Step 1: Define Your Review Scope and Search

Start with a clear research question. For example: “What are the effective interventions for reducing student anxiety in online learning environments?” Enter this question in SciSpace’s search. The AI identifies relevant papers and explains why each matches your question — not just keyword matching, but semantic relevance analysis. You’ll typically get 15-30 candidate papers ranked by relevance.

Step 2: Screen Papers Using AI Explanations

Open each candidate paper and read the AI explanation. This 3-minute-per-paper screening replaces the traditional 20-minute abstract-and-skim approach. Look for papers that directly address your research question with empirical evidence. Discard theoretical-only papers, case studies with too few subjects, or papers about related but different topics. After screening, you’ll have 8-12 papers for inclusion.

Step 3: Select Papers for Comparison

Add included papers to your comparison workspace. Click “Compare” and select 3-5 papers at a time (SciSpace compares up to 5 papers simultaneously). Group them logically — for example, first compare all papers using cognitive-behavioral interventions, then compare mindfulness-based approaches, then compare technology-mediated interventions. This thematic grouping produces more meaningful comparisons than mixing all papers together.

Step 4: Review the Auto-Generated Comparison Table

SciSpace generates a structured comparison table with these dimensions: Research Question, Sample Size and Population, Intervention Type, Study Duration, Outcome Measures, Key Results, and Limitations. Each cell contains an AI-extracted summary from the corresponding paper. The table makes patterns immediately visible — for example, you might notice that studies with sample sizes over 100 consistently report moderate effect sizes, while smaller studies show larger but less reliable effects.

Step 5: Ask Synthesis Questions

Beyond the comparison table, ask synthesis questions in the chat panel: “What are the common findings across all these studies?” “Where do the studies disagree?” “Which intervention type shows the most consistent positive outcomes?” SciSpace analyzes across all selected papers to generate synthesis narratives: “Four of five studies using CBT-based interventions report statistically significant anxiety reduction (p < 0.05), with effect sizes ranging from 0.4 to 0.7. The one dissenting study used a shorter intervention duration (2 weeks vs. 8+ weeks in others).”

Step 6: Extract Methodology Details for Quality Assessment

Systematic reviews require methodology quality assessment. For each paper, ask: “Extract the full methodology details including study design, randomization procedure, blinding, and statistical analysis approach.” SciSpace provides structured methodology extraction that you can use for quality scoring (using frameworks like Cochrane Risk of Bias or CASP). Ask follow-up: “What are the potential biases in this study’s methodology?” to get AI-assisted bias identification.

Step 7: Export Your Review for Manuscript Writing

When your systematic comparison is complete, export the entire workspace — comparison tables, synthesis narratives, methodology extracts, and your Q&A history — as a structured document. SciSpace generates Markdown or PDF output formatted for inclusion in your review manuscript. The comparison tables become your review’s results section tables, and the synthesis narratives provide the foundation for your discussion section.

Pro tip: Use SciSpace’s citation graph alongside the comparison feature. When comparing two papers that reach different conclusions, check their citation networks — if one cites specific foundational works that the other doesn’t, this citation divergence may explain the conflicting findings and adds important context to your review’s discussion of discrepancies.

Step 1: Access SciSpace and Search for a Paper

Visit typeset.io (SciSpace’s web platform) and sign in. In the search bar, enter your topic, paper title, DOI, or author name. For example, search “transformer architecture attention mechanisms.” SciSpace returns relevant papers with relevance scores, citation counts, and brief AI-generated previews. Each preview gives you a 2-sentence overview before you commit to opening the full explanation.

Step 2: Open a Paper and Read the AI Explanation

Click on a paper of interest. SciSpace loads the paper and immediately generates a structured explanation panel with these sections: Research Question (what problem the paper addresses), Methodology (how they approach it, in plain language), Key Findings (the main results with context), Limitations (acknowledged weaknesses and constraints), and Significance (why this matters for the field). Read through this explanation — it takes 3-5 minutes compared to the 2+ hours of reading the original paper.

Step 3: Ask Targeted Follow-Up Questions

Below the explanation panel, there’s a chat interface for asking specific questions. Use this to probe areas the summary didn’t cover deeply: “What specific dataset did they use for training?” “How does their attention mechanism differ from the original Transformer?” “What were the exact accuracy numbers on the benchmark tasks?” SciSpace answers by referencing specific sections of the paper, often quoting relevant passages with citations to exact pages or sections.

Step 4: Clarify Technical Terms and Concepts

When the explanation mentions unfamiliar concepts, highlight the term or ask directly: “What is multi-head attention in this context?” or “Explain their use of positional encoding.” SciSpace provides targeted definitions that relate specifically to how the term is used in this particular paper — not generic Wikipedia definitions, but context-aware explanations that connect to the paper’s methodology and findings.

Step 5: Evaluate Paper Quality and Relevance

Ask evaluative questions: “What are the main criticisms of this paper’s approach?” “How novel is this compared to prior work?” “Is the sample size adequate for their statistical claims?” SciSpace assesses these dimensions based on the paper’s content and citation context. It might note: “The sample size of 50 subjects is small for the type of statistical analysis used, and two citing papers have noted this limitation.”

Step 6: Navigate to Related Papers

Use the citation graph panel on the right side of the paper view. It shows two clusters: papers this work cites (its foundations) and papers that cite this work (its influence). Click any connected paper to jump to its explanation. This lets you trace research lineages — follow citations backward to find foundational works, or forward to discover recent papers that extend or challenge this study’s findings. The graph displays relevance scores, helping you prioritize which related papers to explore next.

Step 7: Save and Organize for Later Deep Reading

Papers that pass your initial screening — those whose AI explanations show strong relevance to your research — should be saved for thorough reading. Click “Save to Library” and assign the paper to a project folder (e.g., “Thesis Literature Review — Attention Mechanisms”). SciSpace preserves the explanation, your Q&A history, and any annotations, so when you return for deep reading, you already have a comprehension foundation to build on.

Pro tip: When evaluating a paper for your literature review, ask SciSpace: “Generate a 3-bullet critical assessment of this paper’s contribution to [your specific topic].” This forces the AI to be evaluative rather than purely descriptive, giving you a quick relevance and quality judgment that helps prioritize your reading queue.