Undergraduates
A first literature review that is actually reviewed literature: sources found, kept for a reason, cited correctly, on a page the faculty will accept.
Describe it in a line. IQThesis plans the research, reads the literature and the statistics where they live, drafts chapter by chapter onto a real page, and cites only what the sources actually say. English, French or Arabic.
Free while in beta. No card. Your writing stays yours.
What a run looks like — the real stages, in order.
| A chatbot | ||
|---|---|---|
| Sources | ReferencesReal published papers with DOIs, and institutional statistics when the brief asks for them | ReferencesOften invented or impossible to verify |
| EvidenceAn evidence line per source — what it establishes — recorded before it is cited | EvidenceNo record of what a source actually says | |
| Your own papersUpload them; they are indexed and searchable by meaning, down to the page | Your own papersPaste excerpts by hand | |
| Writing | The pageA typeset A4 or Letter page, columns, front matter, a bibliography that keeps itself | The pageA chat box; you paste into a word processor |
| The processAn outline you approve, then chapters that land as tinted drafts you accept or send back | The processOne long answer, take it or leave it | |
| LanguagesEnglish, French and Arabic — with right-to-left pages set the way a faculty wants them | LanguagesText in many languages, no page | |
| Citation stylesAPA, MLA, Chicago, IEEE, Vancouver, Harvard — switchable, everything follows | Citation stylesFormatted by hand, checked by hand | |
| Integrity | SimilarityA reading of what matched indexed sources, each match classified and told what fixes it | SimilarityNothing |
| Machine-likenessA score with the signals behind it and the checker’s reasoning — a probability, never a verdict | Machine-likenessNothing | |
| PromisesNone that a supervisor would not accept | Promises“Undetectable”, “plagiarism-free” |
Each one takes a passage or a brief and gives something back you can check. None of them promises what it cannot show.
Sixteen templates across four kinds of document, each with its page already set — margins, columns, front matter, citation style.
The outline is approved once; each chapter lands as a draft, the run parks, and you decide whether to go on, change the plan or stop. Six to nine chapters, level-two parts inside each.
Title, authors, abstract and the sections, written in one pass and set in two columns if the venue wants them — IEEE, ACM or a journal’s own look.
Short, dense, numbered citations in order of first appearance, an abstract that gets written like any other section rather than pasted in afterwards.
Scope, background, findings, recommendations — with the figures the brief asked for, from the institutions that publish them, cited as what they are.
Once deployed, a contract cannot easily be patched, and the assets it controls may be substantial. This paper reviews the guarantees each verification technique offers and the barriers reported by practitioners.
Blockchains were promoted as platforms for trustworthy, automated execution, yet widely used contract systems still fail in surprisingly fragile ways. Programming mistakes and subtle interactions between contracts have repeatedly led to stuck funds[1].
Against this background the question is not whether contracts need stronger assurance, but which forms of assurance are realistic to obtain for real-world code[2].
Surveys report that formal techniques can already prove or refute some classes of property for substantial subsets of widely used languages, though coverage varies between approaches[1][3].
At the same time, foundational questions about how best to validate contracts remain unsettled, and the difficulty of reading counterexamples is cited more often than proof power.
The pageA4 or Letter, your margins, your columns — laid out as you write rather than guessed at and discovered at print time. Two-column articles break across columns and pages the way the printed paper will, and an Arabic paper starts in the right column.
APA, MLA, Chicago, IEEE, Vancouver and Harvard, rendered by the same engine reference managers use. Change the style and every citation and the bibliography follow. Numbered styles count in order of first appearance.
A first literature review that is actually reviewed literature: sources found, kept for a reason, cited correctly, on a page the faculty will accept.
A mémoire with a structure, in French or Arabic, its sources on file with what each one establishes — so the defence is about the argument, not the formatting.
Chapters that land one at a time, an evidence map that grows with the thesis, and a page that stays typeset at 60,000 words.
Articles in two columns, numbered citations, a similarity reading before submission, and rewriting that keeps every claim and every number.
A tool that helps you write for a degree must be one a supervisor would sign off on. These are the rules it is built on.
The writer can only cite what the search found — a published paper with a DOI or an institutional page with a URL. If nothing relevant was found, it says so and writes nothing.
A share that matched indexed sources is reported as exactly that. Each match is named for what it is, and the fix offered is attribution — a citation, a quotation, your own synthesis — never rewording until a number drops.
Nothing the agent writes becomes your text until you keep it. Drafts land tinted, with Accept and Back above the page, and the original is restored exactly.
Figures and tables, equations and cross-references are not built, and neither is collaborative editing. Arabic pages are set right-to-left in the editor, but the Arabic PDF export is not yet final. Everything above this line is working today.
Yes, while it is in beta: no card, up to 30 research runs a month and 200 documents on the free plan. Asking the agent about your document does not count against anything.
No. Every source comes from a search it actually ran — a published paper with a DOI, or an institutional page with a URL when the brief asks for statistics — and each is kept with a line saying what it establishes. If nothing relevant is found, the run stops and says so.
We do not promise that, and we would not trust a tool that did. Detectors are probabilistic. IQThesis gives you the reading — the score and the signals behind it — and a refinement that improves clarity and tone while keeping your claims. It does not sell evasion.
English, French and Arabic. The page follows: an Arabic manuscript is set right-to-left, in an Arabic Naskh face, with its columns starting on the right and its headings in Arabic.
PDF and Word, carrying your columns, your title block, your abstract and your bibliography. The PDF is real selectable text, never a picture of the page.
You do. The drafts are yours to keep or discard, the documents are yours to export and delete, and nothing you write trains anything.