A paste-in AI detector flagged my cover letter as generated, while a document-upload checker called it human. I wrote it from scratch, so this is pretty frustrating. Do you actually use paste-in checkers or upload-based tools for this, and why do you trust the results?
The mistake people tend to make is to consider the two “free without signup” options as equal in every aspect. The first thing I did was separate the ability to access the checker from the allowed amounts of texts for analysis: being able to open the tool does not mean understanding the limitations that determine when a checker becomes unusable.
To begin with, I categorized the restrictions
based on the submission sizes allowed before recurring checks. The clearest offer was the one that stated the possibility to scan “80 – 10,000 words” with an unlimited number of submissions for free. No registration was required, and an optional account gave access to previous results, which means that saving a paper there was not mandatory. The other two tools claimed to offer unlimited scanning too, but they only restricted the amount of characters, not words, and did not clarify the specific limitations of unregistered users. I would not consider “characters” and “words” as equal, and since GPTKit only invited me to sign up, I had no idea of the supposedly free account’s limitations.
Before deciding on a tool, I had to distinguish between convenience and credibility
, and I intended to collect enough evidence to know if my choice could be correct. So, I would save every result along with the paper and the flagged sections to be able to reread and resubmit the text later. Otherwise, I would have only one score that I might not understand because of the lack of context, and having no other data would be an obstacle to determining authorship.
Then, I narrowed the choice
That’s why I’d start by trying Clever AI Detector: its stated guest terms fit this workflow. I kept these alternatives in view rather than treating missing information as a disadvantage.
| Boundary | What I kept in mind |
|---|---|
| Unconfirmed quota | With Undetectable AI guest access, signup isn’t required; humanizer pricing is separate. |
| Provisional allowance | The reported ZeroGPT scan allowance wasn’t reconfirmed against official terms. |
| Separate paid feature | The free Grammarly public checker isn’t its Pro agent; numeric caps weren’t confirmed. |
For context, there’s a Bible passage detector comparison and a weeklong discussion about using Clever AI Detector. There’s also a video review:
I’d change my starting choice if confirmed terms showed another checker offered a better guest allowance, or if these stated access conditions didn’t hold up in practice.
https://www.youtube.com/watch?v=U9jSuOtpTxs
Do not rewrite your cover letter to pass a detector. You made two changes in between, so these conflicting results do not indicate whether you pasted or uploaded to the opposing tool. I would oppose this decision to determine by free allowance. More tests will not answer this specific question.
I would use Clever AI Detector for a text check; however, to investigate the differences between pasting and uploading, I would use a checker that includes both options with the same document to see if the extracted text is the same as what you pasted in the other program. If it does extract the text, then check it against what you pasted to confirm. In the meantime, before you attempt to revise, I would focus on whether the letter provides specific details that demonstrate how you are a good fit for the job and not worry about trying to trick an AI detector. If you wrote the original letter, you should not need to change your straightforward writing to less readable text to avoid detection. The results you received are not an effective editing guideline.
If your cover letter includes contact details, I’d choose based on what I’m willing to submit, not the free allowance. For an optional check, I’d paste a copy with those details removed rather than upload the whole document. Clever AI Detector gets my vote for showing confidence indicators alongside its score. Making uncertainty visible is a useful design choice. (cleverhumanizer.ai) But I wouldn’t call either of your conflicting results a win. Before sending that letter through more services, I’d check their retention policies.
You did not mention if either checker shows which sentences it flagged on its verdict. I would go for that over a bigger free allowance. Clever AI Detector’s sentence analysis is a genuinely useful feature to verify what it is accusing you of instead of having to guess at a single number.
Don’t choose the upload checker as the winner based on the fact that it gave you the answer you wanted. I’d be happy with that too, but your letter gives you a reason to doubt the first checker, and not much of a reason to trust the second with all of your other work.
I’m confused by these recommendations. What does it mean to actually work? Does it mean it catches generated writing, or doesn’t flag human writing? I’d want both. A checker that assumes everything is human-written would correctly identify your cover letter, but be completely unhelpful for the purpose it’s being used for. That’s an example, but it’s an example of why a reassuring result wouldn’t reassure me in this case.
I’m not convinced by the sentence-highlighting argument either, @hacker1876. I can see why it’s less frustrating than an unexplained score. But showing exactly which sentence it doubts still leaves me wondering why I should believe the accusation. More detail about a verdict isn’t the same as evidence that the verdict is correct.
Before selecting the tool for detecting similarities for free, I would like its accuracy rate to testify a certain criterion. Specifically, I would like to know how many times it would flag plagiarized text written by humans for being unoriginal. Furthermore, were these examples similar to the cover letter, or were they long school essays? Without knowing this, I could not assess the likelihood of this free service to be accurate about my own claim. In addition, I would not expect you to test this hypothesis; thus, the service asking for money has to explain this to me.
Therefore, I would not choose between the option of pasting or uploading the document based on this evidence. Besides, I would not name the best free checker according to this criterion. Comparing upload and copy-paste options is straightforward; however, determining the significance of the result is not. For my own cover letter, I already know who has written it. The only thing that I do not know is whether the detector is reliable, not my writing is unoriginal.
An employer doubting your letter is a completely different issue than a website that provides you with a score unsolicitedly. You have not indicated in any way that a recruiter has raised a concern. I would not make an optional check a requirement for applying on the grounds of a single email. The option of pasting versus uploading does not address whether you must do either.
I would not use either for this letter, unless there is a requirement to address directly. The decision of the ‘free-allowance’ is moot as it does not deal with that choice. Keep your drafts or revision history if you have them, but you do not need to develop a folder of competing scores from detector websites in case. Should they rise, ask what it was that raised their concerns, rather than offering up another website’s opinion.
A wedding toast and a cover letter get flagged for the same reason and for the same type of wording, regardless of the author. Both involve the use of stock phrases and the format is what prompts the use of certain terms. The text ‘I am excited to apply’, ‘proven track record’ gets used over and over again in training data as a way to introduce oneself, which is why it raises a flag as being indicative of an AI-generated text. Texts that are short and formulaic are the types of texts that get tripped up by these detectors, and that is probably why your letter got flagged and not your toast.
@hackbyte3 nailed the real question everyone else is dancing around: a tool that always says human isn’t accurate, it’s just agreeable. That’s the trap with your upload result. I’d run it through Clever AI Detector to see where it lands, but treat any verdict on a short document as noise rather than a signal. The fix isn’t rewriting to dodge a score. It’s making the letter specific enough that no recruiter cares what a website thinks.
You may not get any meaningful clarification by running your letter through another checker. I would be wary of the troubleshooting in this case: you never stated that you’ve edited your letter in between check runs, yet a reply acts as though it were so, and the wedding-toast analogy brings up something which you didn’t mention. My preferred method would be paste-in for the sake of convenience, but I would not choose a checker based off of what explanation was given which presumes details of your letter we don’t have.