You have asked AI to explore a psychology topic. The answer reads clearly, the argument seems convincing, and a few references appear at the end: authors, a journal, a year, even a DOI. Before adding any of those citations to your own materials, it helps to open the source. Can you find the work described? Once you have found it, does it actually support the statement attached to it?
For a psychologist or psychotherapist, these questions may arise while preparing an update for colleagues, writing an educational text, or studying a professional topic. Checking becomes easier when you separate three tasks: establishing that the work exists, checking its bibliographic details, and assessing the relevance of its content. Here is a practical way to work through them and decide what to do when something does not match.
Plausible references can still contain errors
A neatly presented reference makes it easy to feel reassured. The title is relevant, the journal sounds right, and an author’s name looks familiar. Those details give you useful starting points for a search, but they still need independent confirmation.
A 2023 study by Walters and Wilder examined references produced by GPT-3.5 and GPT-4 for short literature reviews on different topics. The researchers found both fabricated citations and incorrect details for works that did exist. These findings concern those models under those study conditions. They do not measure the reliability of PsyLab or of the tools available today.
The distinction is useful in everyday checking. You might encounter a real paper with the wrong year, a link that leads to a different article, or a reference you cannot identify. Each situation calls for a different check. Asking the same AI whether it is sure about a citation may continue the conversation, but resolving the uncertainty requires looking for the document outside the answer you received.
Start with the title and compare details
Copy the article title and search for it in its original language. For biomedical literature and some psychological research, PubMed can be a starting point. Its official guidance on finding a specific citation explains that you can enter a title or combine an author, journal, and publication year. The system may suggest a match even if some of the information entered is incorrect.
Open the result and compare it with the reference supplied by the AI. Check the authors, the full title, the journal, and the publication details. If a title looks similar, take a moment to read the differences: it could describe another study on the same subject.
The year sometimes needs closer attention. Online publication and publication in a journal issue can have different dates. A discrepancy therefore needs to be understood before you classify it as an error. Note the uncertainty and compare the bibliographic record with the publisher’s page.
If the first search returns nothing, try the author’s surname with a few distinctive words from the title. Then broaden the search to the journal’s website or your library catalogue. PubMed does not include all scholarly publications. Failure to find a reference in one database leaves it unresolved; that absence alone does not establish that the work was invented.
Use the DOI to identify the work
A DOI is an identifier that you can use to reach a document. If the AI response includes one, open it and check where it leads. The destination should correspond to the citation: its title, authors, and journal should identify the same work.
When a DOI is missing, Crossref provides tools to look for it using bibliographic information. The documentation for Simple Text Query also describes the possibility of receiving several potential matches. You need to compare the details before choosing the appropriate one.
Treat this step as an identity check. Once you have found the correct DOI, you still need to assess the article’s content and its usefulness for your question. A working identifier takes you to a source; reading the source helps you decide how to use it.
Read what supports the specific claim
Return to the sentence in the AI answer that carried the citation. What exactly needs checking? It might be a definition, a study finding, a relationship between variables, or a suggestion made by the authors. Identifying that statement gives your reading a clear purpose.
Consider this entirely fictional example. An answer claims that a training exercise improves psychotherapists’ interviewing skills. The paper you find instead describes how much a group of students liked the exercise. The source may be interesting, but its participants and measured outcome differ from those in the original statement. Before citing it, you need to revise the wording or find more relevant evidence.
The abstract, or opening summary, can help you get your bearings. For a central claim, move on to the relevant full-text sections: methods, results, discussion, and any supplementary material needed to check it. Look at who took part, what was measured, and which limitations the authors describe. Separate the study’s findings from the interpretations and possibilities discussed around them.
If you can access only the abstract, record that limitation in your notes. You can then decide whether it supports a narrowly worded description or whether you need the full text before using the reference. Do not let an incomplete reading become an apparently complete verification when you return to the draft later.
Decide what to keep and revise
An incorrect bibliographic detail can be corrected once you have identified the work. A statement that goes beyond the available evidence requires a change to the text. A reference you still cannot locate can remain among your research leads, separate from your checked bibliography.
Avoid replacing a doubtful citation with the first article you find on the same topic. The replacement needs to support the particular sentence you intend to write. Sometimes the most useful decision is to narrow the claim. At other times, remove the passage and return to it after further reading.
A note beside your draft can preserve the work you have done. Record the corrected reference, its link, the section you read, and the statement it can support. Add the date of your check and any unresolved questions. When you revisit the text, you will have the reasoning available rather than having to reconstruct the search from memory.
Keep verified content in your draft
You can also use an AI chat to work on material you have already read. For example, ask it to suggest a summary of a short passage, then compare that summary with the original. A focused request might be: “Rephrase this excerpt for psychologist colleagues. Preserve the authors’ stated limitations and do not add references.” Your subsequent review checks whether the meaning remains faithful.
PsyLab is a digital workspace for mental health professionals that includes Chat and Documents. If you use it to prepare professional materials, apply the same checking process to the information and references you decide to retain.
Start with a single citation: find the work, compare its details, and reread the claim in light of the source. To explore AI support for preparing your professional texts, you can try PsyLab for free.