Client accuses me of using AI to write article

[#1]

Hi all, Thanks in advance for your help. So a client hired me to write a 1500-2000 word article for him. And I did so, spending over 3 hours researching the topic and writing something that I would be proud to post on my own website (the article was in fact just over 2000 words).

I sent the written article to the client, giving them a chance to review and ask for any changes or edits before submitting it as completed work to be paid for. I know, I know, I should have just submitted it. Client runs the article through an AI detection algorithm (originality.ai) and says the app is only 24% confident it's written by a human. Ok, what? I literally wrote the article word for word, didn't ask AI to generate it or paraphrase it, I would not do that. So I tell him I wrote the article, not sure what proof I can give him. At this point it doesn't even sound like the client has even read the article himself.

I've asked him for specific deliverables - would he like me to rewrite the article, edit it, or what? I said: "I totally understand. Again, I wrote this article, I'll share some of the sources I used along the way. Can you tell me what you would like me to do specifically? Are there specific sections of the article that stand out to you that you'd like me to change or rewrite?"

So, I'm stuck in limbo. Client is accusing me of using AI and hasn't specified what he would like me to do about it. Please advise! I am top rated and really don't want a negative review. Do I report it to Upwork, give him an entirely new article, or something else? Thank you!

[#2]

Wait for answer. Submit job as-is if no answer received.

Dont worry about feedback, you can remove it with top-rated perk if some.

You have job done so must be paid. Maybe client just want it for free, it is happening.

[#3]

This could be a legitimate client who believes you used AI. The detection tools are hit-and-miss. What the client should say

is if there is any plagiarism, because the detectors go off when AI is used, or if it is not up to my standards, it will be rejected. The AI I have seen, even the ones who have studied how to prompt the AI, still reads poorly. I find it hilarious that there are so many learning, training, etc. hoops to jump through to make the program work, then editing, fact-checking, error detection, etc. How is that saving any time?

However, this could also be a client who is trying to steal your work for no pay. It's difficult to sort them out sometimes.

I agree with Mykola. You are top-rated, so submit the work and if you receive a bad review, you can remove it.

There isn't anything to report. The client can be unhappy. At this point, there are no definitive programming to always detect AI, and the degree. The programs may be better than humans unfamiliar with AI, but it varies.

[#4]

Helena:
I don't doubt that you feel disappointed.

[#5]

Thank you all for responding! The client proceeded with sending over "proof" from the AI checker that showed highlighted sections with high vs. low predictability of AI assistance. As anyone who is in the writing business knows, language = patterns, so it's not exactly wrong when words can be predicted. Anyway, I submitted the work as-is and told the client that unless he had any actionable items, that would be my final draft. He stopped arguing and released the payment, so at least I don't have to wait 2 weeks. I will see if he ends up writing a poor review and contact Upwork support if he does!

[#6]

I'm pleased the client has now paid you, and hopefully that is the end of the matter. On a side note, (and i say this as a bystander from an unrelated field) but I assume with the way AI is currently so widely available, similar situations could quite conceivably start becoming more commonplace (not that I think they should). Im not sure how practical it is for you, but would using something along the line of Upworks time tracker tool be worthwhile, just so if the situation ever does rise again, you have instant proof in the form of time stamped photos showing various stages of completion?

It may be totally inpractical for you, or (hopefully) unnesessary, but thought id put it out there as in idea in case it does start happening on a more regular basis.

[#7]

You've been far too kind a patient. Anyone who has invested 6 seconds in research knows that AI detectors don't work. And, if a client can't make a decision about the quality of content without running it through a machine, he's not competent to be hiring writers. 

In your shoes, I would submit the work for payment and let him know that I was more than happy to both dispute and go to arbitration to prove that I'd written the content. But, I would actually be willing to put down the $300+ to do that. And then I'd leave a scathing review to warn off other freelancers who are less well established and might actually be harmed by the client's lazy ignorance.

If you're newer and care about how he rates you and/or don't want to spend the money to crush him like the bug he is, you may have to be more diplomatic.

[#8]

You can't report this to Upwork.

This is not an Upwork matter: Upwork does not make a determination about whether a freelancer's work was generated by AI or not.

This is a matter between a freelancer and a client.

Note that if you did the task asked, the client is required to pay for it whether or not the client claims to like or dislike it, and whether or not the client accuses you of using AI.

[#9]

My father recently completed a master's degree in cybersecurity. For one essay, he wrote the entire piece relying strictly on his personal work experience (Raytheon IT, 25+ years, with US security clearance). The professor ran it through an academic plagiarism software that claimed my father copied passages from a few textbooks/papers on the subject. He then threatened to report it to the institution's academic honesty board, which could have terminated my father's master's program pursuit (which was subsidized by his employer). It was not until my father relayed his own credentials that the professor relented and finally accepted that there tend to be few ways of accurately talking about very advanced niche topics.

Long story short: plagiarism and AI-detection tools only review correlation and CANNOT make causation claims. Anyone who relies on them to do so is not qualified to review that work.

[#10]

If a client hires a freelancer to do some writing for him, and then the client thinks the freelancer used AI, then I don't think the client should accuse the freelancer of anything. The client should simply end the contract.

It would be a waste of time for the client to get into an argument with the freelancer about the use of AI.

The client ALREADY KNOWS that he doesn't like the freelancer's writing. So it doesn't really matter if the freelancer used AI or not. The freelancer doesn't have anything else to offer. That's what you get with that freelancer. So the client needs to end the contract and looks for other people to provide what he needs.

Upwork is NOT GOING TO make a determination about whether any work was created by ChatGPT/AI, or not. And there is no tool that can make that determination. A cilent CAN NOT claim that a freelancer's work was generated by AI and use that as a basis for not paying the freelancer.

If a client hires the freelancer to do a task, and the freelancer does the task, the client needs to pay for the task.

So the whole question of whether a freelancer's work was AI-generated or not is something for the client to talk about with his buddies at lunch time. It is NOT something to discuss with the freelancer or with Upwork.

Jonathan wrote: "Long story short: plagiarism and AI-detection tools only review correlation and CANNOT make causation claims. Anyone who relies on them to do so is not qualified to review that work."

Clearly he does NOT believe the Internet wisdom: "Any time you click a button on an Internet tool, whatever words appear next must therefore be true."

[#11]

I have been a top rated freelancer with a 100% JSS until today. The client claims that my entire content was AI generated when in reality I wrote it all by myself. She even requested a refund. My JSS went down to 92%. I asked the client to grant me access to her AI detector, since I have seen that every AI detector gives a different result. The AI detector I have showed 91%human written content while the client is saying that my content is barely human written. I feel miserable.

[#12]

Yusra: I am really sorry that this happened.

But try not to feel miserable. This is business. Try not to take it personally.

This is simply a sign of the times that we live in. You are dealing with a client who doesn't understand these things very much. Most people don't.

It is important for YOU to understand that the client has NO BASIS for asking for a refund.

IF A CLIENT BELIEVES THAT CONTENT was written by AI, and even if a client has proof that content was AI-generated, that is not a basis for a refund.

So do not agree to a refund.

[#13]

Thank you Preston! Your reply was much needed. I am just worried what if I don't agree to her refund request and she starts a dispute. Would that further affect my JSS? 

[#14]

Has the client already ended the contract? Then the client's feedback has ALREADY been left, and nothing she does will further affect JSS.

Also, you are here to work for money, not stars.

If you did the work, you should be paid for the work.

If she threatens bad feedback because you don't give a refund, that is feedback manipulation. That is a violation of Upwork ToS and you can report her for that.

[#15]

Just ask them to post popular content like the Declaration of Independence, Bible passages into the AI detector to confirm they are crap.

You can also look for articles that support this and send to them.

There isn't a way to detect AI-written content.

[#16]

Ikechukwu: You're right!

This is a preposterous game that some clients are playing.

I think most of those clients really don't understand how this works, and they don't understand that these "detection" tools don't really work So I think the clients are mostly sincere.

But they're still very wrong.

To make things simple:

If a client doesn't like the freelancer's writing, the client still needs to pay for it.

If the client thinks the freelancer's writing was created using AI, the client still needs to pay for it.

If the client can "show" that the freelancer's writing was using AI, the client still needs to pay for it.

The client DOES NOT NEED to use the writing if she doesn't want to.

But she still needs to pay for it.