Page 14 of 32.
A generative AI system is not one fixed thing: providers update and retire models on their own schedule, and a citation without a version string cannot say which one produced your text. Here is where to find the exact label your interface showed, what to do when it will not give you one, and how APA, MLA, Chicago and IEEE each treat the version element differently.
A paraphrasing tool and an AI humanizer get reached for interchangeably, and they do different jobs with different failure modes. Here is the purpose, the input, what breaks in each, and how to tell which one actually solves your problem.
An annotated bibliography and a literature review can be built from the same five sources and still do completely different jobs: one evaluates each source on its own, the other builds an argument out of all of them at once. Here is what changes, source by source, when the same reading list moves from one document into the other.
Introduction, Methods, Results, Discussion: four sections, four different questions. Most structural mistakes are really one section answering another section's question, especially where interpretation, a stray result, or a literature review lands in the wrong place.
A classifier trained to react to statistical pattern does not need to have seen a specific chatbot to flag its output. Here is why that matters for DeepSeek, what Turnitin itself publishes about Claude, Gemini and other models by name, and what it does not publish for any of them.
An academic integrity email naming AI use is frightening precisely because there is no single fact that disproves a percentage. Here is what to do, in order, in the first 48 hours: what to say, what never to admit, what evidence to request, and how to find out whether a score is even allowed to stand alone.