By design, the copilot declines to generate sections from nothing. It verifies, repairs, and reviews — authorship stays with authors.
Verification you can read, on every manuscript.
Start free and run the standard checks on your current draft — the guardrails come with every plan.