Your first research

This guide walks through one full research cycle: open the composer, write a strong question, choose providers and budget, pick a mode, submit, and read the results -- including the merged synthesis, each provider's report, and the structured claims view.

Open the new research dialog

From Research (/threads), start New Research (or use the start-new-research card at the top of the page). You'll see a composer where you type your question and configure the run.

If you're already typing in the main field, you can submit with the keyboard when the app shows the hint: Cmd+Enter (use Ctrl+Enter on Windows). When your query is long enough to refine, Cmd+Shift+Enter (Ctrl+Shift+Enter on Windows) can trigger query refinement -- optional, but useful as you get comfortable.

Write a good research query

Short, vague prompts produce shallow answers. Aim for questions that name the topic, the scope, and what "done" looks like.

Tips:

• Be specific -- "Summarize CRISPR delivery advances in 2024-2025 for liver disease, with clinical trial phase called out" beats "Tell me about CRISPR."
• Say what you need -- Mention format (comparison, timeline, pros/cons) if it matters.
• Constrain when helpful -- Region, industry, regulation, or time window reduces noise.
• One main question -- If you need several angles, plan follow-on research (below) instead of stuffing everything into one prompt.

Choose AI providers

Parallect can run Perplexity, Gemini, OpenAI, and Grok in parallel so no single vendor's blind spots dominate the answer. Each provider contributes its own deep-research pass; you'll see their outputs separately as well as a merged synthesis.

Pick the mix that fits the question -- more providers generally mean broader coverage and more cross-checking, at higher spend for the same tier.

Choose a budget tier

Budget tiers cap how much a single job can spend, from XXS (about $1) through XL (about $60). Higher tiers allow more depth and typically involve more providers. Use the picker below to see how tiers relate to cost and capacity -- your in-app composer uses the same scale.

Fast vs methodical mode

• Fast -- Quicker turnaround; best when you need a solid answer soon and the question is well bounded.
• Methodical -- Deeper, more thorough passes; better for high-stakes or ambiguous topics where extra analysis pays off.

Mode interacts with tier and providers: methodical runs are typically heavier, so balance your time budget and credit accordingly.

Submit and watch progress

After you submit, the job moves through clear stages -- roughly:

• Pending -- Queued and preparing to run.
• Running -- Providers are executing their research passes.
• Synthesizing -- Your unified report is being prepared.
• Completed -- Results are ready to read (and per-provider tabs are populated).

Use the thread view to come back anytime; long jobs don't require you to keep the tab focused.

Read the results

When the job completes, you'll see:

• Synthesis -- The merged report: one narrative that combines and reconciles provider outputs, with citations you can trace.
• Per-provider tabs -- Each model's standalone answer, useful when you want to see who said what before reading the synthesis.
• Claims matrix -- Structured claims with agreement or tension across providers -- ideal for spotting controversy, uncertainty, or consensus.

Suggested follow-on research

Parallect often suggests follow-on topics -- next questions that extend the thread logically. Pursuing a follow-on keeps context in one place and avoids restating background in a brand-new prompt.

---

You've completed the onboarding path. When you're ready to go deeper -- finer control over threads, refinement, tiers, and modes -- continue with the Research section, starting with creating threads and related guides.