One fabricated source
can cost you the grade.

ChatGPT hallucinates citations. Acurio checks every source in your thesis against the original. In five minutes, before your examiner does.

Acurio review view: thesis with color-coded citation verdicts and issue list

16 unsupported citations found

Unsupported

The source never defines the claimed prevention concept. Its focus is staging presymptomatic type 1 diabetes, not primary prevention.

Checked passage

“Primary prevention aims to prevent the onset of islet autoimmunity in the general population (Insel et al., 2015).”

Source: Insel et al. (2015) · Staging Presymptomatic Type 1 Diabetes

140 / 185 citations supported

26 partially supported

140 supported 26 partial 16 unsupported

Used by students at: ETH Zürich, LMU München, IU Internationale Hochschule, Berner Fachhochschule, Universität Wien

Your paper, x-rayed.

This is exactly what you see after upload. Hover over any coloured citation: the original quote, page, and match score appear on the right. This is exactly what you see after upload. Tap a coloured citation: the original quote, page, and match score appear below.

CH4_COGNITIVE_LOAD.DOCX Chapter 4 · Theoretical framework
PAGE 47 / 134
SUPPORTED PARTIAL UNSUPPORTED NO SOURCE

PAGE 47 / 134

4.2 Load on working memory

Cognitive load theory distinguishes intrinsic from extraneous load, a relationship that intensifies with the number of interactive elements [Sweller, 2011]. While intrinsic load arises from the material itself, extraneous load can be shaped by instructional design [Kalyuga, 2007].

Earlier work tied the effect of scaffolded feedback to motivational effects [Deci & Ryan, 2000]. A 28 % reduction in extraneous load, however, is consistently documented [van Merriënboer, 2010]. Some sources cannot be traced [Müller, 2019].

Earlier work tied the effect of scaffolded feedback to motivational effects [Deci & Ryan, 2000]. A 28 % reduction is consistently documented [van Merriënboer, 2010]; other sources support only part of the claim [Kalyuga, 2007].

PARTIAL Match 63%

"Effects were observed on novice learners, though the threshold differs for experts."

Kalyuga (2007) · p. 512

6 citations · 4 supported 4 1 1 0

Three steps. A few minutes of work.

From upload to colour-marked document.

Upload your .docx.

Drag & drop: your Word file with Zotero field codes plus your source library as BibTeX, RIS, or a stack of PDFs. Drag & drop: your Word file with Zotero field codes plus sources as BibTeX, RIS, or a stack of PDFs.

✓ thesis_v7.docx 1.2 MB
✓ references.bib 84 KB
✓ 14 source PDFs 38 MB

Acurio reads and checks.

Every citation is resolved to the matching passage in the source: page, paragraph, match score. Live in the terminal. Every citation is resolved to the matching passage: page, paragraph, match score. Live in the terminal.

Analysing 41/60 … ok
[Renkl, 2014] → p. 147 · 94%
[Deci, 2000] → no passage
[Kalyuga, 2007] → p. 512 · 63%

Read the verdict.

Your paper opens colour-marked in the browser. Hover for the original quote, export as a DOCX report. Your paper opens colour-marked in the browser. Tap for the original quote, export as a DOCX report.

… the load drops by 28 % [van Merriënboer, 2010]

✓ DOCX REPORT ·SHARE LINK
Demo

Free bibliography check.

Paste your bibliography. Acurio instantly checks every entry against Crossref & OpenAlex and flags retracted papers. Try it right here.

Paste your bibliography. Acurio instantly checks every entry against Crossref & OpenAlex and flags retracted papers.

What Acurio actually does for you.

Not abstract promises. Concrete tools that make every citation defensible.

Four colours, no guesswork.

Every citation is underlined beneath the prose: supported, partial, unsupported, or no source. You see more on one page than in ten hours of reading.

Every citation is underlined beneath the prose: supported, partial, unsupported, or no source.

PAGE 47

The study shows a 28 % increase in load [Renkl, 2014], especially among novices [Kalyuga, 2007]. An effect on motivational variables, however, is [Deci, 2000] not clearly supported. Some sources are unavailable [Müller, 2019].

Hover to see the original text. Tap to see the original text.

Instead of jumping between PDFs, you get the matching paragraph inline. With page and match score.

● SUPPORTED · VAN MERRIËNBOER, P. 147, 91%

"Scaffolded feedback reduced extraneous load by 28% in the experimental condition."

German ↔ English.

Mix a German thesis with English sources, no translation step.

… reduziert die kognitive Last messbar. DE THESIS
… reduced extraneous cognitive load. EN SOURCE

Down to the page.

Every verdict ships with a concrete source, page number, and score, no "somewhere in the PDF".

Every verdict ships with a concrete source, page number, and score.

p. 14794% matchParagraph 3

Mixed sources, one pass.

BibTeX, RIS, raw PDFs mixed. Acurio indexes everything in one pass.

.docxBibTeXRISPDF

Your work stays yours.

HTTPS upload, no training use, EU hosting. You can wipe everything with one click, including your sources.

HTTPS upload, no training use, EU hosting. Everything deletable with one click.

An examiner reads the whole thesis.

PRO

Not citation by citation, but like your examiner: overall verdict, strengths, weaknesses, and gaps, each anchored in your text. In two minutes instead of two weeks waiting for supervisor feedback.

Overall verdict, strengths, weaknesses, and gaps, each anchored in your text. In two minutes.

FULL REVIEW · THESIS_V7.DOCX 2 MIN

"Solid argumentation with clean methodology. Two structural gaps before submission."

  • STRENGTH Methods section cleanly operationalised, sample comprehensibly justified Methods section cleanly operationalised, sample justified §3.2
  • STRENGTH Theory section connects both research strands cleanly §2
  • WEAKNESS Research question 2 is never explicitly answered in the discussion Research question 2 is never answered in the discussion §5
  • MISSING Limitations section: examiners expect it before the conclusion

Prioritised next steps

1. Add limitations (before §6)
2. Tie the discussion back to research question 2
3. Sharpen the abstract on the result

Critique per section, with § anchors
Quality filter: only defensible findings
Prioritised next steps
Optional: gender-inclusive language pass

Check gender-inclusive language OPTIONAL

Follows your guideline's convention: findings with citation anchor and a concrete rewording suggestion.

Genderstern*Doppelpunkt:PaarformNeutral

p. 12: "die Probanden" → "die Teilnehmer*innen"

2 Rich Reviews in the Pro pack · 4 in the Lab pack

A second AI assessment, while you sleep.

For every unsupported or doubtful citation, an independent model searches your full library overnight for a passage that actually supports the claim and delivers the result by morning.

For every unsupported citation, an independent model searches your full library overnight for a passage that actually supports the claim.

Run thesis_v7
Re-review 7 unsupported
Delivery 06:42, tomorrow

Suggestion: van Merriënboer (2010), 4C/ID, p. 147 supports the claim. Match 91%.

Start Second Opinion 1× in the Student pack · up to 3× from Pro

What examiners see before they read. PRO

Formal requirements and the oral exam, both prepared from your own thesis.

Your university's rulebook: checked automatically.

Upload your institution's style guide once. Acurio checks every rule: citation style, margins, font size, declaration of authorship. Upload the guideline once. Acurio checks every rule: citation style, margins, font size, mandatory declarations.

UNI ZURICH · GUIDELINE V4.2 ● 23 / 26 MET
Citation style APA 7 Met
Font size 12 pt, line spacing 1.5 12 pt · 1.5
Left margin 2.5 cm 2.3 cm
Declaration of authorship Missing

Practice questions from your own thesis.

Acurio generates the questions an examiner is likely to ask, drawn directly from your arguments and sources, with location and difficulty. Questions an examiner is likely to ask, drawn directly from your arguments and sources.

Q · p. 47 "You claim on page 47 that scaffolded feedback reduces cognitive load by 28 %. What effect would reverse that finding for experts?" "You claim that scaffolded feedback reduces the load by 28 %. What effect would reverse that finding for experts?"
Q · p. 62 "Why did you compare mathematics and physics in Renkl but not in Sweller?"
● 18 QUESTIONS GENERATED ● 18 QUESTIONS 7 HARD EXPORT AS DOCX AS DOCX

Take the findings with you.

Comments like a Lektorat.

Every finding lands as a Word comment on the exact citation, ready to hand to your supervisor.

… the load drops by 28 % [Renkl, 2014]

Acurio

No matching passage in Renkl 2014. Closest hit: 4 % at item level (p. 88).

Built for Zotero.

Field codes in the .docx, BibTeX and RIS exports. Acurio understands your library as it is.

Field codesBibTeXRIS

Your AI assistant, plugged in. PRO

Claude or Codex pull every flagged citation and fix it right in the chat. Read-only.

ClaudeChatGPT (soon)Codex

Every feature.
One line.

What Acurio does, which pack it's in, where it's explained. Click a row to jump to the section.

Tap a row to jump to the section.

  • FREE usable without an account
  • ALL included in every pack
  • PRO from USD 36 / one-time

Pay once. No subscription.

One pack per thesis. Use your analyses up, then buy again, nothing else.

No renewal Nothing to cancel 10 citations free up front

Start free, no credit card.

Upload your thesis. We verify the first 10 citations with full sources: once, free. See exactly how many problems you have before you pay.

We verify the first 10 citations with full sources: once, free. See exactly how many problems you have before you pay.

Start free

Student

USD 24 / one-time

  • 3 full analyses (every citation checked)
  • Second Opinion on disputed citations (1 included)
  • DOCX export
Get Student
MOST POPULAR

Pro

USD 36 / one-time

Your whole thesis, fully checked.

  • 10 full analyses
  • Rich Review: whole-thesis examiner pass (2 included)
  • Second Opinion (up to 3 included)
  • Defense Question generator (2 included)
  • Guideline engine + gender language + MCP
  • 10 full analyses
  • Rich Review: whole-thesis examiner pass (2 incl.)
  • Second Opinion (up to 3 incl.)
  • Defense Question generator (2 incl.)
  • Guideline engine + gender language + MCP
Get Pro → Get Pro

Lab

USD 60 / one-time

  • 20 full analyses
  • 4 Rich Reviews + 5 Defense Q
  • Multi-document workspace for teams / labs (coming Q3)
  • Priority queue (coming Q3)
Get Lab

Lifetime

USD 99 one-time

  • Lifetime access: no subscription renewal
  • Monthly analysis reset
  • All Pro features included
Get lifetime

One-time payments, no subscription. Buy again when used up. Lifetime refills monthly. One-time payments, no subscription. Buy again when used up.

Acurio compared

Feature comparison with the tools students actually use.

Deep analysis (claim vs. source)
Upload your own source PDFs
Commented DOCX export
Guideline check against university rules
Gender-inclusive language check (optional) Gender-inclusive language (optional)
Defense-question generator
Full German support
GDPR-compliant
Swiss-run, EU hosting

NBLM = NOTEBOOKLM, CITE = CITETRUE, TRNK = TRINKA

What's coming next.

Transparent. You can vote on it directly.

● Done Q2 / 26
  • Citation check (DE / EN)
  • Hover excerpts with page reference
  • Overnight Second Opinion
  • Rich Review: whole-thesis examiner pass
  • Defense-question generator
  • Institutional guideline engine
  • Zotero .docx import · BibTeX & RIS
  • Gender-inclusive language check

Citation check (DE/EN), Hover excerpts, Second Opinion, Rich Review, Defense questions, Guideline engine, Zotero import, Gender-inclusive language

● In progress Q3 / 26
  • Lektorat-grade DOCX export
  • Heatmap / effort overview
  • Multi-document workspace for teams

Lektorat-grade DOCX export, Heatmap / effort overview, Multi-document workspace for teams

● Planned Q4 / 26
  • Co-author mode
  • Live Word plug-in
  • Custom LaTeX templates
  • Supervisor API
  • Citation-style editor

Co-author mode, Live Word plug-in, LaTeX templates, Supervisor API, Citation-style editor

What you want to know before you upload.

What is Acurio?
An AI-powered citation-checking tool. It reads a DOCX document with Zotero citations, compares every cited claim against the uploaded source PDFs, and reports per citation: supported, partially supported, or unsupported, with source excerpt and confidence score.
How does the AI citation check work?
Acurio resolves every citation to the matching passage in the source: page, paragraph, match score. A language model actually reads the passage and judges whether it supports your claim. Not a metadata comparison, a substantive check.
Is Acurio GDPR-compliant and good with Swiss data protection?
Yes. Swiss-run, hosted in the EU (Germany), HTTPS upload, no training use of your documents. You can wipe everything with one click, including your sources.
Is Acurio a subscription?
No. Packs are one-time payments per thesis (USD 24-99). Use your analyses up, then buy again, there is nothing to cancel. Lifetime refills monthly.
Is Acurio suitable for a bachelor or master thesis?
Yes, that's exactly what it's built for: .docx with Zotero field codes, mixed German and English sources, typical lengths of 30-150 pages. We check your first 10 citations free.
Does Acurio do plagiarism checking?
No. Acurio checks the opposite: whether your sources actually support your claims. Classic plagiarism software finds copied text. Acurio finds wrongly supported claims.
Which citation styles are supported?
APA, Harvard, Chicago, and more, through the guideline engine, your university's formal requirements (style, margins, mandatory declarations) are checked automatically as well.

Upload your paper.
In five minutes, see where it's weakest.

No setup, no credit card. Just your .docx, your sources, and a colour-coded result.

Privacy first. No credit card. Deletable anytime.