Kristiina Kaju
Head of the Social Sciences and Humanities Information Centre, National Library of Estonia
The rapid spread of generative artificial intelligence has contributed to the creation of large amounts of fabricated and low-quality content — both on Amazon’s book market and among scientific publications, particularly in Google Scholar. Artificial intelligence makes it possible to produce vast quantities of text at almost no cost. This poses a threat to authors in the academic world, misleads information users, and may undermine trust in science.tiklite hulgas, eriti Google Scholaris? Tehisintellekt võimaldab luua suuri koguseid tekste peaaegu olematute kuludega. See ohustab teadusmaailma autoreid, eksitab infotarbijaid ning võib kahjustada usaldust teaduse vastu.
On Amazon’s book market, artificial intelligence has made the creation of fraudulent content simple and inexpensive. While such activity previously required considerable resources, content can now be generated quickly and with little effort. The most common forms of deception include summaries and guides to popular books, which are marketed alongside the original works. Often, these publications merely repeat the main points of the source material without offering any substantive analysis. Unauthorized biographies are also published, using the names and biographical details of well-known authors, although the texts themselves are in fact AI-generated and often inaccurate. In addition, artificial intelligence is used to rephrase original texts in order to sell them as new works, even though their content is based on existing material. Such practices reduce the sales revenue of original authors and mislead readers, who may regard these publications as reliable or authentic sources.
Similar problems have emerged in the academic field in Google Scholar, which uses automated systems to index scholarly texts. Unlike scientific databases with stricter quality control, this approach also allows low-quality or misleading content to circulate. Research has identified numerous suspicious articles and fabricated studies containing phrases characteristic of AI-generated writing. This threatens the credibility of scientific communication and the use of evidence-based information by distorting scholarly debate and making it more difficult to distinguish reliable information from misinformation.
It is especially concerning that many of these articles address politically sensitive topics, including health, the environment, and technology — areas that can be used to influence public opinion. According to experts, this problem requires active intervention. One important step would be the clear labelling of AI-generated content, so that users can assess its reliability more consciously. There is also a need for legislative measures to protect authors’ rights and limit the misuse of their names or creative styles. In the case of scientific databases, it is important to develop more precise filters that can distinguish peer-reviewed scholarly work from unverified material. Attention must also be paid to the academic system itself — for example, to the pressure to publish large numbers of articles, which may encourage the misuse of artificial intelligence.
How can you assess the reliability of a scientific article or other source when searching for information and writing research?
Analyse the language, expressions, and structure of the text
Although artificial intelligence can write in grammatically correct English, AI-generated texts have certain characteristic features:
- Excessive statistical predictability. AI is optimised to predict the statistically most likely next words in order to produce convincing content. Researchers, by contrast, focus on testing hypotheses and describing experiments precisely, which means that their language tends to be more varied and less “predictable”.
- Word repetition and semantic connections. Studies show that AI-generated texts often contain fewer unique words, while those words are more closely connected to each other. This means that the text may appear fluent, but its content can be repetitive and circular.
- Superficial summaries. AI-generated texts often simply restate the main points of the original work without adding new analysis or commentary.
- Typical phrases associated with artificial intelligence.
- AI-generated articles often use general buzzwords in titles and body text in order to appear topical or relevant. Pay attention to words and expressions that sound convincing but lack concrete substance.
Check where the text was published
A large proportion of questionable AI-generated texts do not appear in reliable academic journals.
- Non-indexed journals. Many AI-generated papers are found in journals that are not included in official scientific databases, such as Scopus or Web of Science.
- Google Scholar. Use Google Scholar with caution, as its search results may include texts that have not undergone strict quality control.
Examine the references
Check whether the articles and sources cited actually exist.
Research the author or publisher
It is useful to look for more information about the author or publisher. Find and examine the official website, and check whether the author has previously published other works.
Sources:
AI Is Driving a New Surge of Sham ‘Books’ on Amazon. (2024). AuthorsGuild
Haider, J., Söderström, K. R., Ekström, B. & Rödl, M. (2024). GPT-fabricated scientific papers on Google Scholar: Key features, spread, and implications for preempting evidence manipulation. HKS Misinformation Review.
Constantino, T. (2024). AI Tools Fuel Rise Of Fake Research Papers On Google Scholar. Forbes
Rajkumar, R. (2024). Study finds AI-generated research papers on Google Scholar – why it matters. Zdnet.
Moersen, A. (2025) AI detection for peer reviewers: Look out for red flags. Sage.
The article was originally published on RaRa.ee: “Tehisintellektiga loodud valeinfo teadusartiklites – kuidas eristada õiget? – Eesti Rahvusraamatukogu”, author Kristiina Kaju, published on 27. aprill 2026.
