ReviewAid

ReviewAid suits evidence-synthesis researchers who want configurable AI support and will keep people responsible for review decisions. Its appeal is a free, open-source tool that can run with a cloud provider or locally; its limits are small submission batches and the need to verify its work.

Overview

For systematic reviews, ReviewAid screens full-text papers against user-defined PICO criteria and extracts data into fields chosen by the user. Confidence scores and low-confidence flags help reviewers identify work that merits closer attention, but the maker positions the tool as a supplement, not a replacement, for manual screening and extraction. Reference deduplication is supported; blinded screening is not.

The Apache 2.0 license and local-running option give researchers flexibility over how they use the software. With a cloud provider, paper text is sent to that provider for the run. The maker says API keys are not stored. With Ollama, inference runs on the local machine and paper text stays there, a meaningful option for users who do not want to send papers to an external provider.

Key features

Screening, extraction, and exports

Users provide PICO inclusion and exclusion criteria for full-text screening and define the data fields to extract. This accommodates review-specific questions, but puts the onus on researchers to configure criteria and fields carefully and check the results. Results can be downloaded as CSV, XLSX, or DOCX, which gives teams several formats for continuing their work outside the tool.

Confidence and batch limits

Confidence scores and flags for low-confidence results can help direct manual review; they are cues for checking, not assurances that other results are correct. Batch sizes are modest: the README limits extraction submissions to 20 articles and full-text screening submissions to 10. That makes ReviewAid more suitable for bounded batches than for processing a large review in one submission.

AI providers and local use

Supported providers include OpenAI, Anthropic, DeepSeek, Cohere, Z.ai, and local Ollama. ReviewAid can run locally with Streamlit on Windows, macOS, or Linux, using Python 3.12 or later. The browser-based Streamlit app avoids installation, though it may take about 30 seconds to initialize after a cold start. Choosing a cloud provider means sending paper text to that provider; Ollama keeps inference and paper text on the local machine without an external API key.

Pricing

ReviewAid is free: its Free plan costs nothing and is open source under Apache 2.0. There is no paid plan to weigh against the limits; users should instead account for the 20-article extraction and 10-article screening caps per submission. The supported Ollama option can avoid an external API key, while cloud-provider use sends paper text to the selected provider.

Platforms

ReviewAid is available on the web and can be run locally on Linux, macOS, or Windows. Local installation uses Python 3.12 or later and Streamlit. The hosted browser app is more convenient to start with, but may have cold starts of about 30 seconds.

Who it's for

ReviewAid is a good fit for researchers, especially those conducting evidence synthesis, who want to configure full-text screening and extraction around their own criteria and fields. It also suits users who value an open-source license or want local inference through Ollama. It is a weaker fit for teams that require blinded screening, need larger per-submission batches, or want AI output to replace human review.

Pros and cons

• Pros: Free and Apache 2.0-licensed, so researchers can use an open-source tool without a subscription.
• Pros: User-defined PICO criteria and extraction fields adapt screening and extraction to a review's questions.
• Pros: Local Ollama inference keeps paper text on the machine and needs no external API key.
• Pros: CSV, XLSX, and DOCX exports give reviewers choices for working with results elsewhere.
• Cons: Full-text screening is limited to 10 articles per submission, and extraction to 20, constraining batch size.
• Cons: Blinded screening is not supported, which rules it out for workflows requiring that feature.
• Cons: Cloud runs send paper text to the selected provider, so users who want to keep text local need to choose Ollama and run it locally.
• Cons: The hosted app may have cold starts of about 30 seconds, a small but real delay for browser-based use.

Alternatives

Systematic Review Software is a useful place to compare more tools for this workflow. Choose ASReview instead if you want another free, open-source option and can run it on your own hardware or server. RevMan is an alternative with a free plan and free trial. Universal SLR Assistant is another free, open-source option, though cloud AI provider API fees may apply. Bango is free for researchers with no subscription, paywall, or limits. Rayyan may fit teams that prefer a freemium service, but its free plan caps active reviews at three and reviewers at two. Catchii is a free web option with multiple users supported. Covexe offers a free plan with an AI co-pilot capped at 100 messages per day. HAWC is aimed at exploring public assessments; creating assessments requires an account request.

Verdict

Choose ReviewAid if you conduct systematic reviews and want free, configurable screening and extraction assistance, especially with local inference available through Ollama. Its strongest reason to look elsewhere is equally clear: it does not replace reviewer judgment, has small batch limits, and cannot support blinded screening.