Let us show you around.
Have a look at our video tutorials or simply
Every week, a new AI tool appears promising to revolutionise how we search, sort, summarise, or synthesise information. If one believed the marketing copy, literature reviews should already be fully automated, effortless, and practically self-writing.
So why, in a world where AI features are bolted onto software at remarkable speed, does MemOwl still operate without an embedded AI engine? And—dare we ask—when will we implement one?
Let’s start with our position.
We use AI every day, we test it rigorously, and we fully appreciate the remarkable acceleration it can bring to certain parts of a research workflow. We see its potential in preprocessing, labelling, suggesting, rating, screening, and extracting data.
But we also see the other side of the coin: overlooked signals, hallucinated details, missing nuances, and a tendency—particularly in scientific contexts—to skip the very steps that matter most.
And in literature review work, especially where the stakes involve clinical decisions, policy recommendations, or public health, such oversights are not merely inconvenient.
They are unacceptable.
This is why MemOwl has been built, from the ground up, as a tool for people who insist on precision, transparency, and control.
Below are several reasons—technical, methodological, and ethical—why AI is not (yet) part of MemOwl’s workflow.
In fields such as medicine, health technology assessment, or pharmaceuticals, control is everything. A misplaced inclusion, a mislabelled study, or an imprecise extraction can distort an entire evidence base.
When we decided to launch MemOwl publicly, we committed to offering the maximum degree of control and reproducibility. Users should remain in charge 99.99999% of the time—not the model, not an opaque algorithm, not an automated shortcut.
Until AI can guarantee this level of steerability and auditability, we won’t pretend it can.
This is precisely why MemOwl focuses on speed, clarity, and a friction-free workflow—so that human expertise, not computer inference, remains the deciding factor.
Even the most advanced models can exhibit subtle but consequential biases:
Such distortions undermine the methodological standards that librarians, scholars, and evidence professionals have spent decades refining.
MemOwl’s design deliberately avoids these pitfalls by ensuring that you, not an algorithmic preference, decide what is relevant.
Literature reviews often involve:
Sending such content to external AI services—knowingly or unknowingly—introduces legal, ethical, and institutional complications. Many universities and pharmaceutical companies simply cannot, under any policy, allow sensitive content to leave their infrastructure.
Because MemOwl is used for projects where confidentiality matters, we refuse to compromise on this principle. A tool designed for professionals must treat professional data with professional care.
Systematic and scoping reviews rest on transparent, reproducible processes.
Yet AI models, by design, rely on complex internal representations that cannot be fully inspected, justified, or replicated. This undermines:
Until AI becomes explainable to the degree that a screening decision can be defended in a PRISMA flow diagram or in front of a review board, caution is not only warranted—it is necessary.
MemOwl’s no-AI approach ensures every decision is traceable to a human choice, not an opaque inference.
A growing number of literature-review tools are built with AI-assisted coding and then rely on AI to generate their outputs. This “AI loop” creates an additional layer of uncertainty:
Our experience with AI-generated software components has been, shall we say, mixed. The common issue? A lack of genuine control. This again reinforces why MemOwl favours deliberate engineering over automated shortcuts.
Possibly—when it can be done responsibly.
For us, that means:
And most importantly:
AI that does not interfere with your control.
When such conditions are met, AI features may appear in MemOwl. But they will come slowly, thoughtfully, and only after they have proven themselves reliable in domains where reliability is not negotiable.
And until AI can meet that standard, MemOwl will continue to prioritise what truly counts: clarity, trust, and control.
Don't hesitate to contact us if you have any questions.
