AI Essay Writer That Never Invents Sources
HidenGPT's Essay format drafts argumentative, expository, compare-and-contrast, analysis and reflective essays, research reports and homework answers at your level and word count. It cites only the sources you give it, in APA 7, MLA 9, Chicago or Harvard, and a citation it can't match to them becomes a [source] placeholder instead of a made-up reference.
Try it: Essay
Enter to send · Shift+Enter for a new line
What you get
From one position, two notes and the objections the student expected, HidenGPT wrote this 802-word argumentative essay. No sources were given, so it put [source] wherever a reader would expect a citation instead of inventing one.
Typed into HidenGPT
Argumentative essay, 800 words, university level: should cities make public buses free? My position: yes, but start with free fares for students and people over 65, not for everyone. Notes from my transport policy seminar: Tallinn made public transport free for its registered residents in 2013, and where I live a monthly pass costs $55, which is a lot on a student budget. The other side: fares pay part of the drivers' wages, and free buses could get crowded at rush hour.
Free Buses, Starting With the Riders Who Need Them Most
Public buses are one of the few city services that most residents pay for each time they use them. Roads and street lighting are paid for through taxes, while a bus ride still carries a price at the door. That price is easy to overlook for people who drive, yet for a student or a retiree on a fixed income it can decide whether a trip happens at all. At least one city has already acted on this: Tallinn made public transport free for its registered residents in 2013 [source]. Making buses free for everyone at once, however, would be expensive, and no city could say in advance what it would achieve. Cities should make public buses free, but they should begin with students and people over 65, the riders for whom the fare is the largest barrier, and widen the scheme from there.
For students, the cost of a fare adds up quickly. Where a monthly pass costs $55, a student pays for the bus as regularly as for a phone plan, and the money comes out of a budget that is already tight. For students who also work part-time, the trip to work or class is not optional. When the fare is too high, students tend to walk long distances or skip the trip altogether [source]. Each skipped trip narrows what a city offers them, whether that is the library across town or a better-paid job in another district. Over a degree that lasts several years, those small decisions add up to a narrower education and a narrower start in work. A free pass removes that calculation for the group most likely to depend on the bus in the first place.
Older residents face a different version of the same problem. For those who no longer drive, the bus is the link to doctors and family. A fare that seems small to a commuter can be a real cost on a pension, where daily rides quickly become a monthly bill. Free travel for this group also serves a wider public goal, since older people who can get around easily tend to stay independent for longer [source].
Starting with these two groups costs less than a universal scheme, and it is easier to evaluate. A city can count how many trips students and older residents make before and after the change and set that against the fare revenue it gave up. It can see which routes carry the new riders, at what times of day, and whether the extra trips replace car journeys or simply add to them. With that evidence in hand, the decision about the next group of riders becomes a practical one, based on what the change has shown. Tallinn's decision shows that a city can make public transport free for its registered residents, but a city with a different budget and a different road network cannot assume that the same results would follow. A targeted scheme lets each city learn what free buses do in its own conditions before it commits to the full cost.
The strongest objection concerns money. Fares pay part of the drivers' wages, and a transit agency that gives up that income must either find it elsewhere or cut service. Free buses could also become crowded at rush hour, which would make the service worse for the people who already rely on it. Both concerns are serious. They are also the main reason not to make every ride free at once, since a universal scheme would lose all of its fare income at once and risk crowding on every route from the first day.
A targeted scheme answers both objections better than a universal one. Students and older residents make up only part of the ridership, so the revenue a city loses is limited and can be covered from its general budget without cutting into the drivers' wages. Crowding is less likely as well, because many older riders travel outside the morning and evening peaks [source], and the city can watch the busiest routes closely in the first year. If buses do fill up, the right response is to run more of them on those routes. Full buses are a sign of demand, not an argument against making the service affordable.
For a commuter the fare is a minor expense; for a student or a pensioner it can decide whether the trip happens. Free buses for students and people over 65 would lift that cost where it weighs most, at a cost a city can measure and control. Once the results are in, the case for going further, or the reasons to stop, will rest on the city's own numbers rather than on hope. A city can begin with the riders who need free buses most and let the evidence decide the rest.
How to use it
- 01
Paste the assignment brief or the questions. If you want citations, paste your sources and reading notes under Sources & notes, not into the brief: only that list can be cited.
- 02
Essay is already selected here. Choose the assignment type, your level, the word count, your thesis if you have one and the citation style; More details holds the tone.
- 03
Send it, check the outline, the word count and any [source] placeholders, then ask for changes such as “stronger counterargument” or “shorter conclusion”.
Good for
- An argumentative essay that takes a clear position and answers the other side
- A compare-and-contrast essay with clear criteria
- A literary or case analysis built around a thesis
- A research report with sections and a reference list from your sources
- Worked answers to homework questions, with the steps shown
- A first draft to study from and rewrite in your own words
Why AI essay writers invent citations
A language model writes what text usually looks like. A reference usually looks like an author, a year, a title and a journal, so when a model is asked for one without the source in front of it, it can produce something with the right shape that doesn't exist.
Studies have put numbers on it. In short literature reviews on 42 topics, generated in April 2023, 55% of the GPT-3.5 citations and 18% of the GPT-4 citations were fabricated (Walters and Wilder, Scientific Reports, 2023). When ChatGPT-3.5 wrote 30 short medical papers, 47% of its 115 references were fabricated and only 7% were both real and accurate (Bhattacharyya and colleagues, Cureus, 2023). And when researchers in Montreal asked ChatGPT 20 medical questions, 69% of the references it gave were fabricated, nearly all of them under the names of authors who had really published (Gravel and colleagues, 2023). That is what makes them hard to spot.
A newer model can make fewer of these mistakes, as GPT-4 did in the first study, but the safe setup doesn't change: start from sources you've read and cite only those. That's how the Essay format works.
How the citation check works
What you paste under Sources & notes is the only list the essay may cite. After the draft, the red pen and every fix, each author-year citation in parentheses, such as (Wardle, 2019) or (Wardle & Derakhshan, 2017, p. 4), is checked: if the first author's surname isn't in your sources, the citation is replaced with [source]. A narrative citation that doesn't match, such as Smith et al. (2021), is sent back to be rewritten. With no sources at all, any reference list the model writes is removed, and the red pen leaves citations and placeholders exactly as they are.
So if Wardle (2019) is your only source, (Wardle, 2019) stays in the draft and (Smith, 2020) comes back as [source]. Search the draft for [source], then either add a source you've read or cut the claim.
The check matches author names, not years or page numbers. MLA author-page citations such as (Wardle 12), and a single name before a year such as Wardle (2019), aren't checked automatically. For APA or Harvard, run the finished essay through the free citation checker, which matches each citation to your reference list by author and year and lists years that don't match. More on citing only your sources.
What to paste as sources
For each source, give the author, year, title, publisher or journal (with volume, issue and pages for an article) and the URL or DOI. Under it, add the notes or quotes you plan to use, with page numbers. The notes are what the essay argues from; with a title alone, it has nothing to say about the source.
Paste only works you've read, and put them under Sources & notes rather than in the assignment box: a citation typed into the brief isn't on the list and turns into [source]. There's no file upload, so copy the text in; the field takes up to 20,000 characters.
There's no plagiarism checker here. Still, an essay that grows out of your own notes and quoted pages is much easier to keep honest than one that starts from a blank prompt.
Hitting the word count
Set a target from 150 to 4,000 words and the essay is planned to that length: an outline first from 600 words, and section headings from 1,500, while shorter essays run as paragraphs with no headings (a research report always has sections). The body is counted without the reference list. A draft more than 10% short or 20% long gets up to two automatic fix passes, and between 10% and 20% over, you see a warning with the actual count.
The details are in setting an exact word count, and to see what a target looks like on the page, convert words to pages.
The rules it follows
| Assignment types | Argumentative, Expository, Compare & contrast, Analysis, Cause & effect, Reflective, Research report, Answer questions |
|---|---|
| Levels | Middle school, High school, University, Graduate |
| Word count | 150-4,000 words (800 by default); a draft more than 10% short or 20% long is corrected automatically |
| Tone | By level (default): Plain & clear for middle and high school, Academic for university and graduate. Or pick Academic, Plain & clear or Personal |
| Planning | An outline comes first from 600 words; from 1,800 words each section is written on its own |
| Citations | None, APA 7, MLA 9, Chicago (author-date) or Harvard, only from the sources you give |
| Sources | Pasted under Sources & notes, up to 20,000 characters; no file upload |
| Invented citations | Replaced with [source]; a reference list without sources is removed |
| Red pen | Light by default; citations and the essay's register stay as they are |
| Download formats | Markdown, HTML, plain text |
Questions
Is there an AI essay writer with real citations?
Yes, as long as the citations come from sources you give it. HidenGPT's Essay format cites only what you paste under Sources & notes, in APA 7, MLA 9, Chicago (author-date) or Harvard, and an author-year citation it can't match to your list becomes [source] instead of a made-up reference.
Why does ChatGPT make up references?
Because it writes what a reference usually looks like. Without the source in front of it, a plausible author, year and title can come out for a paper that doesn't exist; in one 2023 study of short literature reviews, 55% of GPT-3.5's citations and 18% of GPT-4's were fabricated.
Will it find sources for me, or can I upload a PDF?
No to both. It doesn't search for sources and has no file upload: paste the details and your notes from works you've read under Sources & notes, up to 20,000 characters. Without sources, it writes [source] where a claim needs one.
Which citation styles does it support?
APA 7, MLA 9, Chicago (author-date) and Harvard, with in-text citations such as (Wardle, 2019), (Wardle 12), (Wardle 2019, 12) and (Wardle 2019). Chicago notes and bibliography isn't included. Pick None if your assignment doesn't need citations; it still won't invent facts.
Do I still need to check the citations?
Yes. The check matches author names, not years, titles or page numbers, and MLA author-page citations aren't checked automatically. Compare each citation with your source, or run an APA or Harvard essay through the free citation checker, which also lists years that don't match.
Can I hand in what it writes?
Only where your school allows AI help, and as your own work only where its rules say so. Many classes ban or require you to disclose AI use; our Acceptable Use Policy asks you to follow those rules. It works best as a draft you study, check and rewrite.