How Research Teams Configure Perplexity Spaces to Combine Live Web Search with Internal Documents

Research teams at investment funds, management consultancies, and corporate strategy groups frequently confront a frustrating workflow split. Proprietary insights—such as proprietary financial models, expert interview transcripts, initiation coverage decks, and due diligence memos—live in closed cloud repositories like Google Drive or Microsoft OneDrive. Meanwhile, fast-moving market realities—breaking regulatory filings, competitor product announcements, real-time pricing shifts, and supply chain disruptions—live on the open web. Manually toggling between an internal document store and search engines wastes hours and introduces blind spots.

Perplexity Spaces solves this disconnect by functioning as a unified research hub that binds internal document retrieval (RAG) with real-time web search. When configured correctly, a single prompt can cross-reference internal baseline projections against live market news, providing a reconciled synthesis supported by verifiable citations.

This guide walks strategic research teams through configuring Perplexity Spaces from scratch: connecting cloud storage providers, structuring document ingestion, drafting robust space-level system prompts, selecting the optimal underlying AI models, and establishing rigorous verification workflows.

1. Foundation: Perplexity Spaces Architecture and Account Tiers

Before configuring files, teams must understand how Perplexity handles multi-source retrieval and how subscription tiers impact data governance.

Core Architecture

Perplexity Spaces (an evolution of the platform's earlier Collections and Projects features) creates a shared, persistent workspace. Within a Space, all queries inherit:

• A shared file index (uploaded files or linked cloud drives).
• A unified system prompt (Custom Instructions) that enforces analytical personas, citation rules, and response formats.
• A default foundation model selected for the team's specific analytical rigor.
• Shared thread visibility for team collaboration.

When a researcher submits a prompt, Perplexity's retrieval engine parses the query, queries both the vector embeddings of internal documents and the live web search index, and feeds the merged context into the chosen LLM with precise source attribution.

Enterprise Pro vs. Individual Pro

For corporate research teams handling confidential memos, M&A due diligence, or proprietary investment theses, the tier selection is critical:

If your team analyzes sensitive non-public material, mandate Perplexity Enterprise Pro to guarantee contractually that internal files and user search queries are never used to train third-party foundation models.

2. Ingesting Internal Knowledge: Connectors and File Hygiene

A research space is only as reliable as the documents that ground it. Teams have two primary mechanisms for mounting internal intelligence into Perplexity Spaces: automated cloud storage connectors and direct file uploads.

Step 1: Authorizing Cloud Connectors (Google Drive & Microsoft OneDrive)

Cloud connectors allow Perplexity to query live cloud directories without requiring researchers to re-upload files each time a folder updates.

1. Navigate to Account Settings and locate the Connectors (or My Connectors) tab.
2. Select your organization's storage provider: Google Drive or Microsoft OneDrive / SharePoint.
3. Complete the OAuth authorization window using your corporate credentials.
4. Once connected, click the settings icon (⋮) beside the active connector to define specific file and folder scopes.
5. In Enterprise Pro setups, IT administrators can configure organization-level connectors via the Admin Security Hub, ensuring enterprise security policies and permissions are mirrored automatically.

Step 2: Direct Local Uploads for Curated Deal Dockets

For self-contained engagements (such as a 4-week commercial due diligence sprint or an activist campaign), creating a dedicated local file bundle inside the Space offers tighter scoping:

1. Open your designated Research Space.
2. Click Add Files in the Space settings sidebar.
3. Select local documents from your research drive. Supported formats include:
   • Analytical text & presentations: PDF, DOCX, PPTX, Markdown (.md), Plain Text (.txt).
   • Quantitative models & data: XLSX, CSV, JSON.
   • Multimedia transcripts: Audio/video files (.mp3, .mp4, .wav), which Perplexity automatically transcribes into text.

Step 3: Enforcing Document Preparation Hygiene

To avoid hitting file ceilings and indexing degradation, adopt these operational habits:

• Run OCR on Scanned Filings: Perplexity parses extracted text layers. If broker notes or legal exhibits are scanned images without OCR, the vector search will return empty results. Run OCR prior to ingestion.
• Merge Fragmented Files: Standard Pro accounts restrict spaces to 50 individual files. Instead of uploading 40 separate 3-page expert call notes, merge them into 2 thematic compendiums (e.g., Expert_Transcripts_Competitor_Landscape.pdf). Perplexity treats the merged file as a single document while indexing all pages.
• Prune Raw Financial Spreadsheets: Massive Excel models containing complex VBA macros or multi-tab link structures can create noise. Convert key financial outputs, historical income statements, and sensitivity tables into clean CSVs or streamlined summary sheets before upload.

3. Configuring Space-Level System Instructions (Custom Prompts)

The system instruction prompt defines the analytical standard for every thread created in the Space. Without explicit rules, general-purpose LLMs tend to generate superficial summaries and over-rely on broad web commentary.

Strategic Objectives of the Research System Prompt

A rigorous research prompt must accomplish four tasks:

1. Define Source Hierarchy: Command the model when to prioritize internal documents versus when to pull live web data.
2. Establish Contradiction Protocols: Dictate how the model should behave when live news contradicts an internal projection or historical thesis.
3. Mandate Analytical Tone & Frameworks: Direct the model to adopt the posture of a senior research analyst rather than an assistant.
4. Standardize Output Structure: Enforce structured comparison tables, bulleted takeaway summaries, and explicit margin/valuation callouts.

Production-Ready System Prompt Template for Research Teams

Copy and tailor this template into your Space's Custom Instructions field:

You are a Principal Investment & Corporate Strategy Research Analyst assisting an institutional research squad.

PRIMARY SOURCE HIERARCHY:
1. Always interrogate internal documents (uploaded files and connected drives) first to extract historical baselines, proprietary estimates, management commentary, and internal investment theses.
2. Concurrently query live web search to identify real-time developments, competitor earnings results, updated regulatory rulings, macro indicators, and consensus revisions.
3. If an internal forecast conflicts with live external data (e.g., internal revenue forecast vs. latest actual earnings report), do not blend or ignore the conflict. Explicitly highlight the divergence, quantify the delta, and detail potential drivers.

ANALYTICAL STANDARDS & TONE:
- Maintain an objective, institutional, and evidence-driven tone. Avoid marketing buzzwords, speculative filler, and vague generalities.
- When numbers are cited, always specify the metric, reporting period, unit (e.g., USD millions), and source context.
- Separate established historical facts from forward-looking estimates.

RESPONSE ARCHITECTURE:
- Executive Takeaway: 2-3 high-impact bullets summarizing the bottom-line conclusion.
- Reconciled Findings Table: Where applicable, provide a Markdown comparison table contrasting Internal Stance vs. Live Market Consensus.
- Detailed Evidence Breakdown: In-depth commentary explaining the strategic rationale, operational drivers, and risks.
- Evidence Footnotes: Explicitly identify the specific internal document title, slide/page reference, and corresponding web source.

4. Model Selection: Matching Engines to Research Workflows

Perplexity Spaces allows teams to choose their default foundation model, preventing researchers from having to reconfigure parameters on each prompt.

Team Recommendation: For strategic consulting, equity research, and private equity due diligence, set Claude 3.5 Sonnet (or newer Sonnet iterations) as the Space default. Its adherence to system prompts and restraint against hallucination make it the industry benchmark for document-grounded research.

5. Executing Dual-Source Inquiries: Search Scope and Prompt Tactics

Once your Space is populated with documents, prompts, and a designated model, researchers interact using Perplexity's dynamic source toggle.

Mastering the Search Source Selector

At the prompt input bar, researchers can control the scope of the retrieval engine:

• Web + Org Files (or All Sources): The default operational mode. Perplexity runs parallel searches across your indexed internal drive/files and the real-time internet index, blending both into a single answer.
• My Files / Org Files Only: Shuts off external web indexing. Use this when performing pure internal audits (e.g., "Summarize all risk factors mentioned across our internal CIM documents").
• Web Only: Bypasses internal files. Useful for broader macro queries that do not require internal context.

Formulating Dual-Context Research Prompts

To extract maximum value from blended retrieval, construct prompts that explicitly tie an internal reference to a real-time market question.

Example 1: Earnings Call vs. Internal Model Reconciliation

"Review our internal financial forecast in Project_Titan_Model_v4.xlsx regarding Q3 semiconductor margin assumptions. Then, search live web filings for TSMC and ASML's latest reported quarterly gross margins. Provide a comparison table displaying our internal forecast against actual reported numbers, explain any variances, and highlight supply chain commentary."

Example 2: Regulatory Change Impact on Internal Portfolio Thesis

"Based on our uploaded portfolio review memo CleanTech_Thesis_2025.pdf, what are our stated investment arguments for green hydrogen? Then, search the live web for the latest European Commission delegated acts on renewable fuels released in the past 6 months. Summarize whether recent policy adjustments strengthen or undermine our internal thesis."

Example 3: Competitive Intelligence and Pricing Shifts

"Extract the pricing benchmarks and tier structures documented in Competitor_Matrix_Q1.docx. Then perform a live web search on each competitor's official pricing page and recent product launch announcements. Identify any pricing increases, feature repackaging, or newly introduced tiers since our document was written."

6. Verification and Audit Workflows: Dual-Stream Citations

In high-stakes investment and advisory work, an unverified AI answer is a liability. Perplexity's primary advantage over closed chat models is its granular citation engine.

Decoding Dual-Stream Inline Citations

When Perplexity synthesizes a blended answer, every material claim receives an inline numerical citation badge:

• Internal Document Badges: Clicking the citation badge reveals the exact internal document title, folder path, and an excerpt of the extracted passage or page number. Analysts can verify whether the model correctly interpreted the context of the internal memo.
• Live Web Badges: Clicking the badge opens the external URL, showing the cached web snippet, domain, and publication date.

Institutional Verification Checklist for Analysts

Before incorporating a Perplexity synthesis into an investment memo, board deck, or client deliverable, enforce this 3-step audit:

1. Audit Internal Anchors: Click through all internal citations to ensure that numbers were not extracted from outdated sensitivity scenarios or draft iterations.
2. Verify External Publication Dates: Ensure live web citations originate from authoritative primary sources (e.g., SEC EDGAR, official corporate press releases, regulatory bodies) rather than unverified aggregator blogs.
3. Inspect the Divergence Gap: If the system notes an inconsistency between internal baselines and live data, confirm whether the discrepancy stems from differing accounting standards (e.g., GAAP vs. Non-GAAP) or genuine market shifts.

7. Strategic Governance and Best Practices for Research Squads

To scale Perplexity Spaces across a multi-tiered research division without creating organizational clutter, institute these operational policies:

Space Granularity: Project-Based vs. Industry-Based

Avoid dumping all company research into a single monolithic Space. Instead, organize Spaces around distinct analytical horizons:

• Deal / Engagement Spaces: Ephemeral Spaces dedicated to a specific live transaction (e.g., M&A Diligence - Target Alpha). Archive or delete the Space upon project completion.
• Coverage / Sector Spaces: Persistent Spaces maintained by sector coverage teams (e.g., Enterprise SaaS Coverage, BioTech Phase III Tracker). Regularly refresh the document folder with new earnings models and initiate coverage reports.
• Internal Knowledge Spaces: Fixed Spaces containing static institutional playbooks, valuation templates, and methodology whitepapers.

Managing Token Limits and Storage Boundaries

While Perplexity Enterprise handles generous document repositories, vector search precision degrades if repositories are clogged with duplicate or superseded file drafts. Designate an "engagement librarian" or lead associate responsible for:

• Removing superseded versions of financial models (retaining only the active golden copy).
• Maintaining clear, descriptive file naming conventions (e.g., 2026-03_MorganStanley_Semiconductor_Initiation.pdf instead of report(1).pdf).
• Reviewing connector permissions quarterly to de-index completed deals.

By treating Perplexity Spaces as a calibrated synthesis engine rather than a passive document folder, institutional research teams eliminate the friction between private intelligence and public market reality—unlocking faster, more rigorous, and fully auditable strategic insights.

References

• Perplexity AI - Wikipedia
• Introducing Internal Knowledge Search and Spaces - Perplexity Blog
• Perplexity Launches Internal Knowledge Search and Spaces to Combine Web and Local Data - Neowin
• Perplexity Spaces File Limit: Upload Caps, Supported Formats, and Storage Workarounds - Fast.io