AI chat for researchers

Research involves a lot of reading, structuring and rewriting, and an AI chat can speed up each of those steps. It can summarize a dense section, suggest search terms, explain a method in plain language or tighten a paragraph. It cannot be trusted as a source: it may misstate findings or invent references that look real. ChatG is free with a daily allowance of messages and keeps your history in your own browser. Here are the workflows researchers use most, each with a prompt to open in the chat.

Reading and planning

Use the chat to get oriented faster, then go back to the original texts for anything you will cite.

Digest a dense paper section

Paste the text itself rather than asking about a paper by title, so the summary is based on what the authors wrote.

Summarize this paper section for someone familiar with [field] but not this subfield. Give the research question, method, key findings and the limitations the authors mention, each in two or three sentences: [paste section]

Try this prompt in the chat or use the Summarizer.

Sharpen a research question

A broad topic becomes workable when you see a few focused versions side by side.

My research topic is [topic]. Help me narrow it into three focused, answerable research questions. For each, suggest a suitable method, the kind of data needed and one likely challenge.

Open this prompt in the chat.

Plan a literature search

Ask for keywords and search strings, not for citations. You then run the search in real databases.

Suggest search keywords, synonyms and Boolean search strings for finding literature on [topic] in [database]. Include related terms from neighboring fields that I might miss.

Try this prompt in the chat.

Understand a statistical method

A small made-up example often makes assumptions and outputs click.

Explain [statistical method] in plain language: when to use it, what assumptions it makes, how to interpret the output and the most common mistakes. Use a small made-up example with [type of data].

Open this prompt in the chat.

Instruments, analysis and writing

Later in a project, the chat helps with drafts you will review closely.

Draft survey or interview questions

Ask the chat to flag leading or double-barreled wording, a common problem in first drafts.

Draft [number] survey questions to measure [construct] among [population]. Mix Likert-scale and open questions, avoid leading wording, and flag any question that could be double-barreled.

Try this prompt in the chat.

Write analysis code

Describe the columns, not the data itself, and ask for explicit handling of missing values.

Write [R or Python] code to [analysis task] on a dataset with these columns: [column names and types]. Add comments, handle missing values explicitly, and produce [plot or table].

Open this prompt in the chat or use the Code Assistant.

Edit academic prose

Ask for concision while keeping technical terms, and have the chat point out claims that need a citation.

Edit this paragraph from my [paper type] for clarity and concision in an academic style. Keep the technical terms, don't change the meaning, and list any claims that need a citation: [paste paragraph]

Try this prompt in the chat or check grammar with the Grammar Checker.

Prepare a conference talk

Turning a paper into a short talk is mostly about choosing one message and cutting the rest.

Turn this abstract into a 10-minute conference talk outline with one main message, 8 to 10 slides (a title and key point for each) and three questions the audience is likely to ask: [paste abstract]

Open this prompt in the chat.

Tips for better results

• Give the field and level. "A second-year PhD student in epidemiology" gets a different, more useful answer than "a researcher".
• Paste the text you mean. Summaries and edits based on the actual passage are far more reliable than answers from memory.
• Ask what is uncertain. Add "tell me which parts you are unsure about" to any explanation of a method or concept.
• Keep a record. Export important chats as Markdown so you can document how AI was used in your project.

What to avoid

• Citing references from the chat. AI chats can produce authors, titles and page numbers that do not exist. Find and read every source yourself before citing it.
• Pasting confidential or participant data. Unpublished results, interview transcripts and anything covered by an ethics approval or data agreement should stay out of the chat unless your protocol allows it.
• Letting it interpret your results. The chat can explain methods, but conclusions about your data are your responsibility and need your expertise.
• Ignoring disclosure rules. Journals, funders and universities have policies on AI use in research and writing. Check them and disclose as required.
• Trusting recency. The chat may not know recent publications or changes in a field. Search current literature directly.

Related tools, guides and prompts

• Guides: fact-checking AI answers, how AI chatbots work and AI privacy and safety.
• Tools: Summarizer, Paraphraser, Translator for foreign-language papers and Code Assistant.
• Prompt collections: study prompts, writing prompts and productivity prompts.