Professionalism and Copy Editing AI-Generated Content
July 9, 2024 11:00 AM Subscribe
I am a freelance copy editor. Some of my clients are very upfront about using AI-generated text and ask me to edit their content to no longer resemble AI. Fine. Others are not upfront, which is a separate matter. Either way, what are an editor's professional, ethical, and legal obligations to address these concerns with their clients?
First things first:
- You are not my lawyer (I am not a lawyer, either). This is not legal advice.
- I don't love AI but I want to be clear that I'm not here to debate whether it should be used; I really do want to stick to the matter of "what are an editor's professional, ethical, and legal obligations to address these concerns with their clients?"
- I've edited enough content that was generated by AI (a client would tell me "I used AI to write this") that I've come to notice its tell-tale patterns and signs. From what I've observed, generative AI's syntax range is not only limited, but profoundly formulaic, and tends to fall back on the same adjectives, idiomatic expressions, narrative structures, and syntax.
- As an editor, I give my clients advice about tone, style, narrative structure, and everything in between. To me, this should include pointing out when writing sounds like AI. I want to be able to say "Hey, I noticed this uses language and syntax that is similar to what generative AI uses. I don't want your audience to accuse you of using AI, which could harm your credibility, so would it be okay if I make some sweeping changes to this text beyond the scope of what we discussed?"
- Yes, I understand assuming something was written by AI is very different than having proof something was written by AI. I know people have been wrongly accused of using AI. I don't want to embarrass myself or someone else, but I'd like to know if there's a way to (gently) thread the needle.
I guess my biggest question is: When a client doesn't disclose the use of AI, yet their text reeks of formulaic AI-isms ... should I just shut up and not say anything? (This is basically what I've been doing up to this point.) Or is it in poor taste to suggest their writing sounds like AI, let alone may have been written by AI? Do I need to tell them to include a disclosure with their writing or discuss it with an attorney?
First things first:
- You are not my lawyer (I am not a lawyer, either). This is not legal advice.
- I don't love AI but I want to be clear that I'm not here to debate whether it should be used; I really do want to stick to the matter of "what are an editor's professional, ethical, and legal obligations to address these concerns with their clients?"
- I've edited enough content that was generated by AI (a client would tell me "I used AI to write this") that I've come to notice its tell-tale patterns and signs. From what I've observed, generative AI's syntax range is not only limited, but profoundly formulaic, and tends to fall back on the same adjectives, idiomatic expressions, narrative structures, and syntax.
- As an editor, I give my clients advice about tone, style, narrative structure, and everything in between. To me, this should include pointing out when writing sounds like AI. I want to be able to say "Hey, I noticed this uses language and syntax that is similar to what generative AI uses. I don't want your audience to accuse you of using AI, which could harm your credibility, so would it be okay if I make some sweeping changes to this text beyond the scope of what we discussed?"
- Yes, I understand assuming something was written by AI is very different than having proof something was written by AI. I know people have been wrongly accused of using AI. I don't want to embarrass myself or someone else, but I'd like to know if there's a way to (gently) thread the needle.
I guess my biggest question is: When a client doesn't disclose the use of AI, yet their text reeks of formulaic AI-isms ... should I just shut up and not say anything? (This is basically what I've been doing up to this point.) Or is it in poor taste to suggest their writing sounds like AI, let alone may have been written by AI? Do I need to tell them to include a disclosure with their writing or discuss it with an attorney?
Or is it in poor taste to suggest their writing sounds like AI, let alone may have been written by AI?
I am only answering from a legal perspective here. I'm not a lawyer, but as you said, you probably shouldn't be looking for situation-specific legal advice on the internet anyway.
I'm not entirely sure what your concern is from a legal perspective. If you are asked to edit content to not sound AI-generated, and that content sounds like it was AI-generated, then you should suggest changes that result in it not sounding AI-generated, regardless of whether the content actually was AI-generated.
I know people have been wrongly accused of using AI
Saying "this content sounds AI-generated" is your professional, highly qualified opinion. It doesn't mean the content was AI-generated. Further, even if it wasn't AI generated, so long as you don't make a public statement that includes no caveats (ie, "I worked for xyz, and their content is AI generated"), it's unclear how it could be construed as libelous.
Privately providing guidance on style is exactly what your job is as an editor. Any client who finds that distasteful is not a client you should work for.
posted by saeculorum at 11:15 AM on July 9 [11 favorites]
I am only answering from a legal perspective here. I'm not a lawyer, but as you said, you probably shouldn't be looking for situation-specific legal advice on the internet anyway.
I'm not entirely sure what your concern is from a legal perspective. If you are asked to edit content to not sound AI-generated, and that content sounds like it was AI-generated, then you should suggest changes that result in it not sounding AI-generated, regardless of whether the content actually was AI-generated.
I know people have been wrongly accused of using AI
Saying "this content sounds AI-generated" is your professional, highly qualified opinion. It doesn't mean the content was AI-generated. Further, even if it wasn't AI generated, so long as you don't make a public statement that includes no caveats (ie, "I worked for xyz, and their content is AI generated"), it's unclear how it could be construed as libelous.
Privately providing guidance on style is exactly what your job is as an editor. Any client who finds that distasteful is not a client you should work for.
posted by saeculorum at 11:15 AM on July 9 [11 favorites]
If they gave you writing that sounded like it was written by a 12-year-old, what would you say? If they gave you writing that sounded like it was written by a non-native English speaker with certain grammatical tells, what would you say? If they gave you writing that was overwrought, florid, purple, dry, inscrutable, whatever... AI-style is another stylistic characteristic. You're right that audiences will react to it negatively in some cases. So it seems you can address it in the same way you address other characteristics that may be detrimental. Stick to talking about how it reads and how audiences may react, not how you think it was created.
posted by whatnotever at 11:18 AM on July 9 [10 favorites]
posted by whatnotever at 11:18 AM on July 9 [10 favorites]
Is your question 'are there legal or ethical concerns with using AI that they should be aware of and if so, what are they and should I be the one to tell them?'
Or is your question, 'I know there are legal or ethical concerns with using AI which I have not mentioned in the question to avoid sidelining this discussion into whether those are really concerns, should I tell them?'
Or is your question, 'I am not sure if my client knows their staff is using AI to generate copy, should I tell them?'
posted by jacquilynne at 11:19 AM on July 9 [2 favorites]
Or is your question, 'I know there are legal or ethical concerns with using AI which I have not mentioned in the question to avoid sidelining this discussion into whether those are really concerns, should I tell them?'
Or is your question, 'I am not sure if my client knows their staff is using AI to generate copy, should I tell them?'
posted by jacquilynne at 11:19 AM on July 9 [2 favorites]
Response by poster: Is your question 'are there legal or ethical concerns that they should be aware with using AI that they should be aware of and if so, what are they and should I be the one to tell them?'
Or is your question, 'I know there are legal or ethical concerns with using AI which I have not mentioned in the question to avoid sidelining this discussion into whether those are really concerns, should I tell them?'
Or is your question, 'I am not sure if my client knows their staff is using AI to generate copy, should I tell them?'
The first and third questions. The second question is something I might bring up in a private conversation with a friend (in other words, yes, I have my "concerns" about what the legal and ethical concerns could be) but I'm responsible enough to know I need to seek the advice of subject-matter experts before presenting those concerns to a client, if at all.
posted by nightrecordings at 11:26 AM on July 9
Or is your question, 'I know there are legal or ethical concerns with using AI which I have not mentioned in the question to avoid sidelining this discussion into whether those are really concerns, should I tell them?'
Or is your question, 'I am not sure if my client knows their staff is using AI to generate copy, should I tell them?'
The first and third questions. The second question is something I might bring up in a private conversation with a friend (in other words, yes, I have my "concerns" about what the legal and ethical concerns could be) but I'm responsible enough to know I need to seek the advice of subject-matter experts before presenting those concerns to a client, if at all.
posted by nightrecordings at 11:26 AM on July 9
Is your question 'are there legal or ethical concerns that they should be aware with using AI that they should be aware of and if so, what are they and should I be the one to tell them?'
Your job as a copy editor is to edit copy and you are not a lawyer. Providing legal advice to your clients is way outside your job qualifications, and could actually result in legal issues to you.
I don't personally think it's appropriate for a copy editor to tell clients it's unethical to use AI-generated copy. You're well within your rights to refuse to work on it, but if you accept that you work on it, then it seems strange to tell clients it's not ethical to use it. I would find it strange to conclude that editing AI content is ethical but using AI content is unethical.
posted by saeculorum at 11:31 AM on July 9 [1 favorite]
Your job as a copy editor is to edit copy and you are not a lawyer. Providing legal advice to your clients is way outside your job qualifications, and could actually result in legal issues to you.
I don't personally think it's appropriate for a copy editor to tell clients it's unethical to use AI-generated copy. You're well within your rights to refuse to work on it, but if you accept that you work on it, then it seems strange to tell clients it's not ethical to use it. I would find it strange to conclude that editing AI content is ethical but using AI content is unethical.
posted by saeculorum at 11:31 AM on July 9 [1 favorite]
Saying "this content sounds AI-generated" is your professional, highly qualified opinion. It doesn't mean the content was AI-generated.
I think this is it. That leaves it up your client about how to proceed.
If I were paying you to help me put out professional content, this is feedback I'd want to hear. You're the editorial expert!
posted by pantarei70 at 11:51 AM on July 9 [4 favorites]
I think this is it. That leaves it up your client about how to proceed.
If I were paying you to help me put out professional content, this is feedback I'd want to hear. You're the editorial expert!
posted by pantarei70 at 11:51 AM on July 9 [4 favorites]
I've overseen content production at a corporation (still do to a lesser extent) and employed countless copyeditors.
If you came back and said "this content sounds AI-generated" that would strike me as a totally normal interaction - through I would also expect that you'd be returning content to me that no longer sounded that way.
If you said anything else - about morals, ethics, legality, etc. - I suspect I wouldn't continue working with you, because that would be both bizarre and beyond your role in the process.
I can create masterful content, rich in data, and with clear insights and guidance. Many of my colleagues struggle with one or more parts of that, but still have something valuable to continue to the content process. The ability to pair then with AI and get a finished product, even if it needs a heavy edit afterwards, will be a wonderful thing.
posted by NotMyselfRightNow at 12:03 PM on July 9 [3 favorites]
If you came back and said "this content sounds AI-generated" that would strike me as a totally normal interaction - through I would also expect that you'd be returning content to me that no longer sounded that way.
If you said anything else - about morals, ethics, legality, etc. - I suspect I wouldn't continue working with you, because that would be both bizarre and beyond your role in the process.
I can create masterful content, rich in data, and with clear insights and guidance. Many of my colleagues struggle with one or more parts of that, but still have something valuable to continue to the content process. The ability to pair then with AI and get a finished product, even if it needs a heavy edit afterwards, will be a wonderful thing.
posted by NotMyselfRightNow at 12:03 PM on July 9 [3 favorites]
I want to be able to say "Hey, I noticed this uses language and syntax that is similar to what generative AI uses. I don't want your audience to accuse you of using AI, which could harm your credibility, so would it be okay if I make some sweeping changes to this text beyond the scope of what we discussed?"
Without getting into the weeds of ethics and legality, you can just say "AI tends to use a very specific style, and so these days any writing in that style sounds like an AI wrote it, which readers sometimes have negative reactions to / which can affect how readers react to the text. How do you feel about me making style changes beyond the scope of what we discussed to make the phrasing less similar to the usual AI style?"
That said, doing that kind of editing might affect the amount you bill them, which would potentially be an ethical concern on your end.
posted by trig at 12:13 PM on July 9 [4 favorites]
Without getting into the weeds of ethics and legality, you can just say "AI tends to use a very specific style, and so these days any writing in that style sounds like an AI wrote it, which readers sometimes have negative reactions to / which can affect how readers react to the text. How do you feel about me making style changes beyond the scope of what we discussed to make the phrasing less similar to the usual AI style?"
That said, doing that kind of editing might affect the amount you bill them, which would potentially be an ethical concern on your end.
posted by trig at 12:13 PM on July 9 [4 favorites]
I think one obligation you have, that you likely already deal with outside of the "AI" lens, is to be sure that each individual client understands the scope of your copy editing does not include content editing or "fact checking." While that is something you surely already deal with in your discussions and client education, the added complexity with AI is that sources cannot be checked (or known), so there's really no way for even your clients to guess what might be an entire fabrication.
posted by neutralhydrogen at 12:17 PM on July 9 [3 favorites]
posted by neutralhydrogen at 12:17 PM on July 9 [3 favorites]
Best answer: An ethical concern for you as an editor is that in a sense, it is not possible to edit AI output. Here's what I mean: normal editing is bringing an unpolished text closer to an ideal—you're preserving the author's intent, and making the text more like what the author would have written if they were better at writing for their intended audience. But AI output has no intent to preserve, it's just plausible-sounding word salad, and editing it is just rearranging the croutons. I'll admit this argument probably won't get a lot of traction among generative AI users.
There may be a business argument against using generative AI: I can only speak for myself, but as soon as I suspect I'm reading AI-generated copy, I stop reading it. I doubt I'm the only one.
posted by adamrice at 12:35 PM on July 9 [9 favorites]
There may be a business argument against using generative AI: I can only speak for myself, but as soon as I suspect I'm reading AI-generated copy, I stop reading it. I doubt I'm the only one.
posted by adamrice at 12:35 PM on July 9 [9 favorites]
Best answer: This sounds like a good question to bring up on the email list for the Editorial Freelancers Association, if you’re a member. Or the Copyediting-L email list (I’d link to these but I’m on a mobile device that’s being wonky). They’re both easily searchable if you’re not already aware of them. I’ve found my membership in the EFA valuable, though I’m not on the email discussion list currently. And over the decades of being in this field, I’ve learned so much from CE-L.
posted by kitten kaboodle at 1:39 PM on July 9 [4 favorites]
posted by kitten kaboodle at 1:39 PM on July 9 [4 favorites]
Best answer: > I want to be able to say "Hey, I noticed this uses language and syntax that is similar to what generative AI uses. I don't want your audience to accuse you of using AI, which could harm your credibility, so would it be okay if I make some sweeping changes to this text beyond the scope of what we discussed?"
As a client this is exactly what I would want (and need) to hear even if that is not really what I want to hear. But from a practical perspective with this kind of writing the harm is #1. Factually incorrect content that everyone along the way just kinda accepts as true because the AI said it and no one knows any different or has a way to fact check it, and then it gets outs there & harms your reputation. And #2. Your ad copy, web page, or whatever sounds like it was written by AI so people immediately click away and also assume your business is run by incompetent nitwit. Or putting it more professionally, it will strongly affect the business's reputation.
FYI any more I click away from a page the instant I detect any trace of AI generated text. The bullshit factor is just far, far too high. I'm not going to waste one second of my time on that garbage and any business that relies on it is just tarring itself instantly as stupid and lazy. (Not to put too fine a point on it.)
So I think you need say and push for exactly that kind of revision and if it is so extensive as to require higher fees, then that as well.
The only dilemma for you is what to do if you present that option and the client refuses it. Are you OK with proceeding with the smaller job and having your work coming out sound like AI? Or will you just refuse the job.
One helpful thing might be to address some of these issues up front as you set expectations with clients. Maybe one the skills in your bullet points is "Ensuring that your communications have the right tone and vocabulary for the intended audience, and does not have any hallmarks of machine-generated text. Maybe one of your pricing tiers is some extra $$$ for "Re-writing as needed to turn AI generated text into text with the appropriate tone, format, and vocabulary to reach its intended audience."
FWIW I think one of the legitimate uses of AI is in helping people who tend to get writer's block or just can't get needed writing done, to get things written more quickly and easily. It can be most helpful in the sense of, "I've put down a few basic thoughts about what I want to get across, now I need to put it in a more presentable format. Or can we make it easier to read or more in style X."
But in my mind, that is more in terms of helping people get to a first draft. There is no way you can just take that output as the final version.
posted by flug at 1:50 PM on July 9 [2 favorites]
As a client this is exactly what I would want (and need) to hear even if that is not really what I want to hear. But from a practical perspective with this kind of writing the harm is #1. Factually incorrect content that everyone along the way just kinda accepts as true because the AI said it and no one knows any different or has a way to fact check it, and then it gets outs there & harms your reputation. And #2. Your ad copy, web page, or whatever sounds like it was written by AI so people immediately click away and also assume your business is run by incompetent nitwit. Or putting it more professionally, it will strongly affect the business's reputation.
FYI any more I click away from a page the instant I detect any trace of AI generated text. The bullshit factor is just far, far too high. I'm not going to waste one second of my time on that garbage and any business that relies on it is just tarring itself instantly as stupid and lazy. (Not to put too fine a point on it.)
So I think you need say and push for exactly that kind of revision and if it is so extensive as to require higher fees, then that as well.
The only dilemma for you is what to do if you present that option and the client refuses it. Are you OK with proceeding with the smaller job and having your work coming out sound like AI? Or will you just refuse the job.
One helpful thing might be to address some of these issues up front as you set expectations with clients. Maybe one the skills in your bullet points is "Ensuring that your communications have the right tone and vocabulary for the intended audience, and does not have any hallmarks of machine-generated text. Maybe one of your pricing tiers is some extra $$$ for "Re-writing as needed to turn AI generated text into text with the appropriate tone, format, and vocabulary to reach its intended audience."
FWIW I think one of the legitimate uses of AI is in helping people who tend to get writer's block or just can't get needed writing done, to get things written more quickly and easily. It can be most helpful in the sense of, "I've put down a few basic thoughts about what I want to get across, now I need to put it in a more presentable format. Or can we make it easier to read or more in style X."
But in my mind, that is more in terms of helping people get to a first draft. There is no way you can just take that output as the final version.
posted by flug at 1:50 PM on July 9 [2 favorites]
Best answer: One other consideration is that AI generated text, for the reasons you mentioned, may require a much more substantial edit to give it a clear compelling voice. So you want to make sure that you are getting paid for that higher level of effort and contribution that would be needed to make the work fully usable. Building on adamrice's comments, you basically become the author (or at least the co-author) and should be paid for that expertise to the extent that it requires more skill and/or more effort than a conventional editing job.
Also, if you don't already have this in your contract, it might be appropriate to a clear disclaimer that the editing process is only about working with the information given to you and the client has a responsibility to review the final material to make sure that is true and accurate.
posted by metahawk at 1:55 PM on July 9 [4 favorites]
Also, if you don't already have this in your contract, it might be appropriate to a clear disclaimer that the editing process is only about working with the information given to you and the client has a responsibility to review the final material to make sure that is true and accurate.
posted by metahawk at 1:55 PM on July 9 [4 favorites]
Or is it in poor taste to suggest their writing sounds like AI
You’re hired to fix text, not investigate where it came from.
[the text’s] syntax range is not only limited, but profoundly formulaic, and tends to fall back on the same adjectives, idiomatic expressions, narrative structures, and syntax.
I’m not sure at what point you’re willing to kick the text back to the business and say that fixing it would require a complete rewrite, but these would seem like reasonable grounds for such an action.
posted by Tell Me No Lies at 2:15 PM on July 9
You’re hired to fix text, not investigate where it came from.
[the text’s] syntax range is not only limited, but profoundly formulaic, and tends to fall back on the same adjectives, idiomatic expressions, narrative structures, and syntax.
I’m not sure at what point you’re willing to kick the text back to the business and say that fixing it would require a complete rewrite, but these would seem like reasonable grounds for such an action.
posted by Tell Me No Lies at 2:15 PM on July 9
Taking the client's poorly communicated, poorly defined brief and turning it into something they can actually use is the service I provide as a freelancer. The AI angle of this sounds like an interesting topic of discussion with your colleagues or professional group but I don't see what the ethical concerns are in regards to your job.
In my creative freelancing practice, it has always been the norm that clients will present you with briefs or pitches that contain the work of other people as references. Sometimes these other people are the creatives you were bidding against on the project! Now, instead of using random work taken from random portfolios/websites, it has become just as common to see AI generated assets used in these briefs.
Which one of these situations is more ethical? Ultimately the reason they are hiring me is because they are unable to communicate effectively what they want, and the references they provide are just a starting point to begin that conversation because they don't have the vocabulary to tell me. Unless someone is literally asking for direct plagiarism I don't see what the issue is.
posted by bradbane at 2:58 PM on July 9 [4 favorites]
In my creative freelancing practice, it has always been the norm that clients will present you with briefs or pitches that contain the work of other people as references. Sometimes these other people are the creatives you were bidding against on the project! Now, instead of using random work taken from random portfolios/websites, it has become just as common to see AI generated assets used in these briefs.
Which one of these situations is more ethical? Ultimately the reason they are hiring me is because they are unable to communicate effectively what they want, and the references they provide are just a starting point to begin that conversation because they don't have the vocabulary to tell me. Unless someone is literally asking for direct plagiarism I don't see what the issue is.
posted by bradbane at 2:58 PM on July 9 [4 favorites]
« Older Help me buy a bed! Any reason not to buy a... | Restaurants with tea leaf tempura in NYC? Newer »
You are not logged in, either login or create an account to post comments
posted by nightrecordings at 11:14 AM on July 9