Analysis Chat Overview

What is Analysis Chat?

Analysis Chat is Scholara's AI-powered conversational interface for data extraction and analysis. Instead of manually filling in extraction forms, you interact with an AI assistant that has read all of your included studies. You can ask it to extract data, generate tables, create plots, run calculations, and even help draft sections of your review — all through natural language conversation.

What You Can Do

  • Extract specific data points from any included study by asking in plain language
  • Generate structured data tables across all your studies (e.g. characteristics tables, outcome tables)
  • Create publication-ready forest plots and funnel plots
  • Run calculations and statistical analyses via code execution
  • Search the web for supplementary information or methodology guidance
  • Upload additional files or documents for the AI to reference
  • Generate your PRISMA flow diagram from the screening data
  • Export any generated table or plot in multiple formats

How It Works

When fulltext screening completes, the assistant transitions into the Analysis phase automatically. Scholara processes your included studies — reading and indexing the full-text PDFs — so the AI can search them instantly when answering your questions. From the user's perspective, you stay in the same chat: the phase indicator at the top simply moves to Analysis, and the kinds of requests the assistant can handle expand to include data extraction, tables, plots, and code execution.

  1. 1When the chat reaches the Analysis phase, ask the assistant for what you need in plain language (e.g., "create a characteristics table for all included studies")
  2. 2The AI reads your studies, runs any code it needs, and streams the result back into the chat
  3. 3Tables and plots appear inline as interactive asset cards
  4. 4Open assets in the side drawer for a closer look, copy them, or export in the formats you need

Understanding the Interface

Chat Panel

The main area where you converse with the AI. Your messages appear on the right, AI responses on the left. The AI can show its reasoning process, display generated tables and plots inline, and cite specific studies with clickable references.

Asset Cards

When the AI generates a table, plot, or code output, it appears as an interactive card within the chat. Cards can be opened in the artifact side drawer for a larger view, copied, or exported in the formats supported by that asset type (typically CSV/Excel for tables; PNG and source-script formats for plots).

Study References

When the AI references a specific study, it displays a clickable PMID badge. Clicking the badge opens the paper review sheet so you can verify the source against the full PDF — the AI's citation lands on the relevant passage, which is highlighted automatically.

AI Capabilities

Study PDF Search

The AI has indexed all included study PDFs and can search them to find specific information. It cites the source study for each data point, allowing you to verify the extraction.

Code Execution

The AI can write and run Python code to perform calculations, statistical analyses, and data processing. Code execution results (including generated plots) appear directly in the chat.

Web Search

The AI can search the web to find supplementary information, methodology guidance, or reference materials. Web sources are cited with clickable links.

File Uploads

You can upload additional files (PDFs, spreadsheets, images) for the AI to reference during the conversation. This is useful for providing supplementary materials, existing extraction data, or reference documents.

Getting Started

Here are some good first messages to try when starting your analysis:

  • 'Create a table of study characteristics for all included studies' — generates a comprehensive characteristics table
  • 'What outcomes were measured across my included studies?' — summarizes outcomes across the review
  • 'Extract the sample sizes and intervention details from each study' — pulls specific data points
  • 'Generate a forest plot for [outcome]' — creates a publication-ready forest plot
  • 'Summarize the key findings from my included studies' — provides a narrative overview

Best Practices

  • Be specific about what data you need — "Extract mean age, sample size, and intervention duration from each study" works better than "Get the data"
  • Always verify AI-extracted data against the original PDFs, especially numerical values — every value the AI produces links back to the source paper for one-click verification
  • Use follow-up questions to refine results — the chat maintains full context, so you don't have to re-explain your protocol or the studies
  • Export tables and plots as you go — they serve as both analysis tools and backups
  • Ask the assistant to summarise progress when starting a new session — "recap what we've done so far" — to get oriented quickly

The Analysis Chat replaces traditional form-based extraction with a more flexible, conversational approach. You can extract any data point simply by asking, without needing to pre-configure columns or forms. The AI handles the complexity of searching across all your studies and presenting results in structured formats.