Skip to main content
Research

Perplexity Grounded LLM

Get AI-powered responses with real-time web grounding using Perplexity's language models

View details
Try it in Ampere

Inputs

Loading input fields...
Execution Steps

Loading workflow structure...

Loading curated examples...

Overview

Perplexity Grounded LLM turns conversation messages into an AI-written answer grounded in web results from Perplexity. Use it for current market questions, competitor context, trend snapshots, source-backed drafting, and research answers when you want synthesis rather than a raw search-result list.

Use cases

  • Ask for a concise answer to a current market, competitor, or category question.
  • Generate source-backed context before writing a brief, article, campaign angle, or founder note.
  • Constrain research to selected domains or recent results for a focused answer.
  • Request related questions or images when planning the next research step.

Input tips

  • Provide at least one user message; add a system message when the answer needs a specific role, scope, or format.
  • Use sonar for the default path, or choose sonar-pro, reasoning, reasoning-pro, or deep-research models for harder questions.
  • Use searchDomainFilter to focus sources and searchRecencyFilter for month, week, day, or hour freshness.
  • Adjust maxTokens, temperature, topP, topK, presencePenalty, and frequencyPenalty only when output style needs tuning.
  • Turn on returnImages or returnRelatedQuestions only when those extra outputs will help the research task.

Expected output

The AI Tool returns a web-grounded response, model, optional finish reason, citation URLs, detailed source results with URL, title, snippet, and date when available, optional related images, optional related questions, token usage, and cost metadata. The output view shows the answer, citations, collapsible source cards, optional image grid, related questions, token usage, and model.

Caveats

  • The answer is synthesized from available web results and should be reviewed before use in public claims.
  • Citations and search snippets can be partial, stale, or unavailable depending on the query and source access.
  • Domain and recency filters can improve focus but may exclude useful sources.
  • Images and related questions are optional outputs and may not be returned.
  • Reasoning or deep-research models can take longer than the default model.

Related AI Tools