The Current and Future State of AI Literature Review Tools

At moara.io, we spend a lot of time talking about AI verticalization, which, in short, involves adapting broad AI models for very specific workflows and applications.

In the areas where verticalization has meaningfully scaled – such as software development and sales – a transition occurred from ChatGPT-like question-and-answer tools to process tools, where LLMs are used in conjunction with significant human intervention and iteration.

We believe academic research is going through a similar transition.

The State of AI in Literature Reviews

Despite being one of the clearest use cases for AI in academic research, literature reviews remain a task still predominantly performed with legacy software. Think, Microsoft Word to local filing systems.

Early AI literature review tools fit the mold of the first portion of this transition – chat engines ringfenced to query only academic papers. Users plug in questions and get sources and AI summaries in return. Many of these tools are excellent for search and quick discovery.

But few have made a dent in improving reference management, collaboration, or structured synthesis. It's difficult to imagine using these products to meaningfully manage a dissertation's literature review or a comprehensive systematic review.

The Next Stage: moara.io as an AI Workflow Platform

The new version of moara.io reflects the next stage of the AI-academic research transition – moving beyond simple search to a true workflow platform for research teams.

Video: moara.io Demo Reel

Below is a short overview of our work:

We created moara.io to complement – and in some ways counter – the AI research tools out there that attempt to automate away the literature review process through (in effect) question-and-answer engines.

Most ignore the fact that meaningful discovery happens at the edge of prior knowledge. The process of actually finding, organizing, sharing, reviewing, and synthesizing papers is a beastly one that, to be sure, needs to be refreshed in a big way.

With newly available LLMs, APIs, data sets, and modern design tools, we're building a workflow platform that helps research teams move through large literature reviews at lightning speed – collaboratively, transparently, and with a repeatable methodology.

How moara.io Works

In moara.io:

• Users decide on the papers they want to include in their review
• Users screen and set the priority levels of those papers
• Users perform a significant chunk of the full-text data extractions from their papers
• Users synthesize findings from papers, and
• Users decide what to do with those findings

The human is in the driver's seat.

That design philosophy – human-in-the-loop AI – guides every part of our product.

Hence, if you're looking for a bot to write your research assignment automatically, moara.io is NOT a good fit.

Who We Built This For

Researchers who will find moara.io most useful share a few core traits:

• They actually read through full-text articles as part of their work
• They use reference managers such as Zotero or Mendeley
• They collaborate with a team on literature reviews
• They treat literature reviews as a formal research method

Earlier tools in this space are excellent for generating a quick scan of the literature on a topic. Tools like moara.io are built for those conducting comprehensive, systematic literature reviews that demand rigor, transparency, and collaboration.

Video: Who moara.io Is For (And Why)

Collaboration as a Core Design Principle

We view collaboration as one of the most underserved areas in the AI research workflow ecosystem. Few-to-no newer AI tools offer a seamless process for teams to come together to tackle a review.

By (a) enabling review leads to add other users to their project, and (b) facilitating a detailed systematic review workflow, we can empower not only co-authors to work together but also instructors with their students, and librarians with their patrons.

Applying AI Thoughtfully

How and where to apply AI are questions we think about very carefully. To the extent possible, we design our LLM outputs to emphasize organization rather than automation.

We offer no models that draw conclusions from the literature – that remains on the user.

Instead, we focus our AI capabilities on workflow steps that benefit most from structure and efficiency:

• Building optimized search queries from your review title and description
• Extracting key metadata from each paper (e.g., relevance summary, study design)
• Generating a structured synthesis or catalogue of all papers, notes, and annotations

Integration as a Core Principle of Vertical AI Design

Building an end-to-end workflow tool means recognizing that we can't (and shouldn't) build everything. A key focus of our work at moara.io is integration – connecting with other best-in-class tools across academic search, reference management, and research collaboration.

Our most important integrations to date include:

• LibKey, enabling users to access full-text articles from their libraries in one click
• Zotero, allowing users to import existing libraries and papers
• Semantic Scholar, enabling natural-language search across a corpus of over 200 million papers

We aim to meet users where they are, to expand on the incredible innovations of our partners, and to make a lasting mark on the research ecosystem.