Tech, ExplainedGaming, streaming and media
Product

Academic writing assistant

Helping non-native researchers polish the paper.

BEFORE ITAFTER ITAcademicwriting assistantCactus CommunicationsEnagoCOVERSGrammar checksJournal rulesAI rewritingHuman editingLarge language model (LLM)↗ GENERATIVE AI APPLICATIONSsuggeststhe rewritessends thefinishedpaperJournal publishing platformCONCEPTS IT USESAI-generated contentBEFORE ITAFTER ITAcademicwriting assistantCactus CommunicationsEnagoCOVERSGrammar checksJournal rulesAI rewritingHuman editingLarge languagemodel (LLM)↗ GENERATIVEAI APPLICATIONSsuggeststhe rewritessends thefinishedpaperJournalpublishingplatformCONCEPTS IT USESAI-generated contentBEFORE ITLarge language model (LLM)↗ GENERATIVE AI APPLICATIONS↓ suggests the rewritesAcademic writing assistantCactus CommunicationsEnagoGrammar checksJournal rulesAI rewritingHuman editingAFTER IT↓ sends the finished paperJournal publishing platformCONCEPTS IT USESAI-generated content
The academic writing assistant marks grammar and style problems in a researcher's draft, checks it against the journal's rules, and the polished paper is submitted to the journal.
opens a page inside this one a page in the same scene a page in another scene a concept page a label only, with no page

An academic writing assistant is software or a service that helps researchers write their papers in clear, correct English for a journal. Most of its users are researchers whose first language is not English, in India, China, Japan, Europe and elsewhere. It comes in two kinds. Grammar tools are AI software that checks each sentence as the researcher writes. Editing services are human editors, often with a degree in the same subject, who correct a whole paper for a fee based on its length.

Why do researchers need a special tool? Scientific English has its own rules. Many journals write "the data were", not "the data was". A paper describes its methods in the past tense. It must not claim more than its results show. A general grammar checker does not know these rules, or the special words of each subject. An academic writing assistant is trained on a very large number of published papers, so it suggests the phrasing that journals expect.

Say a PhD student in agricultural science at a university in Ludhiana has written a 6,000-word paper. It reports how much water a new wheat variety needs. She pastes the paper into the assistant, and it marks 140 places. Most are missing or wrong articles, the words "a", "an" and "the", which many speakers of Indian languages find difficult. Some sentences are too long, and the assistant suggests splitting them. It also checks the paper against the journal's rules. It notes that the abstract is 310 words long, while the journal allows 250. It notes that two references cited in the text are missing from the reference list. She accepts most of the changes. She rejects a few that changed her meaning.

The newer tools are built on a large language model, so they can rewrite a whole paragraph, not only correct words. This creates two risks. First, a model can invent facts or references that do not exist, a problem called AI hallucination. So a careful assistant edits only the researcher's own text, and does not add new claims. Second, most journals now ask authors to declare whether they used AI, and how. An AI tool cannot be listed as an author. Using AI to polish the language is usually accepted. Using AI to write the research is not. This line is discussed in AI-generated content.

Researchers also need their work kept secret. An unpublished paper contains results that nobody else has yet. So a researcher should check that the assistant does not store the text, or use it to train its model.

After the polishing, the author submits the paper on the journal's journal publishing platform, and the assistant's work ends. The journal makes its own language corrections later, during production, but only for papers it has accepted.

Tech, Explained is a part of ingrid.fyi. Every page explains one piece of technology in plain words.