How we select, check and update research resources

GAI for Research helps researchers learn practical uses of AI, follow the evidence behind a recommendation, and keep control of their research decisions. This page sets the editorial standard for new and substantially updated resources from 7 November 2026. Older pages are being reviewed progressively.

Start with a research task

We select resources for a concrete job: finding literature, checking a citation, analysing data, developing qualitative interpretations, improving writing, or documenting AI use. A useful guide explains its inputs, steps, expected output and verification checks. Tool popularity alone does not establish research quality.

Start with our research learning paths, explore the Toolkit, or browse practical tutorials.

Make the level of checking visible

Executed example. We identify the data, code or procedure actually run, the expected result, and the checks performed. The CSV-analysis exercise includes synthetic data, reference code and tests. Passing those checks supports that small example; it does not validate every use of an AI tool.

Source-checked guidance. We check relevant claims against the named papers or official documentation and describe what remains untested. The Inciteful exercise includes editor-selected relevance decisions. It does not claim those papers were returned by a live search.

Curated resource. We identify the original creator, link to the source, explain who it may help, and add an exercise or a reason to read or watch. We state the extent of our review when it matters. Linking a recording does not imply we reproduced its results or watched every minute.

Historical archive. We label older interfaces and discontinued workflows, preserve useful context and original media where appropriate, and direct readers to current guidance.

Keep evidence and interpretation traceable

Research summaries should identify the source, publication date and version. We distinguish published articles, preprints, working papers and provider documentation. A study's setting, sample and limitations determine how far its findings can travel. We link to supporting material and mark unresolved claims clearly.

For tool descriptions, current official documentation supports feature claims. Prices, usage limits, availability and account-specific data policies can change. Check the provider before choosing a paid plan or uploading research material. We avoid numerical rankings or “best” claims without a stated comparison method.

Credit creators and disclose relationships

Our curation uses original summaries, source links and permitted embeds. Full republication needs permission or a licence that permits it, with its conditions followed. A public webpage or video is not automatic permission to copy it in full.

Shubin Yu is the founder of this initiative and is involved in QualiTaTi and MimiWork. Resources involving those products should carry a relevant relationship disclosure. Any sponsored placement or affiliate relationship should be identified where it appears. Tool inclusion is an editorial selection, with evidence and limitations open to inspection.

Show how AI was used

AI assistance is used in drafting, translation and site maintenance. A page's checks should be specific: source comparison, calculation, code execution or visual inspection. Readers should not infer independent peer review or human-only authorship from an AI-assisted page. Scholarly responsibility for a research project remains with its researchers.

Worked datasets are labelled synthetic when they are invented for teaching. Reusable prompts ask for evidence locators and explicit missing information. Research files, participant data and unpublished manuscripts require an appropriate account, permission and data-handling decision before upload.

Correct errors and maintain useful pages

Use the contact page to report an error, broken link, outdated step or accessibility problem. Include the page URL, the passage or step, and a source or reproduction detail if available. Share a minimal example that excludes confidential or participant information.

Substantive revisions should show their review date and explain a material correction. We preserve useful URLs and distinguish the original publication date from later review. English, Norwegian and Chinese versions may be updated at different times; check the review date on the version you use.

Contribute a tutorial or recommend a resource

Send a focused proposal through the contact page: the research task, intended reader, prerequisites, source links, a worked example or expected output, checks performed, limitations, and any relevant commercial relationship. For a video, include the creator and original link, plus useful timestamps if you have reviewed them. For code, include a small reproducible example and clear reuse terms.

Our priority is material a researcher can try, inspect and learn from. A stronger library comes from clear evidence, useful exercises and visible corrections.