How To Critique A ChatGPT Essay For Commentary
Quick answer: to critique a ChatGPT essay well, check its claims against primary sources, test every reference for existence, accuracy and relevance, list what it leaves out, then use those gaps as the structure of your own essay.
You have asked ChatGPT for an essay on your chosen lecture topic, and now the brief wants a 1,000-word commentary on it, followed by your own 2,000–3,000-word essay that goes further. The commentary is where many students lose the plot: they summarise the AI text instead of judging it.
This guide is based on the Values, Ethics and IPR written report brief for a Masters information studies course in Singapore. If you want to see the four parts laid out in practice, our ChatGPT commentary essay sample is available for reference only, not for submission.
Need a second pair of eyes on structure? Our report writing help team can talk through how the four parts fit together.
Where the marks are
The brief does not state a mark split, so do not assume one. Check your course handbook or ask your tutor. What the brief does make clear is that there are four deliverables, and each tests a different skill.
| Deliverable | What it tests | What markers will look for |
|---|---|---|
| 1. Chosen question, rationale and overview | Topic selection and scoping | A focused, answerable question tied to the lecture; a reason for choosing it; a clear preview of the answer |
| 2. Query written for ChatGPT | Reading the brief and prompt design | A query that names the topic, scope and referencing style, so the output can be judged |
| 3. Commentary (about 1,000 words) | Critical evaluation | Accuracy, nature of references, gaps and omissions, other observations, all backed by evidence |
| 4. Own essay (2,000–3,000 words) | Research, argument, academic writing | Own words, wider scholarship, a clear structure (comparative or interrogative), APA 7 referencing |
One thing is worth saying plainly. The commentary is short, but it is the part that shows you can think independently of the tool. Spend your time there.
Step 1: Choose a question the tool can get wrong
Pick a question that is specific enough to be checked. “Ethics of AI” is too wide, and any answer will sound fine. “Should works generated by AI be protected by copyright?” is better: it has law, cases and scholarship you can test against. The same applies to other Values & IPR topics, such as data ownership, open access or plagiarism.
Write two or three sentences on why you chose it, then give a short overview of the answer. Markers want to see a position at this stage, even a provisional one.
Step 2: Write a query that gives you something to test
Treat the query as part of the assessment, because it is. State the topic, the word length, the audience, the jurisdictions to cover and the referencing style. Asking for in-text citations and a reference list is useful because you then have claims and sources to check.
Keep a note of the date, the tool and any follow-up prompts. Save the whole output. If the brief lets you include it in an appendix, do so.
Step 3: Write the commentary (about 1,000 words)
The brief points to four areas. A sensible budget looks like this.
| Section | Words | Question to answer |
|---|---|---|
| Short overview of the AI essay | 80–100 | What does it argue, and how is it organised? |
| Accuracy | 250–300 | Which claims are correct, overstated, outdated or wrong? |
| Nature of the references | 250–300 | Do they exist, are the details right, do they support the claim? |
| Gaps and omissions | 200–250 | What would a Masters-level answer include that this one does not? |
| Other observations | 100–150 | Tone, bias, structure, hedging, what it means for how you will use the tool |
Accuracy
Pick five to eight claims and check each against a primary source: a statute, a case, a regulator’s report, a peer-reviewed article. Do not check them against another website that may itself be AI-written. Report what you find with specifics. “The essay says courts have held that AI cannot be an author, but does not name a case” is a finding. “Some parts are inaccurate” is not.
Watch for three habits in AI text. It generalises one country’s law into universal law. It states uncertain or changing law with total confidence. And it merges two separate questions into one. In copyright, for example, who owns an AI output is a different question from whether copying works to train the tool is lawful.
Nature of the references
This is the section students tend to under-do. Research testing ChatGPT’s bibliographic output found that it can fabricate citations, with believable authors, titles and journals, and that real citations often contain errors (Walters & Wilder, 2023). So test every reference in three steps.
- Existence: search the exact title in your library database, Google Scholar or Crossref. Check the DOI resolves to the same article.
- Accuracy: do the authors, year, journal, volume and pages match the real record?
- Relevance: does the source say what the sentence claims it says? A real article cited for a point it never makes still counts against the essay.
Then describe the pattern. How many references were real, partly right, or not found? Are they peer-reviewed, or mostly general commentary? Is there any primary law or data? A small table in the commentary, listing each reference and your verdict, is an efficient way to show this, though check whether the word count includes it.
Gaps and omissions
Ask what an informed reader would expect and does not find. Typical gaps: no local context (for Singapore, that might mean the Copyright Act 2021 rather than only US or UK law), no recent developments, no counter-argument, no ethical analysis that actually applies values to people or groups, and no thesis. Each gap you name becomes a section in your own essay.
If you want help structuring the critique so each point is evidence-led, you can discuss it with our essay writing support team.
Other observations
Use this space for what the other three do not cover. Is the tone more confident than the evidence allows? Does the essay hedge everything with “it is important to consider”? Does it show bias towards a particular legal system or industry view? Finish with one line on what this means for responsible use: the tool can start a draft, but it cannot be your source.
Step 4: Turn the gaps into your own essay
The brief says to use the AI output as a starting point, but to rewrite in your own words and extend beyond it. Copying AI text counts as plagiarism, so do not paste and paraphrase sentence by sentence. Instead, work from your commentary.
Your gap list becomes your plan. If the AI essay missed local law, add a section on it. If it merged inputs and outputs, separate them. If it had no thesis, write one. This also means your essay and commentary support each other: the marker can see the judgement in Part 3 turn into argument in Part 4.
Choosing a structure
The brief suggests two. Pick the one that suits your question.
| Structure | Use it when | Example |
|---|---|---|
| Interrogative | You have one clear question to answer | “Should AI-generated works be protected by copyright?” Thesis, then evidence, then counter-argument, then answer |
| Comparative | You are weighing two phenomena or approaches | Comparing how two jurisdictions treat AI authorship, point by point on the same criteria |
Whichever you use, stay on target. Every paragraph should help answer the question you set in Part 1.
Mark-maximising checklist
- The question is narrow enough to answer in 2,000–3,000 words.
- The query to ChatGPT names scope, jurisdiction and referencing style.
- The commentary covers all four areas the brief lists, within about 1,000 words.
- At least five claims from the AI essay are checked against primary sources.
- Every AI reference is tested for existence, accuracy and relevance, and the results are reported.
- Your own essay uses sources the AI did not, including at least some peer-reviewed articles and books.
- No sentence is copied from the AI text, and all sources are cited in APA 7.
- Every in-text citation appears in the reference list, and the reverse.
Common mistakes
- Summarising instead of evaluating. A commentary that retells the AI essay earns little.
- Vague criticism. “The references are weak” needs numbers and examples.
- Trusting a reference because it looks right. Fabricated citations are designed, in effect, to look right.
- Checking against secondary blogs. Use statutes, cases, official reports and journal articles.
- Rewriting the AI essay rather than extending it. If your essay has the same structure and the same points, it is not new work.
- Ignoring local context. For a Singapore course, show you know where Singapore’s rules differ.
What students actually struggle with
Three problems come up repeatedly. First, students cannot tell how critical a 1,000-word commentary should be, and either praise the output or dismiss it wholesale. A balanced verdict, backed by evidence, is stronger than either. Second, they run out of words and cut the references section, which is often the most valuable. Third, they worry about using ChatGPT at all. The brief requires it here, but your institution’s wider rules still apply, so read them. Our related guide on using ChatGPT for assignments in Singapore covers that question separately.
If your topic touches media or publishing ethics, the ideas in our media law assignment help page may also be useful background.
Frequently asked questions
How long should the ChatGPT commentary be?
The brief asks for about 1,000 words. Plan roughly 100 for an overview, then 250–300 each for accuracy and references, around 200 for gaps, and the rest for other observations. Stay close to the limit, because concise evaluation is part of the skill being assessed.
Can I quote the ChatGPT essay in my commentary?
Short quotations are normal in a commentary because you are analysing the text, so quote the exact words and say they come from the AI output. Do not reuse its wording in your own essay. Check with your tutor how to cite the tool and whether to include the full output as an appendix.
What if every ChatGPT reference turns out to be real?
Then report that, but keep testing. Check whether the details are accurate and whether each source supports the claim. Also comment on the type of sources: a list of real but general or dated works still has weaknesses, especially if peer-reviewed articles, primary law or recent material are missing.
Which structure is better, comparative or interrogative?
Neither is better in itself. Use interrogative when you have a single question and a thesis, and comparative when your topic naturally has two things to weigh against shared criteria. Pick the one that lets you stay on target for the whole essay.
How much of the essay can come from the ChatGPT output?
The brief treats the output as a starting point only. Use it to see what an ordinary answer looks like and where it falls short, then write in your own words from your own research. Copying AI text is treated as plagiarism, so none of its sentences should appear in your essay.
Do I need to use books and journal articles?
Yes. The brief asks for peer-reviewed journal articles and books in the research stage. Search your library databases, check each source exists, and cite it in APA 7. Use news pieces and websites sparingly and only for context.
Last step before you start: open the AI essay, pick your first five claims, and find the primary source for each. That single hour of checking will shape the rest of your report. If you want a structured plan, look at our research paper help page or ask for guidance through the question page above.
About the expert
According to his profile, Mark Wong holds a Master’s degree in Journalism and Mass Communication from the University of Wisconsin-Madison, and his areas include media law and ethics, public relations, journalism and advertising.
References
Walters, W. H., & Wilder, E. I. (2023). Fabrication and errors in the bibliographic citations generated by ChatGPT. Scientific Reports, 13, Article 14045. https://doi.org/10.1038/s41598-023-41032-5
Deadline Close? Let an Expert Handle It.
Submit your assignment details and get a free quote within 10 minutes — 100% human-written, plagiarism-free, and delivered before your deadline.
Submit My Assignment Now →