AI Homework Helper
The AI Homework Helper is evolving into a central tool in academic life in Winter 2025/26. It accelerates processes such as research, structuring, and language optimization, with responsibility for content and accuracy always remaining with the students. Universities are increasingly integrating AI into their guidelines, focusing on transparency and independent work, and defining its use not as a niche, but as part of studies and exams.
Source: Own Representation
Fundamentals & Guidelines
Universities are no longer treating AI as a niche topic, but as an integral part of studies and exams. The University of Zurich has adopted guiding principles for AI in research and teaching and is working on a university-wide policy. This shows that it's not about "whether" to use AI, but "how".
International organizations such as UNESCO have also published guidelines on Generative AI in education and research. These aim to develop rules and competencies in a way that GenAI empowers learners and educators ( UNESCO Document ).
Local regulations concretize these requirements. The Lucerne University of Applied Sciences and Arts emphasizes in its guidelines on AI tools that AI results are not scientific sources and that students are responsible for relevance and accuracy. The ETH Zurich is publishing its own guidelines for Generative AI in Teaching & Learning, highlighting responsible use, transparency, and fairness ( ETH Guidelines
An AI Homework Helper is therefore part of a practice that universities are regulating, often with a focus on declaration and independent work ( ETH Library ).
AI Use in Research
Essays often fail due to literature research. An AI Homework Helper can serve as a research assistant here, not as a text generator. Google Scholar provides many hits, but AI helps to sort them.
Elicit positions itself as a tool for scientific research, enabling interactive screening and data extraction to narrow down literature in real-time. It supports steps of systematic reviews such as searching, abstract screening, and extraction ( Elicit Blog ), including full-text screening as an intermediate step ( Elicit Support ).
Source: robots.net
An example: A student searches for "fluid balance" in heart failure and gets results that are too broad. The AI Homework Helper helps to categorize hits by study design, population, or limitations, based on metadata and abstracts, which are then manually checked.
Connected Papers is a visual tool for finding and exploring relevant papers. The ETH Library describes it as a search tool that graphically displays relevant articles. This allows the "neighborhood" to be explored starting from a central paper and important works to be found more quickly.
Consensus is an academic search engine that provides answers based on peer-reviewed literature. However, the ETH Library warns to check AI-assisted outputs and not to blindly adopt them. Consensus provides a rough map, while the original papers provide the details.
The AI Homework Helper accelerates the selection and organization of sources. The student makes the decision about the relevance of the evidence in the full text.
AI Use in Writing
When writing, the usefulness of AI is quickly limited when it comes to generating entire chapters. University guidelines require transparency and responsibility. The HSLU clarifies that AI results are not scientific sources and students remain responsible for truth and accuracy. The ETH also emphasizes responsibility and transparency.
AI may improve language, but not replace evidence. DeepL Write improves spelling, grammar, and phrasing. LanguageTool is an AI-based grammar checker for grammar and spelling. Such tools make arguments readable without changing the content.
An example: A technically correct but overly protocol-like paragraph can be streamlined with DeepL Write. Every word must be checked to avoid diluting the technical language. LanguageTool then corrects minor errors like duplicate words or commas. The content remains independent, while the presentation becomes professional.
For clean citation, Zotero is useful. It offers Word processor plugins for Word, LibreOffice, and Google Docs that dynamically update bibliographies when new citations are inserted ( Zotero Installation ).
Challenges & Detection
With increasing AI use, universities turn to detectors, but their reliability is often uncertain. OpenAI has removed its own "AI classifier" due to insufficient accuracy. Turnitin states a false-positive rate of about 4% at the sentence level for its AI writing detection. Vanderbilt has deactivated Turnitin's AI Detector due to a lack of transparency and the risks of erroneous decisions.
Source: updf.com
The debate is being conducted in the media. WIRED reports on Turnitin numbers as an indication of mass usage. The Guardian describes cases of false accusations and stress caused by detectors. A later Guardian-Artikel emphasizes that universities cannot solve the problem with simple technical solutions and points to the limits of detection and new examination formats.
For practical purposes, this means: those who want to work fairly document their AI use, work in a traceable manner, and keep drafts, notes, and version statuses ready. This serves as self-protection in an environment where "proof" has become more complicated.
University Requirements
Universities are formulating concrete requirements. The HSLU warns that papers without complete labeling of sources and aids can be considered deception. Another HSLU-Richtlinie states that text passages from generative AI tools cannot be cited as sources but must be transparently labeled.
The ETH Library recommends transparently declaring usage: which tool, when, for which tasks, and for which parts of the work. The UZH recommends that instructors must not require students to use paid or data-protection-relevant AI tools and emphasizes the students' obligation to be informed about allowed AI use in the respective course.
The University of Lucerne allows AI for idea generation or brainstorming but requires the indication of its use and critical verification of accuracy. The AI Homework Helper can assist but does not absolve from verification and responsibility.
Faculty documents, such as the Guidelines of the English Department of UZH , show that "using AI" is no longer equivalent to "having AI write".
The AI Homework Helper is an organizational and quality tool in Winter 2025/26. It accelerates research through systems like Google Scholar, , structures literature work with tools like Elicit , and helps to visualize paper networks, for example with Connected Papers. It facilitates source criticism when scite makes citation context and supporting/contrasting citations visible ( scite Blog). It smooths language with DeepL Write and LanguageTool , and keeps citations stable with Zotero .
At the same time, the environment has become rougher: detectors are not reliable enough to serve as judges, which even providers and universities openly discuss ( OpenAI, Turnitin, Vanderbilt). Those who work cleanly today document tools, check content in the original, and write transparently – exactly as universities repeatedly demand in their guidelines (" ETH, HSLU, UZH).