GatwAI

SciSpace

SciSpace combines PDF chat, literature research and AI agents. Review workflows, conflicting credit descriptions, data handling and documented limitations.

Provider sources checked

What does SciSpace do?

SciSpace combines scholarly search, conversations with PDFs, structured extraction and writing tools. Its agent can connect several steps, while standalone tools remain distinct. This is useful when initial sources need to become a traceable comparison table or a reviewable draft. The breadth makes it easier to move between reading and preparation, but does not provide scientific validation. A clear explanation can oversimplify an equation or lose a methodological qualification. Automated reviews particularly need documented scope, selection and evidence checks. Writing features do not remove the rules of your institution or publication.
Editorial research · 2026-11-14How we research →

Who is this tool for?

  • Researchers and students actively reading and verifying papers.
  • Teams needing repeatable extraction and clear boundaries for automated work.

Strengths

  • PDF comprehension and literature work share one product family.
  • Structured extraction supports a user-defined comparison matrix.
  • An agent can prepare multi-step tasks with visible usage accounting.

Limitations and requirements

  • Agent consumption depends on the task; retries can consume credits again.
  • Published credit tables partly conflict.
  • Readable wording guarantees neither correct citations nor complete literature coverage.

Who is behind the product?

Company
PubGenius Inc. (SciSpace)
Headquarters
Milpitas, California, United States: published contact address, not independently confirmed operating headquarters.
Company founded
The Typeset predecessor began in August 2015 according to a team post; this is not a separately verified PubGenius incorporation date.
Product launch
The reviewed sources do not establish an exact initial launch date for today's AI platform.

Earlier work on a research editor developed into a broader research offer. Typeset's early history is not the age of every current AI feature.

Suggested ways to use it

Suggested workflows or verification steps. These are not hands-on tests; for unverified products they are not a recommendation to use the service.

Suggested workflow: understand a method

  1. Upload an approved paper and inspect its import.
  2. Request an explanation of only the methods section, then ask about variables and assumptions individually.
  3. Compare the explanation line by line with the original, tables and relevant supplement.

Goal: Your own checked reading notes; not a test performed by GatwAI.

Suggested workflow: prepare a small literature matrix

  1. Choose five known studies and fixed columns such as sample, measurement and limitation.
  2. Explicitly constrain the agent to those sources and fields.
  3. Read cited passages, mark missing values and inspect consumption before expanding.

Goal: A verified small matrix as the basis for further research.

Pricing, allowances and data

What should you check before choosing a plan?

Basic is free. The guide lists Premium at US$20 monthly or $12 with annual billing, Advanced at $90/$70 and Max at $200/$160. Pricing lists 1,200 Premium credits while the guide describes unlimited Lite use subject to fair use; older help has other figures. Confirm binding model, credit and fair-use limits before purchase.

What should you check about data and permissions?

Help describes uploaded PDFs for free and paid accounts as private and excluded from its AI training. The linked April 2024 policy describes personal data, providers and purpose-based retention with exceptions. The PDF commitment does not establish zero retention for all agent interactions; sensitive materials require separate assessment.

External reviews and available evidence

A May 2026 JMIR original study found errors in SciSpace 1.5.1's handling of retracted papers.

Reported positives

  • Documented comparison; no conflicts declared.

Reported cautions

  • 2025 data, 15 selected papers and free access; not a current comprehensive product test.

These reports reflect the linked authors’ experiences. Test date, plan and task affect results; they do not establish a representative overall rating.

Product development and notable changes

  • Agent consumption explained in guidance

    The dated guide describes monthly usage, fair use, parallel tasks and separate agent consumption. Some figures conflict with other provider pages.

    Source for this change ↗

Which alternative fits which task?

Elicit →

Compare for structured study extraction and documented reviews.

Frequently asked questions

Does every feature consume agent credits?

The guide ties credits to agent tasks and treats standalone tools separately. Still check account-specific plan and model limits.

Is a generated literature review ready to cite?

Only after checking original sources, claim attribution and your scientific requirements.