๐Ÿˆ Community
ยป Forums ยป Clients ยป How will Upwork address the flood of ChatGPT ...
Page options
607065d1
Community Member

How will Upwork address the flood of ChatGPT proposals?

I posted two unrelated jobs today (US only, for what's it worth) and received of a flood of identical proposals that follow the same, obviously AI-generated format. Two are below, but there were many more. I remember a forum post saying that ChatGPT proposals were verboten, but it's prevalent and done openly with impunity. I asked one why he didn't simply submit a normal proposal and he said: "I am using chatGPT to help me respond better. I am not very good at explaining details but if you wonโ€™t pick me thatโ€™s fine. Have a nice day."

 

That response shows me that, assuming Upwork is legitimately forbidding ChatGPT in proposals and content, there needs to be a better messaging mechanism if these guys don't know they're violating the Terms. 

 

As a small-time client, it's annoying to pick through ChatGPT proposals to find the real ones. But more importantly for me as a freelancer, it's giving professional clients a poor experience that's going to make them avoid the platform. 

 

**edited for Community Guidelines**

โ€ƒโ€ƒ

63 REPLIES 63
prestonhunter
Community Member

I have hired over 180 freelancrs on Upwork.

 

I totally understand that ChatGPT use is causing problems for you and others.

But it does not affect me because I don't read proposals.

You may want to consider changing how you hire freelancers.
I have long found that reading proposals is a waste of my time. So I don't read them, and I block them from being written when the user interface allows me to do so.

"But it does not affect me because I don't read proposals."

 

Great. But other clients do. If they have a poor experience and leave the platform, it will undoubtedly affect you. I'm not asking how you or others manage ChatGPT proposals client-side, but rather whether Upwork has plans to address its use.

re: " I'm not asking how you or others manage ChatGPT proposals client-side, but rather whether Upwork has plans to address its use."

 

I don't know if there is anything Upwork can do to prevent it.
I believe that clients will need to be the ones who decide for themselves how to handle it.

"I don't know if there is anything Upwork can do to prevent it."

 

I didn't expect you to. The post is targeting the official moderator staff who do know what initiatives are or are not in development. 


Preston H wrote:

I don't know if there is anything Upwork can do to prevent it.


Make it a ToS violation and suspend or ban freelancers when clients flag their proposals.

How, though? Freelance writing groups are a sea of seasoned writers talking about how long-time clients have accused them of using AI because they ran the content through one of the wildly inaccurate AI detectors--false positives would be off the charts. And, ChatGPT allows you to request that it write content that doesn't appear to be written by AI, and seems to do so fairly successfully.


Tiffany S wrote:

How, though? Freelance writing groups are a sea of seasoned writers talking about how long-time clients have accused them of using AI because they ran the content through one of the wildly inaccurate AI detectors--false positives would be off the charts. And, ChatGPT allows you to request that it write content that doesn't appear to be written by AI, and seems to do so fairly successfully.


I think it would be pretty easy to spot AI in a proposal situation, because freelancers would all be using the exact same project description and therefore generate very similar results. (Quoting the OP: "I received a flood of identical proposals.") I don't expect that Upwork would actually suspend anyone who got caught - just like they don't really enforce their ToS in other cases where people are caught plagiarising their bids and profiles - but it would be nice to at least have a rule in place, in case it serves as a bit of a deterrant.

Unfortunately Upwork deleted the screenshots from my post despite PII sanitizing them. But the two example proposals, which were representative of the many, were structurally the same and just very obviously inorganically generated. Most tellingly is that both regurgitated quirks in my own communication as keywords from the post and used them out of context. 

Moreover, both examples and all others flat-out misrepresented actual freelancer ability. One said he had a soundproof, top-notch recording studio and the other said he had 5 years of professional journalism experience. Would a professional with those experiences and qualifications, in either case, rely on ChatGPT to generate proposals? Absolutely not.


Which is why, even if there's an argument to be made that the technology is useful as an assistive function, outright fraud and misrepresentation in these cases is unequivocally a violation of the spirit and letter of Upwork's terms and should be prosecuted on that basis alone. 

Consider me as your new freelancer.

 

What exactly do you do then? Check just the Profile?

re: "What exactly do you do then? Check just the Profile?"

 

Well, I definitely don't read proposals. Waste of time.

 

I look at a freelancer's portfolio.

 

I look at their skills list.

 

If I am trying to hire an illustrator or artist, for example, I hire a freelancer whose skills list identifies her as such. I don't hire a freelancer whose skills list identifies her as a writer or Excel expert or baker.

 

I might look at work history.

 

The most important thing is evaluating freelancers AFTER hiring them.

 

I don't assume that I can predict which freelancers will work out for a project. If I hire them and they do great work, then I let them continue. If I don't like their work, I end the contract.

Hi Perston,
But most of the clients read the proposals and select the freelancer on a proposal basis. 
If some are not reading like you, its ok. But those who read, what will they do?

 

re: "But most of the clients read the proposals and select the freelancer on a proposal basis. If some are not reading like you, its ok. But those who read, what will they do?"

 

If reading proposals has been beneficial to other clients, that's great!

 

If reading proposals is now becoming frustrating to a client because he sees a deluge of AI-generated content, then the client should pivot... change his strategy. I'm saying from first hand experience that a client can be very successful WITHOUT reading proposals.

 

It is very freeing for a client to know he can hire freelancers without any commitment, and fire them quickly if they don't work out. That is one of the great benefits of using Upwork. It is very unfortunate when clients don't understand the power they have.

Are you looking to hire more freelancers, consider me?

colettelewis
Community Member

I doubt if Upwork has the mechanism (yet) to prevent it.  I suppose all a client can do is to put in their job offer that any AI proposals  will be binned, or report the freelancers who do it, which would be too time consuming.

 

The  examples you have given look more like proposals for B & M jobs; they are overlong, overstuffed, and rather old-fashioned.  But to get dozens of this kind of proposal must be incredily irritating. One can only hope that the novelty of ChatGPT will eventually wear off - especially if Google gets hoist with its own petard and has to cope with a tsunami of identical articles and websites that are already overcrowding the web. 

Upwork staff posted in the forum that ChatGPT is forbidden unless disclosed; if they're making pronouncements disallowing certain actions there should be an actual mechanism to address it.  


Nichola L wrote:

I doubt if Upwork has the mechanism (yet) to prevent it.  I suppose all a client can do is to put in their job offer that any AI proposals  will be binned.


That's a great idea.

 

6bfcdaf8
Community Member

A client "could" use the proposal content to assess how good this person is in terms of expressing themselves in written format. Also the language skills as most freelancers wouldn't have english as a native language.

 

If they used google translate, you'd easily understand because it wasnt very successful. Now that they use ai tools i believe you can still understand the unnecessary doublespeak in ai generated proposals, it is now less useful in understanding freelancers skills.

 

But i mean does it really have any impact on who gets hired? You'd anyway have a short zoom call and see who this person actually is.


Alper D wrote:

But i mean does it really have any impact on who gets hired? You'd anyway have a short zoom call and see who this person actually is.


I wouldn't bother having a Zoom call with anyone who's so lazy and/or inarticulate that they can't read a project description and write their own proposal. If clients are flooded with a bunch of identical spam proposals, they'll be more likely to get fed up and not hire anyone. So, that's bad for all freelancers and for Upwork. 

"more likely to get fed up and not hire anyone. So, that's bad for all freelancers and for Upwork."

 

Yep, that's critical and why Preston's "just don't read proposals" is useless in toto. I don't care about reading through AI proposals as a client; I'm small-time on that end and often just looking to clear some things off my plate. But other clients, "real" clients, will either see the flood of regurgitated proposals that say nothing or hire off of one and be sorely disappointed by the results. That client, that firm, their associates and friends - all have a tainted view of Upwork in the end, avoid the platform, and affect my bottom line as a full-time freelancer. 

I think I've had Zoom calls with three Upwork clients before they hired me...in nearly 7 years.

feed_my_eyes
Community Member

You can try flagging these proposals, and maybe if Upwork receives a flood of complaints from clients, they'll take it more seriously and ban these people.

That's about all I can do, and have been.

Please let us know if you hear anything from Upwork or if you notice any impact from doing this.

shehanfernando
Community Member

Maybe Upwork could include a tool like ZeroGPT that can detect the amount of AI generated content in a proposal. That'll make the clients skip through the ones that have high AI generated content  ๐Ÿ™‚ 

Thanks to you i learned something new today!

You're welcome Alper ๐Ÿ™‚ 

Unfortunately, at the moment, those tools are horribly inaccurate.

Agree Tiffany, let's hope these tools will improve in the future. Here's a preview of Turnitinโ€™s AI detection tool - 

**Edited for Community Guidelines**

When there really aren't any unique clauses left in the English language (or most major languages, for that matter) unless a new word gets added to the language, I don't see how these tools won't flag most work as somehow plagiarized once their dataset reaches critical mass.

 

Less than 5 years ago, my father wrote a grad school essay for cyber-security that was flagged by software as plagiarizing 5 different textbooks. The professor was going to report him for academic dishonesty until he saw my father's job description (Raytheon, IT dept) and realized that he wrote that essay based entirely on his niche work experience. The conundrum? There are only so many ways to describe a topic, especially when the description is instructional.

There is a very simple fix for the "conundrum." Cite your sources.

When the only source is the experience of the person writing, that's not a possibility.

Citing sources doesn't seem to appease this type of tools. I've seen many freelancers flagged by Copyscape for a direct, attributed quote.

6a182ed6
Community Member

I use Upwork quite a bit as a "paid StackOverflow". I post concise questions that an expert can easily answer; and I pay for their expertise.

 

The thing that is most frustrating with ChatGPT is the plethora of situations where:

1. freelancer is not competent to solve task

2. freelancer feeds prompt into ChatGPT

3. freelancer doesn't test the ChatGPT 'solution', doesn't realize the solution is broken, and has no idea how to fix the solution

4. freelancer sends me the ChatGPT 'solution'

5. the solution uses the right keywords / libraries, so it sounds plausible on a superficial glance

6. freelancer disappears the moment I drill down on problems with the 'solution'

 

The issue here is that there is zero downside for the freelancer to spam untested/incorrect ChatGPT 'solutions'

If you pay them and leave poor feedback explaining the problem, that will hit their JSS (assuming they aren't Top Rated (Plus) with an available feedback removal perk).

Absurdly short-sighted because it assumes clients:

 

1. Will pay to be Upwork police by leaving feedback for freelancers violating the letter and spirit of Upwork's direct communication on the matter.

 

2. Will try again on the platform and post another job with the known potential of wasting more time and money on fraudulent freelancers.

 

You absolutely cannot, and should not, expect clients to do either. "Just pay and tank his JSS bro" is not a viable solution to a problem that could lead to an exodus of quality clients and affect my bottom line.

It is an established psychological fact that people will hurt themselves to see another person hurt more, if they are angry at that person. Multiple studies have demonstrated this.

 

But that's not the point. The point is, TongKe believed that there is no consequence for freelancers who lie by submitting AI-generated answers - which is not true. There are consequences, so long as the Client leaves feedback. Even if the FL uses their Top Rated (Plus) feedback removal perk, that is still a consequence, because now they must wait 3 months minimum before they can use it again.

No. The defrauded client will dispute and win arbitration in a heartbeat. Then they'll be off Upwork forever, punishing the rest of Upwork's legitimate contractors as that revenue stream is lost forever. I can't fathom why you're so enthusiastic about being pedantically correct over the most minor of points while completely missing the forest for the trees. But whatever floats your boat

I believe that it would be very unwise for any client to think that they can accomplish anything through dispute or arbitration if it turns out that a freelancer they hired used ChatGPT or similar AI tools.

 

Every client should proactively accept this:
"I am not getting money back."

 

To be cliear: If I ask a freelancer to do something, and the freelancer uses ChatGPT, and I found out that the results are unsable, I AM NOT GETTING MONEY BACK.

 

Upwork is not going to look at my "proof" of ChatGPT use and give me money.

That is not how it works.

 

If a client is worried about unusable work produced by ChatGPT, the client needs to evaluate work that freelancers submit.

Upwork is NOT going to evaluate a freelancer's work.

Upwork has no way to physically block a freelancer from submitted ChatGPT work, regardless of what its policies state.

Upwork is not going to give clients money if they discover any work they received was generated by ChatGPT.

Latest Articles
Upcoming Events
Learning Paths