Back to Reference
Contents
Reference
466 words
Jun 25, 2026

Research Review Studio

The Review Studio is a per-project workspace where supervisors, co-authors and the researcher discuss the work, and where you can stress-test a project against a simulated tough reviewer before a real one sees it.

Open it from a research project at Research > (your project) > Review Studio, or directly at /research/projects/{id}/review.

It has two halves. The comment half works with or without AI. The reviewer-simulation half uses AI and is always clearly labelled as machine-generated.

1. Comment threads

Comments let your supervisor and co-authors leave feedback that is anchored to the exact thing being discussed.

  • Anchor a comment to a claim so the discussion sits on the specific assertion it is about, or leave it project-level for general feedback.
  • Reply to any comment to build a thread.
  • Resolve a thread once it is dealt with (and reopen it if it comes back). Resolving a thread also resolves its replies, so a closed thread reads as one unit.
  • Filter by claim, and optionally hide resolved threads, to focus on what is still open.
  • Threads are ordered by time, giving you a lightweight revision history of the conversation.

This half needs no AI and never depends on the gateway being available.

2. Adversarial reviewer simulation

The reviewer simulation reads your project brief and your recorded claims and asks a language model to play a tough peer reviewer, so you can see the objections coming.

Pick a persona and run the simulation:

  • Methodologist - attacks sampling, validity, reproducibility, and whether the evidence actually supports each claim.
  • Theory purist - attacks the conceptual framing, definitions and theoretical grounding.
  • Statistician - attacks the quantitative reasoning, sample sizes and over-claiming from thin data.
  • Reviewer 2 - the maximally sceptical, hard-to-please peer reviewer.

The result is grouped into: major concerns, minor concerns, likely objections, required revisions, rejection risks, strongest contribution, weakest section, and missing literature. Every run is saved in the simulation history so you can re-open it later.

Important: this is AI, not a human reviewer

Every simulation output is labelled "AI reviewer - via the AHG gateway, not a human reviewer." It is generated by a language model and may be wrong, incomplete or overstated. Treat it as a prompt to stress-test your work, never as an actual peer-review outcome.

All AI calls route through the AHG AI gateway. If the gateway is unavailable you get a clear message and no run is saved - your comment threads are completely unaffected and stay usable.

Tips

  • Add a project brief (the project description) and record some claims first - the simulation has more to work with and produces sharper feedback.
  • Use comments to turn the simulation's findings into action: anchor a comment to the claim a finding targets, and resolve it once you have addressed it.