Perplexity — Tool Review — Pub-AI: AI in Science

What it does

Perplexity takes a natural language question, searches the web in real time, and synthesizes a cited answer — each claim linked back to a numbered source so you can see where it came from. Unlike asking a general-purpose AI chatbot (which answers from training data alone, with no live sources), Perplexity retrieves actual current documents and grounds its response in them.

For researchers, this makes it most useful in a specific phase: orienting yourself in an unfamiliar area before committing to a formal literature search. You can ask broad questions, get a conceptual overview with cited sources, identify the relevant terminology and key researchers, and surface a few landmark papers to start from — all faster than a cold start in PubMed or Web of Science.

Best for

Background orientation before a formal literature review. Building a vocabulary of field-specific terms so your database searches are better calibrated. Quick fact-checks on claims you’ve encountered. A conversational starting point when you need to understand a new subfield quickly before digging into primary literature.

Pricing

Freemium. The free tier covers standard web search and answer generation. Perplexity Pro (~$20/month) adds more thorough search, file upload, and access to more capable underlying models. The free tier is sufficient for most orientation tasks.

Strengths

Cites sources inline with numbered references — easy to follow claims back to their origin
Searches in real time, not just training data — surfaces recent papers, preprints, and news alongside older content
Conversational: you can follow up, narrow, or redirect without starting a new query
Covers academic content as well as web content, including arXiv, PubMed abstracts, and institutional pages
Fast for building a mental map of a field you’re new to

Limitations

Source quality is mixed. Perplexity cites web pages, news articles, and preprints alongside peer-reviewed journals without clearly distinguishing them — always check what type of source each citation is before treating it as established knowledge
Not a database replacement. Perplexity cannot replicate the controlled vocabulary, MeSH terms, and exhaustive coverage of PubMed, Web of Science, or Scopus. Use it for orientation, not for systematic search
Recency bias. Recent web content is overrepresented relative to foundational work from 10–20 years ago that isn’t heavily indexed on current web pages
Hallucination risk. Perplexity can misread or misattribute sources, and sometimes generates plausible-sounding citations that don’t match the actual document. Check linked sources for any specific claim before citing it
Not suitable for sensitive data. Queries are processed by a third-party service; don’t use it to look up or discuss unpublished results, confidential research, or data subject to confidentiality agreements

How it compares

vs.	Key difference
Semantic Scholar	Semantic Scholar searches only academic literature with a controlled index; Perplexity searches the broader web. Use Semantic Scholar for systematic literature discovery, Perplexity for initial orientation
Elicit	Elicit is for structured extraction across a large paper set; Perplexity is for open-ended background questions. Different phases of the workflow
Claude / ChatGPT	General-purpose AI chatbots answer from training data without live retrieval; Perplexity retrieves current documents. Better sourcing, but same caveat about needing to verify
Google Scholar	Google Scholar is exhaustive for academic search but returns raw results without synthesis; Perplexity synthesizes but trades comprehensiveness for speed

Tutorial: Using Perplexity for Background Research Before a Literature Review
Tool: Semantic Scholar — for structured academic discovery after your orientation
Tool: Elicit — for systematic extraction once you have a candidate paper set
Comparison: AI Tools for Literature Review