Questions, answered straight

An AI evidence platform that runs systematic reviews end to end. It drafts a PRISMA-structured protocol, searches the literature, screens records, and performs risk-of-bias appraisal with RoB 2 and ROBINS-I. You get a verified, traceable evidence base with every signalling answer linked to a source quote.

Regulated evidence teams first: medical device clinical evaluation and regulatory affairs, HEOR and market access, and evidence synthesis groups whose reviews have to survive an audit or a notified body. Systematic review teams and medical writers use it for the same reasons. Academic researchers and graduate students use it too, across health, social science, and policy.

Yes. Explore is a free end-to-end evaluation: one bounded question covering 1,000 AI-assisted screening decisions and 12 full-text papers, with no card required and no time limit. Many teams use it to replay a review they have already completed and compare the results, study by study. Beyond that there are two options: a Review Project at $999, a one-time purchase that runs one substantial review from protocol through a source-linked evidence base, and an Organization agreement from $10,000 a year for teams running evidence work across multiple reviews. Full detail is on the Pricing page.

No, and that is by design. Aibstracts produces a defensible first pass, and you inspect, override, and sign off. The methodological accountability stays with you, which is exactly the point.

Cochrane RoB 2 for randomized trials, including the parallel-group, cluster, and crossover variants, and ROBINS-I for non-randomized studies of interventions. Studies that fit neither framework, such as animal, in vitro, qualitative, economic, or policy work, receive a structured Evidence Profile rather than being dropped.

Every signalling-question answer is anchored to a quote and its location in the source PDF. The domain and overall judgments are produced by the published RoB 2 and ROBINS-I decision algorithms, which are deterministic rules rather than an LLM guess. You can override any answer, and the judgments recompute instantly.

Yes. The most reliable test is to re-run a review you have already completed by hand and compare the appraisals study by study. Because every judgment traces to its signalling answers and the source quote behind them, you can see exactly where and why Aibstracts agrees or differs from your own call.

Every record carries its source provenance and every screening decision keeps its reason, so the PRISMA flow is generated from the real audit trail.

Aibstracts focuses on the protocol-to-appraisal pipeline and a structured, exportable evidence base. Synthesis support is still evolving, and we describe its current scope plainly during onboarding. We would rather under-promise here than overstate it.

Six sources, including PubMed, the Cochrane Database of Systematic Reviews, trial registries, and a grey-literature lane covering HTA bodies, regulators, guideline developers and preprints. Each is reached several ways at once rather than through a single query: the database search, citations followed forward and backward from studies you have already included, similar-paper recommendations, and records your team adds. Results are deduplicated, every record keeps the route it arrived through, and journals flagged as predatory are marked at discovery for the reviewer to judge rather than dropped. Embase, CENTRAL and Web of Science are not connected, and we have not measured recall against a gold-standard reference set, so we do not describe the search as comprehensive.

The managed infrastructure the platform runs on encrypts data in transit and at rest, sign-in runs through a dedicated identity provider with short-lived sessions, and your reviews, documents, and decisions stay yours. We maintain a register of the sub-processors that power the platform and share it on request, together with our security documentation. See our Privacy Policy for the detail.

No. Aibstracts does not train models on your content, and the model providers that process it run on paid API terms that exclude customer content from training. That is a contractual commitment rather than something we enforce in code, so we will put it in writing in the data processing agreement and name every provider in the sub-processor register. Providers may retain content briefly for their own abuse monitoring under those same terms; "not used for training" is not the same as "never stored", and we would rather draw that line ourselves than have you find it.

Export of your structured evidence base is supported today, and broader integrations are on the roadmap. Tell us which tools matter to you, since it helps us prioritize.

One-to-one onboarding, walkthroughs, and direct email support. We are small and responsive, and your feedback genuinely shapes the roadmap.

In the app, go to Account Settings and you can delete your account yourself at any time.

Use the contact link in the site footer for support, feedback, or methodology questions.