← series index
# AI for Researchers — A 7-Session Webinar Series
*Landscape as of August 2026*
## Series Pitch
AI is now part of the research lifecycle whether a researcher has adopted it deliberately or not — literature searches quietly favor semantic tools, editors quietly draft with a chatbot open, journals quietly write AI-disclosure policies. **AI for Researchers** is a seven-part webinar series that turns that casual, ad-hoc use into systematic, defensible practice. Across seven 60-minute sessions, academic researchers in any discipline will build a working mental model of how these tools succeed and fail, learn citation-grounded workflows for discovery, reading, writing, and data analysis, and finish with a personal map of the specialized AI tools transforming their own field. The series is fully tool-neutral: every topic surveys the current landscape of options — general chatbots, research-specific tools, and deep-research agents alike — so attendees leave able to choose the right tool for the task, not just the most familiar one.
## Audience & Prerequisites
- **Audience:** Faculty, postdocs, and PhD students in any discipline.
- **Assumed baseline:** Casual AI users — attendees have tried ChatGPT-style tools but use them naively (no systematic prompting, and little awareness of research-specific tools, publisher/funder policies, or common pitfalls).
- **Prerequisites:** None. No coding experience is assumed or required; the series is demo-based, not a hands-on coding workshop.
- **What you'll gain by the end of the series:** the ability to (a) explain how LLMs work and where they fail, (b) run a citation-grounded literature workflow, (c) use AI in writing within publisher policy, (d) apply AI to data and code with reproducibility guardrails, and (e) navigate the specialized AI landscape of your own discipline.
## Format
- **7 sessions, 60 minutes each:** ~20 minutes of literacy foundation + ~25 minutes of a live workflow demo + ~15 minutes of Q&A.
- Sessions build on each other, but every session opens with a 2-minute recap of the series so far, so drop-in attendees can follow without having seen prior sessions.
- Delivered as a self-contained slide deck (printable to PDF) plus a take-home handout for each session.
---
## Session 1 — AI Foundations for Researchers
**Abstract:** What is actually happening inside an LLM when it answers your research question — and why does it sometimes confidently invent a citation that doesn't exist? This session builds the mental model every later session depends on: how these models are trained, why hallucination is a structural property rather than an occasional glitch, and how to map the current tool landscape so you know what kind of tool you're reaching for. You'll leave with a working framework for judging when AI is safe to trust, when it needs checking, and when it shouldn't be used at all.
**Learning objectives:**
1. Explain, in plain terms, how large language models predict tokens inside a context window, and why hallucination is a structural property rather than a bug.
2. Explain how extended thinking works now that it is on by default — and what it still does not fix.
3. Classify a given AI tool as a general chatbot, research-specific tool, or deep-research agent within the 2026 landscape.
4. Apply a capability/failure map to judge whether a specific research task is safe, unreliable, or dangerous to hand to AI.
5. Apply systematic prompting techniques — context, role, constraints, iteration, output evaluation — to turn a weak prompt into a strong one.
6. Evaluate an AI-generated citation for signs of hallucination using a verification method.
**Topic outline:**
- Mental model of LLMs: training, tokens, context windows, why hallucination is structural, how default-on extended thinking works and what it still doesn't fix.
- Capability/failure map for research tasks: what AI is reliably good at, unreliable at, and dangerous for.
- The 2026 tool landscape taxonomy: general chatbots vs. research-specific tools vs. deep-research agents.
- Systematic prompting basics: context, role, constraints, iteration, output evaluation.
**Demo:** The same research question is asked badly vs. well across two tools, side by side — followed by a live demonstration of a hallucinated citation and exactly how to catch it.
**You will leave with:** the CRIT prompt template (Context / Role and register / Instructions and constraints / Task, then iterate), **The Capability/Failure Map for Research Tasks** (Safe / Unreliable / Dangerous), **The 2026 Tool Landscape Taxonomy** (Tier 1 — General chatbots / Tier 2 — Research-specific tools / Tier 3 — Deep-research agents), a hallucination-spotting checklist, and a further-reading list.
---
## Session 2 — Literature Discovery & Synthesis
**Abstract:** Your next relevant paper might not show up in Google Scholar — and the one AI hands you with confidence might not say what it claims. This session compares citation-grounded discovery tools (Elicit, Consensus, Semantic Scholar, SCiNiTO, Scite, ResearchRabbit) against deep-research agents, showing where each shines, where each misleads, and how to verify what either one gives you. You'll leave able to run a real literature search through both kinds of tools and check the results before you trust them.
**Learning objectives:**
1. Explain the difference between semantic and keyword search and why each surfaces literature the other misses.
2. Compare citation-grounded discovery tools (e.g., Elicit, Consensus, Semantic Scholar, SCiNiTO, Scite, ResearchRabbit) on coverage, grounding, and fit for a given task.
3. Run a literature search on the same question using both a citation-grounded discovery tool and a deep-research agent.
4. Evaluate deep-research agent output for coverage bias, missing paywalled literature, and unverified claims.
5. Apply a citation-verification workflow to confirm a claimed source actually supports the claim made from it.
6. Identify caution zones — systematic reviews, coverage bias, paywalled literature — where AI literature tools need extra scrutiny.
**Topic outline:**
- Semantic vs. keyword search; why AI search finds papers Google Scholar misses (and vice versa).
- Citation-grounded discovery tools: Elicit, Consensus, Semantic Scholar, SCiNiTO, Scite, ResearchRabbit.
- Deep-research agents (ChatGPT / Gemini / Claude deep research): what they do well, where they mislead, how to verify.
- Caution zones: systematic reviews, coverage bias, paywalled literature.
**Demo:** One real research question is run through a citation-grounded discovery tool and a deep-research agent side by side, with a live citation-verification pass on the results.
**You will leave with:** **The Literature-Discovery Tool Matrix**, a citation-verification checklist, prompt templates for both discovery tools and deep-research agents, and a further-reading list.
---
## Session 3 — Reading, Notes & Knowledge Management
**Abstract:** AI can chat with a PDF, answer questions grounded only in your own corpus, and pull a comparison table out of twenty papers you'd otherwise read by hand — but summaries can flatten the methodological nuance that matters most. This session covers AI-assisted reading and grounded notebooks (Gemini Notebook-style, formerly NotebookLM), and shows how to turn a stack of papers into structured, comparable extractions without losing the details that change a conclusion. You'll leave having built and interrogated your own grounded notebook.
**Learning objectives:**
1. Explain how AI-assisted paper Q&A / PDF-chat tools work and where summarization flattens methodological nuance.
2. Build a source-grounded notebook (Gemini Notebook-style, formerly NotebookLM) from a small set of papers.
3. Run structured queries against a grounded notebook to extract methods, effect sizes, and populations.
4. Apply AI-assisted extraction to build a cross-paper comparison table.
5. Evaluate when an AI summary requires returning to the primary source rather than trusting the synthesis.
6. Integrate AI reading tools into a reference-manager workflow (e.g., Zotero).
**Topic outline:**
- AI-assisted reading: paper Q&A / PDF chat; when summaries flatten nuance (critical-literacy segment).
- Grounded notebooks (Gemini Notebook-style, formerly NotebookLM): source-bound Q&A over your own corpus.
- Structured extraction across many papers (methods, effect sizes, populations → comparison tables).
- Integration with reference managers (Zotero and similar).
**Demo:** Building a grounded notebook from 5 papers live, interrogating it with follow-up questions, and extracting a comparison table across all 5.
**You will leave with:** a grounded-notebook setup checklist and prompt template, the Extraction Column Template, **The Summary-Trust Triage** (when an AI summary obliges you to reopen the paper), and a further-reading list.
---
## Session 4 — Writing, Publishing & Integrity
**Abstract:** Every major publisher now has an AI-use policy, and the rules genuinely differ — from outright prohibition to structured disclosure to permissive-with-transparency — but almost none of the researchers using AI to write are actually disclosing it. Measured against statistical evidence of AI-assisted writing, formal disclosure ran at roughly **40:1 underreporting** in 2025 Q1, the quarterly disclosure rate having risen only from 0.01% in early 2023 to 0.43% within the study's full-text sub-corpus (cumulatively, 76 of 75,172 post-2023 papers, ~0.1%) — from almost nothing to nearly nothing (He & Bu, *PNAS*, 2026; footnoted in the Session 4 deck and `sources.md`). This session maps the 2026 policy landscape (guidance from the ICMJE — the International Committee of Medical Journal Editors — the publisher spectrum, and the universal rule that AI cannot be an author), explains why AI-text detectors can't be relied on to enforce any of it, and walks through what a compliant workflow and a correct disclosure statement actually look like.
**Learning objectives:**
1. Explain the distinction between using AI as an editor versus a ghostwriter, and where each falls within discipline-appropriate norms.
2. Compare the publisher policy spectrum — from prohibition through structured disclosure to permissive-with-transparency — and locate a given journal on it.
3. Apply ICMJE-aligned rules to determine when and how AI use must be disclosed, given that AI can never be listed as an author.
4. Evaluate why AI-generated-text detectors are unreliable and should not be relied on for policy compliance.
5. Run a policy-compliant AI-assisted editing pass on a paragraph of academic writing.
6. Draft a correct, publication-ready AI-use disclosure statement.
**Topic outline:**
- AI as editor vs. ghostwriter; discipline-appropriate uses, including translation/polishing for non-native English speakers.
- The 2026 policy landscape: ICMJE guidance; the publisher spectrum from prohibition (Science/AAAS) through structured disclosure (Elsevier, Springer Nature, Wiley) to permissive-with-transparency (JAMA); the universal rule that AI cannot be an author.
- The disclosure gap as a teaching hook: most journals have AI policies, yet very few papers disclose use — about one formal disclosure for every 40 papers showing statistical evidence of AI usage (2025 Q1), with the quarterly disclosure rate rising from 0.01% to 0.43% between early 2023 and 2025 Q1 (He & Bu, *PNAS*, 2026; paired in the deck with Siler, *PNAS*, 2026: 57% of articles showing LLM influence by 2025).
- Why AI detectors are unreliable; peer-review confidentiality rules; funder policies.
**Demo:** A policy-compliant editing workflow applied live to a rough methods paragraph written for the session (never a real author's text), run twice — "suggest edits" versus "rewrite this" — and diffed against the original, followed by drafting a correct disclosure statement for it in one publisher's required format.
**You will leave with:** **The Publisher AI-Policy Comparison Table**, disclosure-statement templates, a policy-compliant editing-workflow checklist, and a further-reading list.
---
## Session 5 — Data, Code & Your AI Workflow
**Abstract:** AI can generate analysis code, suggest statistical approaches, and build visualizations — but only reproducible if you build in the right guardrails from the start. This capstone session covers AI-assisted data analysis (with and without code), agentic coding/analysis tools as they work today — and the gap between what they already do and what still needs you — and closes by helping you assemble your own personal AI workflow across the whole research lifecycle. You'll leave having run a dataset through an AI-assisted analysis and checked the result for reproducibility.
**Learning objectives:**
1. Explain the four reproducibility guardrails — keep the code, pin the versions, set the seed, verify one number by hand — needed when using AI for data analysis or code generation.
2. Run an AI-assisted statistical analysis or visualization task on a sample dataset.
3. Apply a no-code analysis path to a dataset without writing code directly.
4. Evaluate the output of an AI-assisted analysis for correctness and reproducibility before trusting it.
5. Assemble a personal AI workflow spanning the discovery, reading, writing, and analysis stages of the research lifecycle.
6. Identify strategies for staying current as agentic coding/analysis tools continue to evolve.
**Topic outline:**
- AI for data analysis: code generation, statistical assistance, visualization — with the four reproducibility guardrails (keep the code, pin the versions, set the seed, verify one number by hand).
- No-code analysis paths for non-programmers.
- Agentic coding/analysis tools in the present tense: what they already do, what still needs you.
- Capstone: assembling a personal AI workflow across the whole research lifecycle; how to stay current.
**Demo:** A dataset is taken live through AI-assisted analysis to a checked, reproducible result.
**You will leave with:** **The Four Reproducibility Guardrails** (keep the code · pin the versions · set the seed · verify one number by hand), analysis prompt templates, a no-code tool comparison table, and **The Personal AI Workflow Canvas** — a fillable one-page sheet covering the whole lifecycle, whose last two rows you complete in Sessions 6 and 7.
---
## Session 6 — AI for Multidisciplinary Research
**Abstract:** AI is unusually good at what researchers often struggle with alone: getting quickly oriented in an unfamiliar field. This session shows how to use AI to translate another discipline's jargon and methods, find adjacent literature and unexpected connections across fields, and synthesize evidence from disciplines with different methodological standards — plus how AI can support team science, from shared knowledge bases to finding collaborators. You'll leave having watched a full cross-field orientation workflow run start to finish.
**Learning objectives:**
1. Apply AI tools to translate an unfamiliar discipline's jargon, methods, and core literature into accessible terms.
2. Run a citation-graph or cross-field semantic search to find adjacent literature and unexpected connections.
3. Evaluate evidence from fields with different methodological standards for compatibility before synthesizing it.
4. Apply AI to build a cross-disciplinary synthesis memo combining evidence from multiple fields.
5. Explain how AI can support team science: shared knowledge bases, collaborator discovery, and cross-expertise communication.
6. Produce an annotated reading list to orient into an unfamiliar field.
**Topic outline:**
- Crossing field boundaries: using AI to translate unfamiliar disciplines' jargon, methods, and core literature.
- Finding adjacent literature and unexpected connections (citation-graph tools, semantic search across fields).
- Cross-disciplinary synthesis: combining evidence from fields with different methods and standards.
- Team science: AI in collaboration — shared knowledge bases, finding collaborators, communicating across expertise levels.
**Demo:** A researcher from field A uses AI to get oriented in field B, live: jargon translation → landscape map → annotated reading list → cross-field synthesis memo.
**You will leave with:** **The Cross-Field Orientation Ladder** (Translate → Map → Read → Synthesise), with a prompt template and a check for every rung, the Cross-Field Evidence Compatibility Check, row 6 of your Personal AI Workflow Canvas, and a further-reading list.
---
## Session 7 — Discipline Deep-Dives: Specialized AI by Field
**Abstract:** Beyond chatbots, specialized AI is transforming individual fields in ways most researchers outside that field never hear about — from protein-structure prediction to theorem provers to large-scale qualitative-coding tools. This closing session tours specialized AI across medicine, biotechnology and life sciences, engineering, physical sciences and math, and the social sciences and humanities, so every attendee leaves with a map of their own discipline's AI frontier — and a plan for tracking it as it keeps moving.
**Learning objectives:**
1. Identify the specialized, non-chatbot AI tools transforming your own discipline.
2. Explain at least one specialized AI application in each of medicine, life sciences, engineering, physical sciences/math, and social sciences/humanities.
3. Evaluate a specialized, field-specific AI tool relevant to your own research area.
4. Apply strategies to track your field's AI frontier — key venues, benchmarks, and review articles.
5. Compare specialized discipline-specific tools to the general-purpose tools covered earlier in the series.
**Topic outline:**
- Medicine & clinical research: diagnostic/imaging AI, clinical-trial matching, PubMed-grounded assistants, regulatory context.
- Biotechnology & life sciences: AlphaFold 3 and protein structure, protein/enzyme design, AI drug discovery.
- Engineering: simulation surrogates, generative design, CAD and engineering-code copilots.
- Physical sciences & math: materials discovery, AI in astronomy/physics pipelines, theorem provers and math assistants.
- Social sciences & humanities: large-scale text analysis, AI for qualitative coding, digital humanities, survey/agent-based simulation.
- Closing: how to track your field's AI frontier (key venues, benchmarks, review articles).
**Demo:** Two short deep-dives, chosen live by audience poll, run in field-specific tools.
**You will leave with:** a personalized map of specialized AI tools in your discipline, the Five Questions to ask of any field-specific AI tool, **The Field-Frontier Tracker** (venues, benchmarks, review articles), the completed final row of your Personal AI Workflow Canvas, and a further-reading list covering all five fields.