Hire an AI research evidence report writer

Everything it can hand back

Your task sets the actual outputs. The specialist picks the methods that cover them.

• A report with a source after each claim
• A ledger of every claim, its source, a short quote and its status
• A "Not sourced" list for what could not be settled in three rounds
• Points where sources disagree, with both sides

How it does the job well

• Pin down the exact question, scope and the standard of evidence needed before searching.
• Prefer primary sources, and say how strong each source is.
• Separate what is established, debated and unknown, with confidence levels.
• Cite specifically: author, work, date or page where possible.

Mistakes it avoids

• Saying a claim is true. Say what the source says and where.
• Counting a link you did not open as a source.
• Dropping a hard claim quietly. Move it to "Not sourced".
• Running a fourth round. Report what is open.
• Using this for a question you can answer from one known page. Just read the page.
• Presenting one source or a popular claim as settled.

What a good result looks like

An example, not your task: You need a report you can defend, and each claim must trace to a source.

• Report. The full write-up, with a source after each claim.
• Ledger summary. Counts of claims, sourced, dropped and still open, plus how many rounds ran.
• Ledger. One row per claim with its source, a short quote from the opened source, and its status.
• Not sourced. Each claim still open, why it is open, and the kind of source that would settle it.
• Where sources disagree. Each point where sources conflict, with both sides shown.
• Files and missing. The ledger and round files, then every MISSING line in one list.

How the work is checked

• Every sourced claim has a source you opened and a quote from it.
• The counts match the ledger rows.
• No more than three rounds ran.
• No claim was made firmer than its source.
• States which checks passed, failed or could not run, and names missing inputs.

These checks have not been run against a model yet. Your AI uses them to check its own work.

A second opinion

Before it hands back, your AI reviews the draft the way Rand Fishkin would. This is a way of reviewing the work. It does not mean they wrote or endorse it.

Capabilities

• Check for earlier work. Reads the topic's ledger if there is one, so a second run starts from the open claims.
• Draft the report. Runs the deep research skill to get a first cross-checked report with its sources.
• List the open claims. Runs the claim spotter. Each claim goes in the ledger, and counts as sourced only with an opened source and a matching quote.
• Choose a path for each claim. Science and health claims go to the paper search. A claim that rests on one study goes to the study checker. The rest get a narrow search.
• Update and loop. Adds sources, shows disagreement, rewrites or drops what has none, then checks again. It stops after three rounds.
• Hand back. Gives the report, the ledger counts, the Not sourced list and where sources disagree.

How the role works

1. Checks your task, context and available tools.
2. Picks the methods needed and prepares a work plan.
3. Carries out supported tasks, with each method's limits and review gates.
4. Checks the real deliverables and corrects supported faults.
5. Hands back files, evidence and unresolved gaps.

Your specialist's working setup

Download the base specialist without a task form or account. README.md and SETUP.md explain how to use the folder. SOUL.md defines its scope. INTAKE.md asks missing questions inside your AI. PLAYBOOK.md, SKILLS.md, BEST_PRACTICES.md, CHECKS.md and HANDOFF.md guide the work and its checked deliverables.

Complete associated skills, supporting files and licences are included. Choose only the methods needed for the actual task. Keep context, working files and accepted feedback in your AI project.

Use this specialist through MCP or the API

Call prepare_specialist with role_id cited-research-assistant only. No task is required to load it. Your AI asks what job you need done, what you want back and which relevant context it can use.

For task-led hiring, call find_specialist, explain the suggested specialists and let the user accept or change the team.

Complete ZIP: Download base specialist. API: GET /api/v1/specialists/cited-research-assistant/pack. The host AI supplies its own model, tools and execution; a pack does not connect accounts or run a task.

Inputs and helpful context

The specialist starts from the task brief. Only active methods require their inputs.

For the primary method, when used

• Your question, who the report is for, and the time period or place if it matters
• Papers, pages or notes you already have (optional)
• How strict to be, if you want something other than numbers, studies, quotes and named claims (optional)

Helpful context (optional)

• Papers or pages you already trust on the topic
• A note on who will read the report and what they will do with it

Tools and access

• Claude Code: It can keep the research folder and ledger between runs and open sources as files.
• A web search tool: Without search access it can only list the claims and the kind of source each one needs.
• Claude, ChatGPT or Gemini with web search on: It needs live search to find and open real sources.
• A built-in deep research mode, if your AI has one: It can run longer and read more sources. Still open the links yourself.

Functions and deliverables

• Research evidence report writer: Draft a research report, check each material claim against opened sources and show what remains unsourced after at most three rounds.
  • A report with a source after each claim
  • A ledger of every claim, its source, a short quote and its status
  • A "Not sourced" list for what could not be settled in three rounds
  • Points where sources disagree, with both sides
• Deep research reporter: Give it a research question. Get a sourced report that compares evidence, flags conflicts and lists remaining gaps.
  • A list of the sub-questions it will research
  • An executive summary
  • One section for each theme
  • Conflicts and gaps, with disputed claims labelled
  • A source list with link, author and date for each
  Source: Burhan Sebin.
• Unsupported claim spotter: Find the claims in your draft that need a source before you publish. Each one is ranked by risk, with the kind of source to look for.
  • A summary of how many claims need support
  • A table of claims ranked by risk
  • The kind of source that would settle each
  • Internal conflicts in your draft
  Source: Pro Skill Packs.
• Scholarly database searcher: Search PubMed, arXiv, OpenAlex and 15 more free sources for papers, citations and open full text. Every query is shown so you can repeat it.
  • A list of matching papers and preprints
  • Open access links where they exist
  • Citation data
  • The queries used, so you can repeat them
  Source: K-Dense.
• Study and claim evidence reviewer: Share a study or claim. Get a review of its design, bias, statistics and limits.
  • A list of biases and confounders
  • Notes on weak or misused statistics
  • An evidence-strength grade
  • A plain-language verdict
  • Questions to ask the authors
  Source: K-Dense.
• Research paper and code note writer: Share a paper, repo or messy notes. Get claims, methods and evidence sorted into a clear research folder.
  • A research folder with claims, methods and results
  • Every claim linked to its supporting table or log
  • Results tables copied with exact numbers
  • A record of dead ends and decisions
  • A short report on what was filled in and what was missing
  Source: Orchestra Research.
• Rand Fishkin: Plan audience research and pick marketing channels with Rand Fishkin's recorded approach. Find where your buyers pay attention and tie it to a goal.
  • Your problem worked on the way Rand Fishkin would: their questions, their method, their push-back
  • The answer or the work, in Rand Fishkin's voice, using their ideas by name
  • Anything that goes beyond their public record, marked
  • One suggested next step

Tasks you can use it for

• You need a report you can defend, and each claim must trace to a source.
• You have an AI draft and do not know which sentences are backed by anything.
• You are researching a topic over weeks and want each round to start from the last ledger.

Limits to check before starting

• You can answer the question from one page you already know: just read that page.
• You want a quick first draft with no source checks: ask your AI for a plain research summary instead.
• You need someone to publish or send the report: do that step yourself, since this agent never sends anything without a yes.

Best way to use this role

Give it one clear outcome, the actual files and a definition of done. Keep the work in one AI project so it can reuse checked context and earlier files.

• Start a chat with the agent. In Claude Code, add it as a skill or save it in .claude/agents, then give it your question.
• Give it the sources you already trust, so it starts there.
• Let it run. It will ask before it pays for access, logs in to a site or sends anything.
• Read the "Not sourced" list first. Decide which of those claims to cut and which to chase by hand.
• To update the report later, run it again on the same topic. It reads the ledger and starts from the open claims.

Read it as data

Read the complete role setup as JSON. Connect your AI to the AI Blocks MCP library and read aiblocks://roles/cited-research-assistant for its role instructions and pinned sources.

This is a prepared AI role. No task has run. Your AI carries out authorized work using its available tools and reports the actual deliverables and checks.