Latest / Talk Copy to Me | Content + Copywriting Podcast / The Importance of Editing and Fact-Checking Your AI Content

Transcript

The Importance of Editing and Fact-Checking Your AI Content · Talk Copy to Me | Content + Copywriting Podcast

• 0:00
  Your AI just wrote you a blog post and it sounds great.
• 0:05
  So what are you doing?
• 0:07
  Are you about to hit publish?
• 0:09
  Don't do it because here's the thing, AI makes stuff up.
• 0:14
  It gets facts wrong, and sometimes it loses your voice completely.
• 0:19
  Worst, it can copy what other people have previously created before you, and you won't catch those problems unless you know what to look for.
• 0:40
  We have spent the last few episodes talking about building your AI foundation, training it with your brand voice, using it for strategic brainstorming and repurposing content across formats.
• 0:54
  Using ai but now comes the part that most people skip.
• 0:59
  I guarantee you have skipped this before.
• 1:02
  That is editing and fact checking, AI produced content.
• 1:09
  Here's what I see happen.
• 1:10
  Constantly.
• 1:11
  Someone gets AI to write content.
• 1:14
  It looks pretty good.
• 1:16
  Maybe they fixed a few typos, maybe they adjusted the intro, let's say, and then they publish it.
• 1:23
  However, they later find out that maybe AI included outdated information made up a statistic, or let's say it quoted someone who actually never said that thing.
• 1:37
  I actually see this in podcasting quite a bit where people will use AI tools to repurpose their podcast.
• 1:44
  Transcripts, let's say into show notes or into social media posts, and it will create quotes that actually were not said in the podcast episode.
• 1:54
  It might be generally about what the, the meaning was.
• 1:57
  It might be paraphrased content, but it will act as if it is an exact quote where it is not.
• 2:05
  Not cool ai.
• 2:07
  So what's happening is when people see this content, they're seeing content that technically sounds fine, but it isn't correct.
• 2:17
  And imagine how you'd feel if you found out quotes were made up or stats were wrong, or you were showcasing something incorrectly.
• 2:25
  Well.
• 2:26
  Today we're gonna talk about quality control.
• 2:28
  This is the most important thing you can be doing if you're using ai.
• 2:35
  We're gonna talk about how to edit AI content for messaging, voice strategy, and how to fact check everything before it goes live.
• 2:46
  Because this is not about being paranoid, it is about being professional.
• 2:52
  It is about being trustworthy, being credible, because AI may be a powerful tool.
• 2:59
  It needs human oversight.
• 3:02
  And it's what I've been saying.
• 3:03
  All of these episodes be strategic, be the conversation leader, be the decision maker.
• 3:09
  Now you have to be the editor and the fact checker.
• 3:15
  So let, let's talk about why.
• 3:17
  We'll start with that.
• 3:18
  Why does AI content need humans to edit it?
• 3:22
  Your AI will generate content based on the patterns of its training data, and you know, it's really good at sounding confident, at sounding authoritative.
• 3:33
  Even when it is completely wrong, you are gonna hear an example soon on the podcast, um, when Emily Rittenbach joins me and she talks about how she asked Google, oh gosh, I'm gonna get this wrong, but how to turn.
• 3:47
  A Google feature off, I think it was like AI within her Google email, and it gave her all the instructions on how to actually turn it off.
• 3:56
  Turns out, oh, it, it was the AI overviews.
• 3:58
  Turns out you can't turn it off, like literally can't turn it off.
• 4:02
  But Google's AI tool gave her complete instructions on how to do something that don't you think it should known.
• 4:09
  That cannot be done so.
• 4:12
  It m, it's very easy to get missed by you if you're reviewing the content that it creates.
• 4:18
  But beyond just the factual accuracy, AI does not understand your business context the way that you do.
• 4:26
  And if you're not giving it good context, it definitely does not understand.
• 4:31
  It does not know the nuances of what matters to your audience.
• 4:34
  It doesn't catch when something is technically correct, although it is maybe sending the wrong message.
• 4:44
  It doesn't understand when you need to break your own rules too.
• 4:47
  You know, like what in the writing world?
• 4:49
  We talk about how we learn creative writing rules, so that we can break them.
• 4:54
  You know, we're taught in high school, in elementary school not to start sentences with the word and or, but not to have run-on sentences.
• 5:02
  There are so many specific rules that we have about writing.
• 5:07
  Only when we know them can we start to break them for creative or professional reasons, but AI doesn't actually know that.
• 5:16
  It doesn't know when it can break the rules, or in some instances it doesn't even actually know what the rules are if you're not feeding it well-written, style guidelines and something that people don't talk about often.
• 5:28
  AI can lose your voice even if you've trained it well.
• 5:32
  I say this a million times.
• 5:33
  I might have said in the last episode that, I don't know, maybe I did not.
• 5:37
  But you can have a good output from the beginning of a conversation with one of your AI tools and notice that the more editing you're doing within the tool, the more chance there is to lose your voice and tone.
• 5:51
  Because maybe when you're giving it instructions to fix something, it then defaults to a more corporate language.
• 5:57
  , Maybe it uses words that you would never use.
• 6:00
  Maybe it also softens your perspective so that you seem more agreeable or worse.
• 6:06
  It creates a very challenging and headstrong tone, and that's not what your, your meaning is behind what the message you're trying to put out there.
• 6:18
  These types of changes can actually go unnoticed, and what happens is it's adding up to content that just doesn't feel like you.
• 6:25
  In the end, when I'm working with my copy coaching clients, a huge part of what we're doing together is reviewing and refining their writing.
• 6:33
  Even good writers need another set of eyes on what they're too close to see.
• 6:39
  It is why it is so hard for all of us to write our own websites and AI generated content needs that same level of attention and care and editing.
• 6:52
  Actually, it needs more.
• 6:54
  So let's start with my two layer editing process.
• 6:57
  And like I said, this is not gonna cover everything that I wish I could tell you, but I think it's a good starting place for you to make sure that you're.
• 7:06
  Fact checking and editing as possible.
• 7:09
  First, you need to approach your AI content editing in two distinct layers or phases.
• 7:16
  The first is the in tool editing through better prompting and refinements.
• 7:22
  And the second is the human element of editing everything that you can't handle within the AI tool, and you absolutely need both layers.
• 7:32
  Let's start with in tool editing, this is about using AI to improve its own output before you even start doing any human editing.
• 7:42
  It's faster than fixing everything manually, so I definitely highly recommend you start within the tool.
• 7:50
  But it also teaches AI what better actually looks like for your content.
• 7:55
  And it gives you an opportunity to ask it to help you, you know, refine your prompting so you get better output in the future.
• 8:03
  It is vital that when AI gives you your first draft, you do not accept it, even if it looks good, from a quick once over.
• 8:10
  I want you to give it specific feedback about what needs to change or has a possibility, like an opportunity to change.
• 8:19
  Not just saying, please make this better.
• 8:23
  That is useless.
• 8:24
  That's useless.
• 8:25
  We talked about this last time, but if you, if we say something like the tone is too formal.
• 8:29
  Let it know that say this is too corporate.
• 8:32
  Rewrite it using a more conversational language.
• 8:35
  Like I'm talking to a friend who feels frustrated with this problem I'm addressing.
• 8:40
  If the examples are too generic, you could say something like, you know, these examples are too vague or too generic.
• 8:47
  Please use this specific business example.
• 8:51
  , And maybe you then give the example.
• 8:53
  Or if you mention something in the prompt, say, please refer back to the prompt for the specific instructions, if the structure isn't working, say this jumps between ideas, you know, there's no transitions here.
• 9:07
  Add clear connections between each section that explains why we're moving from one topic to the next.
• 9:15
  You are coaching AI to improve.
• 9:18
  The same way that I coach my copy coaching clients.
• 9:22
  You need to be specific about what's not working and what success looks like so that they can achieve.
• 9:30
  When I say that, I mean that's my job as a coach is to help my clients become better at whatever it is that they want to improve at.
• 9:39
  But if I'm not specific with them to tell them how to make those improvements, or if we don't find what that success looks like, it's impossible for us to both be successful in our jobs.
• 9:49
  So now you are gonna think of yourself as the copy coach and the AI is your client.
• 9:55
  And the first thing that you need to do is in tool editing.
• 9:58
  As your first pass, get the AI as close as possible to what you need before you start manual editing.
• 10:06
  It's gonna save you time and it's gonna create a better final result.
• 10:11
  But now comes the part that absolutely requires you to put in some effort.
• 10:17
  Human editing will cover messaging, storyline, voice style, tone, grammar, fact checking, but we'll get to that in a second.
• 10:28
  You'll notice I'm not talking about fact checking for a second because I'm gonna go into it deeper.
• 10:33
  If your AI helped to create the content, what you are doing now is making sure it actually is serving the business goals.
• 10:39
  So let's start with messaging.
• 10:42
  Does the content say what you actually want it to say?
• 10:46
  That's important.
• 10:47
  Is the main point clear?
• 10:49
  Are you leading with benefits?
• 10:51
  Are you just getting bogged down in features?
• 10:54
  Is the value proposition obvious?
• 10:57
  All those questions are really important to ask when you're editing AI content for messaging.
• 11:04
  I see this all the time.
• 11:05
  It is technically about the right topic, but the messaging is off, or maybe it's explaining how something works when what you really need to focus on is why it matters in the first place.
• 11:19
  Another one that I see happen all the time is that it's talking to the wrong audience segment.
• 11:24
  It could be trying to solve a problem that your audience doesn't actually have.
• 11:28
  So read the AI content through the lens of your own business strategy and not just for accuracy.
• 11:36
  And you're thinking right now, I know you are messaging.
• 11:38
  Why did she not talk about voice and tone?
• 11:41
  Well, that's 'cause I'm going to talk about it now.
• 11:43
  Point two is it's not just for the actual message, it's for the way that the message is shared.
• 11:52
  And we do that by things like our brand guidelines.
• 11:54
  So what is the voice and tone?
• 11:58
  Did it lose your voice?
• 12:00
  Are there words and phrases you'd never use?
• 12:03
  Does it actually sound like you or is it just generic business content that's really, could be stated by anyone?
• 12:11
  Is there actual personality that is consistent throughout the piece, but key point here, it's also consistent with the other content you have created.
• 12:23
  One thing I highly suggest doing is reading the content out loud.
• 12:28
  Literally, I did this through my entire way through graduate school of writing my thesis, running a literary journal, and reading through hundreds.
• 12:38
  I mean now it's thousands and thousands of submissions, choosing what we're gonna publish and then reading our final edit.
• 12:45
  We always, we as in writers, always read out loud.
• 12:50
  Does it sound like something you'd actually say?
• 12:52
  Is the tone off are there contractions that need to be made?
• 12:56
  I find often that it's usually instead of the contraction, haven't, let's say I see a lot of AI output will have, the two words have not, or I will instead of I'll it will be, I will.
• 13:11
  You'll notice when you read those out loud that without the contraction, it feels off.
• 13:16
  Because we, as humans use contractions in our normal day-to-day language.
• 13:21
  If you want it to sound like you, you need the contractions.
• 13:25
  And final note here, watch for AI's tendency to kind of hedge everything with words like perhaps or may or might.
• 13:36
  I think it does that because it wants to, again, impress us.
• 13:40
  It wants us to feel like it's doing what we're asking it to.
• 13:45
  So it doesn't necessarily like stake its claim in the ground, especially if your voice and tone is more friendly.
• 13:52
  It does not want to sound too direct, but by using words like that, you're watering down your message.
• 13:59
  All right, let's move on.
• 14:00
  Let's talk about editing for Story and Flow.
• 14:03
  Does the content move logically from one point to the next?
• 14:06
  This is literally one of the biggest things that I always see as issues.
• 14:12
  No transitions and not a logical like.
• 14:16
  Start here and end here.
• 14:18
  Format.
• 14:19
  It's disconnected between ideas.
• 14:21
  There's no clear, middle, beginning, and end.
• 14:24
  All of the content could really be good.
• 14:26
  It just needs to be reorganized and transitions need to be added.
• 14:31
  Let's move on to style.
• 14:33
  Is the formatting appropriate?
• 14:35
  Are paragraphs a readable length or do you need to break up paragraphs more and give them more white space?
• 14:42
  Are there headers and subheads that make scanning easily?
• 14:46
  Is emphasis used appropriately or is.
• 14:51
  Everything bolded in everything italics and tons of exclamation points.
• 14:59
  Emphasis is there for a very specific reading reason.
• 15:03
  So don't let your AI tool run wild and add emphasis wherever it wants because it will be well over emphasized.
• 15:14
  Before we stop talking about formatting, a lot of AI tools are now using, a watermark, which could be a space.
• 15:22
  Every time there's a new paragraph, so you know, you press the return button, a new paragraph starts.
• 15:28
  There is one space to the left before the writing starts.
• 15:34
  We all know a lot of the times there are quotes in our AI output, so you wanna make sure those quotes are gone.
• 15:41
  You wanna make sure that, . You don't have any of those digital watermarks that showcase that it's AI content.
• 15:49
  But in regard to style, it's also a place where you can catch things like repetitive sentences, repetitive paragraphs that I feel like that's something I also notice a lot.
• 15:58
  You know, AI's trying to make a good point and it makes it by saying the exact same thing in two paragraphs in a row.
• 16:06
  Like, we got you.
• 16:07
  Like, that's another way to show emphasis, right?
• 16:09
  Like.
• 16:10
  You made your point, please move on to the next point.
• 16:13
  Before I move on to fact checking, let's talk about editing for grammar and clarity.
• 16:17
  This is really the final polish you don't need to be going through when you're doing all of the other things and changing, you know, periods and commas and m dashes.
• 16:28
  Because m dashes are a real thing.
• 16:30
  People like, we use them in the creative writing world all the time, . You don't need to touch those things while you're editing for messaging or style or tone or story.
• 16:41
  Now's the time you fixed everything else.
• 16:44
  So get in there and fix the typos, adjust the punctuation, smooth out, awkward phrasing if it's there.
• 16:51
  And just make sure that what you're saying is clear and necessary and don't get stuck here.
• 16:57
  I see people obsess over commas, for example.
• 17:01
  But what they're actually missing is bigger picture issues.
• 17:04
  Grammar does matter, but it's your last layer.
• 17:07
  It's not your first.
• 17:09
  All right.
• 17:09
  Now's the very important part.
• 17:12
  That's hard for me to figure out how to teach in one podcast episode, and I, I don't think I'll be able to, but what is critical?
• 17:23
  Critical if you're gonna use AI is fact checking because this can seriously damage your credibility if you skip it.
• 17:34
  So your job is to verify everything that sounds like a fact before you hit publish.
• 17:40
  What, Erin, am I fact checking?
• 17:44
  I'm glad you asked.
• 17:45
  I want you to fact check anything that seems like a statistic or a data point.
• 17:52
  Any quote that is attributed to someone.
• 17:55
  Any claim, this is where people often don't, don't follow through, but any claim about.
• 18:04
  I would say anything, but we could at least to get you started, say a claim about research or about studies, any reference to a historical thing, event or a current event, especially current events, any.
• 18:22
  Uh, process description or, you know, technical information, kind of like how I mentioned Google, giving Emily instructions on how to do something that you cannot do within Google.
• 18:33
  Oh, if there's, you wanna scan and see if there's any legal or regulated information.
• 18:38
  This can be scary to everyone, but I think you for the most part know if you are in an industry that has some regulated information and what you can and cannot say.
• 18:47
  Uh, for example, nutritionists or people who may be in weight loss ish positions know that they cannot make specific claims about, you know, products guaranteeing.
• 19:01
  Weight loss or anything like that, your AI tool might do that.
• 19:05
  So you need to make sure it does not, because well wouldn't, and one more.
• 19:09
  Anything that's time sensitive, why is this important?
• 19:14
  Because.
• 19:14
  Claiming that you don't know or you didn't write it or that your AI tool did, it is not an actual excuse.
• 19:20
  It will never work.
• 19:22
  So you have to fact check so that you don't end up in that situation.
• 19:28
  How do you fact check?
• 19:29
  Well go to the primary sources.
• 19:34
  I have had a bone to pick with the world for well over a decade now about primary sources.
• 19:41
  Friends, you cannot like link to a, you know, M-S-N-B-C article about a scientific study and, and say that that's actually correct information because the news reported it.
• 19:53
  You can't.
• 19:55
  You can't, you need to go to the actual study.
• 19:57
  The study, which is likely in a journal is the primary source, so when you find things online that you do want to link to, and I highly recommend that because that's what Google wants with their EEAT.
• 20:10
  You know, being able to present quality information, present quality primary sources, if it's something that you have to, because it's not your own expertise, so you want to be doing these things.
• 20:22
  But it needs to be a primary source because if AI is citing a study, find the actual study and verify what it says.
• 20:31
  If AI is quoting someone, find the original quote.
• 20:36
  If AI gives you a statistic, track that thing down.
• 20:40
  Don't just Google and see an AI overview and assume that it's correct.
• 20:44
  AI overviews are also wrong sometimes.
• 20:48
  Or often you need to actually verify the information from trustworthy sources for technical or specialized information.
• 20:56
  You know, if it's in your field, you can use your own expertise.
• 21:00
  Does this match what you know to be true?
• 21:03
  If something feels off, you need to investigate it, but if it does match what you know to be true, cool.
• 21:09
  That is your own expertise that counts.
• 21:11
  You are the primary source.
• 21:14
  Fact checking.
• 21:16
  Is really hard.
• 21:18
  Any editor out there that works for big brands, that works for the newspapers, that works for any media outlets, media outlets that actually do fact check their sources.
• 21:27
  But they know this takes time like an effort and it's exhausting, which is why I stopped editing things myself in a career because.
• 21:40
  Sometimes it takes in, in my opinion, from being both the writer and the editor, it can take just as long as it would take me to write an article that I know has good sources, because I wrote it myself.
• 21:51
  I researched it myself as it takes a fact checker or an editor to make sure that actual information is correct, to make sure what I'm saying from my sources are correct to fact check everything.
• 22:04
  So.
• 22:06
  I mean, this is kind of like why I didn't do an actual episode on just content creation in this miniseries, because it's complex.
• 22:14
  You know, I want you to know about fact checking.
• 22:17
  I want you to know about editing because even in repurposing content, it's vital.
• 22:22
  You do these things, and I say that because you might be thinking, but it's my content.
• 22:27
  Like, do I really need to fact check my own content?
• 22:29
  Yes.
• 22:31
  Yes you do.
• 22:32
  Yes, with an exclamation point.
• 22:33
  You need to fact check everything.
• 22:36
  So I don't know.
• 22:38
  Just put the effort in here and spend a lot of time.
• 22:43
  Any claims, any sources, any facts, check them.
• 22:50
  Well, let's talk about a couple of, AI fact problems, so you understand what to look for, and it's not just me preaching at you all episode.
• 22:58
  Here's some things that actually goes wrong.
• 23:00
  AI will combine information from different sources, and when it does that, it often creates new facts that are not true.
• 23:08
  So two pieces of relevant information could be true, but when it's put together, it might make a statistic from one study and a, excuse me, and a conclusion from another study, feel connected, where in fact the statistic from one study could have the complete opposite result from the conclusion in, in the second study.
• 23:29
  AI also tends to present information.
• 23:32
  Again, like I said this before, in a very certain matter of fact way, if you're a researcher, you might say something, and I've said this as a writer who's done an extensive amount of research, you might say something like, this is suggesting a possible connection between this and that, right?
• 23:51
  Quality researchers know when they can, when they need to be clear that there's some ambiguity.
• 23:59
  I'm not sure if that's the right word here, but AI will use words like, well, this proves that X, Y, Z. No, it doesn't prove anything.
• 24:08
  Sometimes research just shows how things are related, but it's not proving anything.
• 24:15
  AI training data has a cutoff date.
• 24:18
  We all know that.
• 24:19
  We never seem to know when it changes, so.
• 24:22
  It doesn't have information about what happened after that date, and it might generate content that sounds current, but is incorrect . I'm trying to tie this up.
• 24:32
  I know I'm, it's getting long, but you know, editing and fact checking cannot be something you do.
• 24:36
  Just when you remember.
• 24:38
  It needs to be built into a workflow as a non-negotiable step.
• 24:43
  So take those SOPs out right now if you've already created them and, and add these notes in, or,, take a damn post-it note and write these things down.
• 24:53
  Step one in tool refinement.
• 24:56
  After the content is generated or repurposed, review and give specific feedback for improvement.
• 25:03
  Step two, a strategic review.
• 25:07
  So it's not just feedback to the output, it is a review to make sure it's still matching messaging, strategy, business goals, that it's talking to the right audience.
• 25:19
  Even step three, voice and flow.
• 25:22
  Does it sound good?
• 25:23
  Does it sound like you, you know, does it flow?
• 25:27
  All extremely important.
• 25:29
  Step four, a fact checking pass.
• 25:32
  Verify everything, no exceptions.
• 25:36
  Step five, final Polish grammar formatting, clarity.
• 25:41
  It's your job now to make it publication ready and.
• 25:46
  If possible, have someone else review it, . Make that part of the process.
• 25:50
  If you have team, team members, you do that in tool review.
• 25:54
  You do the quick fact checking to review, to make sure it's correct, then pass it off to someone else and have them determine whether it sounds like you have them determine if you know it needs any content added or or explanation somewhere.
• 26:09
  It might sound like a lot.
• 26:10
  I know it sounds like a lot.
• 26:11
  It is a lot, but each step is gonna get faster as you are, doing it over time.
• 26:17
  And in truth, it doesn't make a difference how much work this is.
• 26:22
  If you need to protect your credibility, which you need to as a business owner, this is worth the time investment.
• 26:31
  All right.
• 26:32
  Let me address the one key thing that I do think you should worry about, and that is plagiarism.
• 26:40
  AI does not intentionally plagiarize, but it can reproduce content that it was trained on, especially for.
• 26:49
  Obviously common phrases, but well-known concepts.
• 26:53
  But if it's taking a heading or if it's taking notes from someone else in your field, from their own website, from their own blog content, from their podcast or YouTube channels, that's a problem.
• 27:07
  So.
• 27:08
  That's an episode on its own.
• 27:10
  We have had had an episode on plagiarism before.
• 27:12
  I'll link to it in the show notes, but use plagiarism checkers, which are not always correct if you are concerned, especially for your longer pieces of content.
• 27:22
  And more importantly, make sure that the final content is actually reflecting your own unique.
• 27:28
  Personality and experience and voice.
• 27:31
  This is why in the example, last week, for short form to long form content, I left in an actual placeholder to add a personalized personal story in to add a.
• 27:44
  My own expertise in to the end result.
• 27:47
  So that way, no matter what, it could still be my unique perspective.
• 27:53
  I need you to do the same thing, right?
• 27:55
  You can worry so much less about plagiarism if you are making sure you're doing these things to be a good citizen of the internet, to be a good citizen business owner, Your goal is to create content that adds your unique value and perspective.
• 28:13
  That's what makes content worth reading, and that's what helps protect you from plagiarism concerns.
• 28:19
  But while this episode cannot really address and full the plagiarism question and talk about it, I do want you to be aware.
• 28:29
  It's definitely a possibility.
• 28:31
  Is why fact checking is also so important.
• 28:35
  We are actually too long here, so you know, summing it up, editing and fact checking, they're non-optional.
• 28:41
  They are essential if you are gonna be using AI responsibly for content creation, repurposing.
• 28:47
  You need to do them because the goal is not just to make perfect output.
• 28:51
  The goal is to use AI to speed up these things that you're doing in your business, like creating content for social media, like creating emails for your newsletter.
• 29:02
  You wanna maintain quality, you wanna maintain accuracy, and you want your audience to get an output that it deserves.
• 29:11
  In next week's episode, we're gonna be talking all about AI for SEO, the basics of how traditional SEO may differ from AI optimization and what you need to know before we dive into the AI overview episode that I hinted at today with Emily.
• 29:27
  That will come just a few days following that.
• 29:30
  Episode, so two episodes for you next week.
• 29:33
  Just a reminder, ai, it's a tool.
• 29:35
  It is not a replacement for your judgment, your expertise.
• 29:40
  Edit accordingly.
• 29:42
  Thanks for listening.
• 29:43
  I'll see you next week where we keep talking.
• 29:45
  Copy.