Aibstracts
Start a review
Methodology & rigor

Fast is only worth it if the result is defensible

Aibstracts is built so a methodologist can interrogate every output and a peer reviewer cannot fault the trail. The speed comes from automation. The trust comes from transparency, established frameworks, and you in the loop.

  • Four layers, and only one of them is a model.
  • Published decision rules set the ratings, run as deterministic code.
  • You stay the reviewer of record and own the sign-off.
How the appraisal is actually produced

  1. Inputs

    Source PDFs, search results, and screening decisions.

  2. AI proposes

    the only model in the pipeline

    Reads each PDF and answers every signalling question, anchored to a quote.

  3. Published rules decide

    RoB 2 and ROBINS-I algorithms, run as deterministic code, not a model.

  4. Reviewer signs off

    Override any answer and the judgments recompute live. You own the call.

Result An auditable, source-anchored appraisal.

In detail

Built on methods reviewers already trust

  • Established appraisal frameworks

    RoB 2 for randomized trials and ROBINS-I for non-randomized studies, each run by its published algorithm and routed by study design. More frameworks on the way.

  • Structured protocols

    Build the question on PICO, PECO, SPIDER and more, with a Boolean strategy and MeSH hints from the start.

  • PRISMA provenance

    Your PRISMA flow built from the real trail, never reconstructed at the end.

  • Evidence Profiles

    Studies outside RoB 2 and ROBINS-I get a structured profile, never dropped.

  • Reconciliation

    Full-text conflicts with the abstract stage are surfaced, not buried.

  • Predatory-source filtering

    Journals flagged at discovery against Beall’s list, Scimago, and Scopus, for your call.

Responsible AI

Built to the standard our field is writing for itself

Cochrane’s RAISE initiative (Responsible AI in Evidence SynthEsis) sets out what an AI tool must do to be trusted in a systematic review. Aibstracts is built the same way: the model proposes, published rules decide, and you sign off.

We treat RAISE 2 as a design brief, not a badge: aligned with its principles where the platform genuinely earns it, and honest about the rest. Protocol, search, screening, and appraisal are in production; synthesis support is still evolving, and we say so.

Generative AI

Reads the sources and proposes answers

  • Traceable Every answer is anchored to the exact quote in the source PDF.
  • Rule-bound Published RoB 2 and ROBINS-I algorithms set the verdict, not the model.
  • Human-signed You are the reviewer of record and own every sign-off.
  • Honest by default Flags what it cannot ground in the source instead of inventing it.

AI output is never the final word.

Trust & security

Enterprise-grade where it counts

Your reviews, documents, and decisions stay yours. Data is encrypted in transit and at rest, access runs through single sign-on with short-lived sessions, and your content is never used to train AI models.

Built on managed, reputable infrastructure with a full decision audit trail. For procurement, we can provide a sub-processor list and data processing terms, and walk your security team through the architecture under NDA.

Your workspace
  • Reviews
  • Documents
  • Decisions
Yours, and only yours
  • Encrypted In transit and at rest
  • Single sign-on OIDC, short-lived sessions
  • Audit trail Who decided what, and why
  • Never trained on Your content trains no models
For teams whose evidence has to hold up

Put the methodology to the test

Run a review you have already done by hand and compare the appraisals study by study. Every judgment traces back to the quote behind it.

Book a demo

A free plan is available, no card required