We challenged ourselves to write a Grow
Convert blog post that's published on
the GNC site, and this is the first time
we're telling anyone that we published
a post that was written with AI.
As I was writing the article,
I did start at the top, and I
went down through the sections.
Towards the end, I was getting
better at using AI, and I was
learning, like, when it was useful
to use Notebook versus Claude.
And then you sent it to me.
My first, first reaction was
like, oh, this is pretty good.
And then, as I started editing and reading
it, Me knowing that Caitlyn normally is
like a great writer, I started to get
annoyed where I was just like, man, I
gotta like edit like all this stuff.
And if Caitlyn had just written the
piece, I wouldn't have to edit all this.
We've
been getting a bunch of questions about
AI and more specifically AI writing.
Uh, we've done the last two videos AI and
then we talked through, uh, AI search.
And in this one, we wanted to talk
through AI writing specifically.
We challenged ourselves to write
a Grow and Convert blog post
that's published on the GNC site.
And this is the first time we're
telling anyone that we published
a post that was written with AI.
And so we brought Caitlin, uh,
again to, to explain how she
wrote this post category keywords
that is published on our site.
And so we kind of just wanted
to talk through how we, how we
approached, uh, writing it with AI.
What tools we used.
Uh, the pros and cons of using AI, whether
we're going to continue doing this and
just kind of talk through all of the
experience using AI to write content.
And importantly, uh, also how
much of it was written by AI.
We, we tried our best to write it with AI.
Caitlyn edited it.
I edited the AI.
You edited me.
Yeah.
Caitlyn edited the AI and then I edited
some version of the AI slash Caitlyn
and made predictions along the way,
some of which were not correct about
what was written by AI versus Caitlyn.
And I should note that Caitlin is,
uh, an incredibly talented writer
on our team, so, um, I thought I
could very easily distinguish, but,
um, some of them were harder, but
some of them were pretty obvious.
Um, Caitlin, starting off with kind
of the beginning of the process of,
would be great, like, what you remember
we asked you to do, and then, and how
you, how you started thinking about it.
Sure.
So I think we had, um, some really
brief conversations just about, You
know, what would it look like to bring
AI in and getting that started and I
think you mentioned a couple things
about, like, hey, you can't just
tell it to write the whole article.
You know, it's going to be
better if you give it a bit
of an outline and all of that.
Um, because I was coming
at this brand new.
So, like, I hadn't done anything with AI.
Um, I wasn't, I hadn't even opened the
most popular, you know, AI chat tools.
Um, so I was starting from ground zero.
And so, anyone who's worked with AI,
you know, naturally is going to know
that there's a big learning curve there.
Um, and, yeah, that's, that's exactly
what I found, is that there's just
so much to learn in terms of how to
prompt the AI to get what you want.
Um, That it's, it's really it's own skill.
Okay, which, which AI tools
did you start with first?
And I should mention to everyone
that the keyword that we were
going after for GrownConvert's
own blog was category keywords?
Was that the actual SEO keyword?
Is just the term category keywords?
Well, it's one that we're
trying to coin, right?
I think it has volume.
There is volume, but it's not used in
the same way that we're talking about it.
And there's like a lot of
different, if you look at the
cert board, when I was writing it.
I'm not sure what it looks like now.
Um, every post was a little bit different.
And like, even when I went to run
ClearScope, so I could add some
relevant terms in, one of the
relevant terms was like running shoes.
And I was like, what does
this have to do with anything?
You know?
So, it just, there wasn't any,
like, Set understanding, um, from
what Google was producing anyway.
Okay, I think, I think that's
a good jumping off point.
I actually think instead of, um,
talking about which tools we used at the
beginning, although maybe we can, it'd
be good to just get to the crux of the
matter for everyone listening, which I
think you kind of hit at, and maybe we
can use that as a launching point to
get into the details of what you did,
which, as you said, that I mentioned.
And by the way, when Caitlin
said, I came into this new and
Davis mentioned, you can't just,
um, ask it to write a blog post.
You have to ask it to outline.
I'm also new at it.
So just for the record, we
have published zero pieces for
clients that are written by AI.
Our natural growing converts culture
is that we are, um, very particular,
one may say elitist about our writing.
We're kind of proud of that.
I'm happy to take responsibility
for setting that culture.
Cause I'm kind of proud
of setting that culture.
And so like, well, that's our natural
thing, but we don't want, want to be,
you know, Um, ostriches with our head in
the sand as AI is, is, is taking over.
And we're getting close
to publishing this survey.
We, we did on a bunch of writers
and how much they're using AI.
And the long story short is
that a bunch of like 80 percent
plus are using it in some way.
And I think the better way to think about
this is this is AI assistance in writing.
Almost no, we did not in this,
even in this little project
we're talking about today and in
the survey, nobody's just like.
Well, I shouldn't say nobody, but
the vast majority are indicating
that they're not just like.
Hey, like just AI, just write
it and they're just using it.
Copy and paste from
chat, GBT into the post.
So this outlining, um, can you walk
through or do you remember how much
you outlined and what you gave it?
Yes.
While Caitlin's pulling that up just
so that everyone can follow along, the
vast majority of people listening and
following us are in month marketing,
specifically content marketing.
So it'll help you be able to follow
the details of what we talk about in
terms of the writing being good and bad.
If we just give you some
background context of this keyword.
So you can follow along if you
don't know, grow and convert, right?
We write bottom of funnel content.
Um, meaning we write for keywords
that we think the act of Googling,
it shows massive buying intent.
And one of those, we
call that pain point SEO.
You can Google that to,
to read more about it.
One of those pain point
SEO, like sub categories.
Is category keywords.
There's three of them, category keywords,
comparisons, and jobs to be done.
And category keywords in our terminology
means that very specific thing.
Keywords that when you type it into
Google, it's like, you're looking
for a product in that category.
Examples we always use is best
accounting software for us.
It would be content marketing
agency, things like that.
And in fact, you mentioned
running shoes, running shoes.
In my mind is kind of a category keyword,
you know, because that means there,
someone is, if you are an e commerce
company that sells running shoes,
that's a category keyword for you.
So I will say that.
When I started out here, you know,
I was experimenting and I wanted
to see how far I could push it with
the AI or how much I could have the
AI do the work for me, essentially.
And so, I was trying to give it as little
as possible and then by the end, you know,
I was, uh, doing a lot more on my own and
just having it help me word little things.
Um, but if we look at the beginning here,
so I started, this is my initial prompt.
I told it I was going to have it help me
write an article, and then I described,
uh, a little bit about who the audience
is, what their understanding is.
Wait, let me pause you there just for,
so, this episode will be one that it's
Useful to look on YouTube, but I know
a decent amount of people just listen
to us on Spotify and Apple podcasts.
This is not intuitive.
I think and so If you're using AI and
you want to like for writing assistants,
you want to get better This itself is
interesting when Caitlin says I described
a little bit about the audience Let me
just read the beginning of her prompt
out loud for you Help me write an article
that targets the keyword category.
Keywords fine.
The reader is marketing professionals.
Um, specifically those involved
in SEO and content marketing
for SAS and B2B companies.
These individuals are likely to, and she
has 1, bullets that I can see 6, 7, 8, 9.
Wow.
Bullets and, and they're very detailed.
Like.
Like some of 'em are.
These people are likely to be focused
on generating leads and conversions.
They're tasked with
driving business results.
Another bullet, and that's only part of
it, is be frustrated with low conversion
rates from their current SEO efforts.
They're struggling to
get tangible results.
These people are likely to not know
what we mean by category Keywords are
likely to underestimate the importance.
I mean, there's like multiple
sentences, like this is carefully.
Thought out.
That's really interesting.
Yeah.
I also think that's important just
because most people don't like, even
when I've prompted, I don't even think
of giving it this background information,
but the more information you give it,
I think we found that it really just
helps you get better results from it.
And what's interesting about your prompt
is you didn't, it's not like you said
that and then there's like space and
then you're like, now write this article
with this outline, you just hit enter.
You didn't even really ask this.
And by the way, she's
showing this on Claude.
This may be a separate discussion
of comparing different tools.
I think we should focus on the
process and what we wrote here.
But we've noticed that, and we've said
this in previous videos this year, when
we were talking about AI, that Claude
writes better, at least in our taste.
Is that intentional, Caitlin, that
it's not like you wrote this and
then you said, And, and now, now here
are the characteristics of my target
audience, now do X for me, or whatever.
You just hit enter.
Yeah, it just got to be a lot to
put all in, um, one prompt, I guess.
I think maybe I was running out
of character count or something.
Um, what I wish I would have done, looking
back, is asked it, like, I mean, it
kind of does spit back what it thinks.
You're saying, which is kind of
nice, but I wish I would have
asked it more specifically.
Like, okay, what are you learning from
this or, um, what are you not going to
include because of this or something?
You know, I'm not sure, um, obviously
I still have more to learn there, but
it would have been more interesting
to like check its learnings
essentially at this point, which
I didn't do, but looking back, I
think that would have been helpful.
Um, but I think I was just trying to like
get that out there and then, It spit back
a response without me asking for anything.
We're really expecting
it to give me anything.
And then I did give it this outline.
Yeah.
If you scroll back up for a second,
it just, it just sort of says.
Yeah, yeah, that's fine.
I can help you write this and then it
just is like I've already structured
the article Right and and and what's
interesting is it says it's just
it doesn't give an outline But it
structures it and it says the content
and number one of its five bullet
Structure I guess is of what it thinks
the article should do acknowledges
their current approach and pain points
Benji That's what you're saying.
That's actually kind of interesting
I've never done this like as Caitlin
saying I played around with AI a
little bit for the same reason to
I don't think it would have done that.
Yeah, it wouldn't have.
If you hadn't spent ten bullets
or whatever we said being like,
I'm looking to write this article.
Here is a ton of information about
the people's like pain points.
And in our writing process for years.
We have taught and, you know, in our
course and to people on our blog and
everything and to our own team of pain
point resonance and starting these really
understanding the client's pain point.
That's really interesting.
And you're saying this, now
what you're showing, this is
what that sample article was?
Before you gave a outline or anything?
Yes, it does actually give
more than just this outline.
Oh, it straight up wrote an article?
Yeah.
Yep.
Without you asking it to write an article.
Interesting.
That's really interesting.
And then it says like that,
it already has its category.
What are category keywords is the first
H two, why most B2B SEO fall short.
Which I, again, I would say is
probably guided by you doing that.
That's actually kind of interesting.
Yeah.
I mean, if you just scan the headers,
like at first glance I was impressed.
I was like, oh, this is interesting.
Yeah.
That's how I feel right
now looking at this.
I haven't seen this before.
Even the how to part
looks pretty interesting.
Yeah, how to identify category keywords.
Start with your product category.
Yeah.
Okay, and then if we keep going in
your story, what Yeah, what was next?
What did you do next?
So then I gave it an outline
that I was working on.
Let's see if I can see it.
Is that because you looked at what
it produced and you were like, you
said at first glance it was good,
but then you decided it was bad?
Or you were just like, no,
I already wrote an outline.
I'm not gonna use what you produced.
Yeah, I think it was
missing, if I remember right.
It was missing quite a few
points and just a lot of detail.
How close is this to an outline that
you would typically do for an article?
What she produced, you mean?
Yeah, like, I'm curious, just, I mean,
we have a step in our own writing process
where we do a questionnaire and a lot
of people do an outline in that step.
Everyone does.
Okay, so everyone does.
And I'm just curious.
Did you do, did you think of the outline
slightly differently because you were
doing it for AI or is this pretty close to
what you would do in your questionnaire?
That is a very good question.
But I'm actually looking at this
and it's saying the sources also
highlight that something, and
that sounds like notebook to me.
It was constantly referencing,
uh, sources, and so I'm wondering
if I'm missing a step here, and
that's what I'm trying to find.
Or, or you used notebook to
help you create this outline.
So you probably fed it some
material, and then Maybe ask
for an outline and notebook.
Benji, we should mention to the
audience, they're talking about notebook.
L M, which is another.
I think it's another Google product,
but I think both Claude and and notebook
are Google products are anthropic.
I think is the one who owns maybe
Claude is made by anthropic, and I
think which just Google has funded
or Google VC or someone has funded.
So there's some relationship between them.
But notebook is straight up
a Google product notebook L M
and notebook L M. Thank you.
I forgot how we heard about it,
Benji, but it's, I heard about
it through a friend and it's,
what I heard is it's really good.
It's like a research AI.
And so what's really interesting about
it is compared to chat GPT or Claude,
you can actually drop links in here
as sources so you can upload a YouTube
video or a podcast and it straight learns
everything from those sources and so
you could put like 10 different sources
so you could put, let's say, One of our
YouTube videos that describes our product.
So we have a video on competitor
keywords that we've done that
explains what competitor keywords are
and all of our thoughts around it.
And then you could put individual blog
posts that grow and convert has written
in here, uh, and then use it to understand
a bunch of information from all those
sources and then query it based on that.
Yeah.
That's where I found, actually, Notebook
was really helpful with that, because it
would only pull from those sources, then.
Whereas, like, Claude kind of, you know,
pulls from wherever and makes stuff up.
Like, Notebook would actually,
um, cite its sources.
You can see here, like, these numbers
reference the sources over here, and I'm
not sure how to, like, get into here.
Oh, saved responses are view
only, so that's the issue.
But I was able to, before, click on these
and it would take me right to where in
the source it was pulling that from.
And so I found that It produced better
content in that way where it was
more accurate and more detailed and
sometimes it would even pull details
from the sources that I had forgotten.
And so that was a nice thing because it's
able to like keep track of all the details
better than a human mind in a lot of ways.
Hold on, this is actually
a really big point.
I'm thinking about this as the
person who's been reviewing those
survey results of the 50 writers
and writing up the analysis of that.
And, and I don't recall, uh, and
I've read every single long form
response to our questions on that.
Um, I don't recall anyone mentioning this.
I mean, as Benji is quick to note, I
also have the worst memory ever created.
So I reserve the No, but, but I
think this tool is something that
most people don't know about yet.
It's pretty, it's pretty new.
I think, I think when we started using
it, I think it was the first week or month
that it was really Getting the future.
Yeah.
No, it was, we were really early on to it.
I mean, it might've been around for
a while, but I think people were just
starting to talk about this, and I
think even still, most people don't
know about it because it's not, it's
not like chat GPT where it's, you use it
for, I don't know, solving math problems
and just doing anything that you want.
This is, it's has a pretty specific
purpose, which is uploading, which
is this research summarization thing.
Give it a bunch of sources.
And then it, so let me just, let me
just say the big picture again that I
want to emphasize that Caitlin said.
And then we'll get into the details of
what we're showing on the screen here
from Notebook for, for, for everyone.
Including people that are viewing because
they may be like, what's going on here?
A huge thing is that Caitlin
said, this Notebook takes a bunch
of sources that you give it.
And it like the purpose of what it's
doing and what Benji saying is like it
seemed the vibe we're getting and I think
notebooks like landing page Benji is a
consistent with this is that it's almost
like for students that it says like if
you're researching this thing, you can
upload a ton of stuff, your videos,
some class worksheets, or I don't know,
I don't know what you do, but like you
put all the materials and that's there's
like three columns we're looking at.
That's the left column.
And then on the right most column.
What is that?
Caitlin is like, It's
summary of all of that?
Yeah, it's understanding of
everything that you gave.
So you don't have to like,
ask it to do anything?
It just The notes?
Yeah.
This is pieces of the chat
that I was able to save.
Yeah, so you can go back, but you
can ask it like, it's understanding
of the various sources and stuff.
So the way it works is you upload
as many sources as you want on
something, and then you can chat
with it, and like, query that?
Is that the idea?
Yeah.
Yeah.
So like, yeah, you can see
here with these questions.
And the, and so that's what the
bigger picture is like, this is a huge
difference that you should be thinking
about if you're, if you're, if you're
exploring AI assistance in writing
chat, GBT, Claude, the two big other
ones is you're writing something.
And when you ask it, either of
those two, like to help you write
it, it's drawing from, it's like.
Huge amount of training data, which
has basically been fed of like
every written work or, or whatever
in human history, plus like Reddit
and the web now and all this stuff.
Right?
And so that's why Claude in that
first example was gave a draft
basically of a category keywords post
without Caitlin having asked it to
because it's like category keywords.
Yeah, it's in my training data of what
like the Internet has said about it and
humans have said about it historically.
And I think that's been our grow
and converts frustration is that
a thesis of growing convert.
It's so strong to our philosophy in
content marketing is interview subject
matter experts and the brand and
express their views in our pieces.
And when we.
Use the term Google research paper in
kind of a derogatory or mocking way.
And you can Google that exact
phrase and site grown convert.
com to figure it out.
Is we we're saying everybody else,
this mirage content we've been talking
about is writers just always kind of.
regurgitating what the first few
results on Google say about a
topic, which are beginner levels.
And no matter what topic you're
writing about, it's beginner level.
And so if you use Claude and chat, GBT
to write without feeding it a bunch of
information, you end up doing that Google
research paper, but just AI assisted
version of it because Claude and chat
GBT are just going to regurgitate.
What they've been fed
on a particular topic.
So if we, Grow and Convert, have a very
specific view of this term category
keywords, that view is not going to be
expressed if you asked Claude and ChachiBT
to write it unless you very carefully
hold their hand to exactly what you
want it to say and you teach them that.
Because they're going to
resort to like what the entire
internet has said before you.
Versus Notebook LM is only, it just
takes the sources you give it, and
then you can chat with it, and it's
just drawing upon those sources
to answer all of your questions.
So then, you can build an outline that's
based on the knowledge of, from the things
you give it, which is very different.
This is bringing it back now.
Going back to the outline, or working
with Claude, is The article that it gave
me, it was missing all of that unique
voice or messaging that is, you know,
in all of these resources over here
from your episodes, from, you know,
the articles we've written everything.
It's much more than voice.
It's, it's the argument.
It's all the arguments.
And so that was the difference.
Even when we thought about how
do we use AI to write, we thought
this was very important was to.
To give it all of our past information.
So to give it videos on category
keywords, where we describe category
keywords in detail to talk, to give
it pain point SEO, which category
keywords is a part of that framework.
So if it has all these different
sources of our arguments around what
category keywords are, then it should
be able to write a much better post
than if you were to try to query AI to
write it without all this information.
I would say better in the
sense that it's more accurate.
The arguments are more
what we want them to be.
But what I did run into with
Notebook is that, like, you
can't change the language of it.
You can't tell it, like, right in a,
you know, professional but casual tone.
I tried to even at one point, you
know, I mentioned earlier that, um,
It would constantly, like, say, from
this source, such and such a thing.
Like, I tried to specifically tell it,
don't ever mention a source, don't use
the word source, and it kept doing it.
So, um, there's certain It's
not a writing tool, right?
Yeah.
Right, right, exactly.
And so that's where I did keep
using Claude, and I would feed
it, as we saw here, what I was
able to pull from Notebook, then
I would feed it into Claude.
And, uh, see if it could give me the
same information, but in a, I don't
know, tone, I guess, that I was wanted.
I will say, um, I, I noticed that
in the survey responses of other
writers that are, that are using
AI, there was a decent amount that
liked using ChatGPT, no, not a single
person, I think, mentioned NotebookLM.
So you're right, Benji, like, I
didn't realize it was that new, but
we seem to be The ones hijacking a
like student research tool for writing
because we it's so important to us
that you form content pieces based on
background information from the brand.
Uh, a lot of people liked it because
they actually liked querying the Internet
because they don't have those sources.
I'm just noticing like a lot
of the writers who were who
are responding to the survey.
It's like.
They're being tasked to
write Google research papers.
This is what clients are paying them for.
Like that, like that's what they're given.
And they're like, well, this is,
this is how I make my, like, like,
this is what I'm being asked to do.
And so I almost don't blame them.
We're like, and that's what we've
said forever is clients are literally
like, here's the keyword, look
forward to the post, you know?
And it's just like, like, they're not a.
I don't know, like a, a CrossFit expert,
but the brand, the client is a CrossFit,
you know, equipment brand, and they need
to write this thing, and so they're just
like, okay, like I'm gonna write this
fitness thing, and they're not the fitness
expert, so it's almost convenient to
them that they can be like, hey Claude,
write about this thing, because they
would normally have to Google, step one,
then kind of summarize their own findings
in some, some fashion, however they do
it, step two, Then consolidate that in
their heads and write it, step three.
That's like the traditional writing
process for humans forever, and this kind
of lets you do all three steps in one.
Yeah, I mean I can see from that point
of view it being really attractive.
Yeah.
It could be worth getting through
the pain of learning how to prompt
it, if suddenly you can produce
ten times as many articles.
You know, I don't know.
But if, if you are a brand or a writer
or whoever that wants to up level your
writing, because if you, you do it in
that way, if you're asking chat GPT and
Claude to also give you the background
information to also do the web research,
basically, and write in one, like, like
you tell me all of the arguments in this
post, like, it's like, give me a, give me
a thing on category keywords by definition
is going to just be, Summarize what the
average of what the internet has said
on category keywords in the past, like
if that's not good enough for you, like
it's not for us and you want to implement
the brand's perspective, then like
this is actually very, very necessary.
You either need to do what Caitlin's
saying she did for this experiment,
which is use some other AI tool
like notebook LM to do the summary.
Of, you know, and, and she had on the
screen, like, you must've had what?
10 things on their notebook.
LM, you had like multiple YouTube videos,
multiple articles that we had published.
That's a lot, multiple
articles and YouTube videos.
So if you want to do it that way,
and you're uploading a bunch of your
client or your company's information.
You either need to figure out how
to do that inside ChatGPT and Claude
natively, which would be like, they're
not gonna accept YouTube videos, right?
Or like podcasts, so you're gonna need to
like copy and paste or transcribe them.
Yeah, the difference would be in
Claude, you would have to copy and
paste the entire text of the article
or the transcript from videos.
And then you need to tell Claude, you need
to be, in my experience, you need to be
very careful and be like, Use only, like,
use only these arguments from in here.
You need to ask it first,
like, summarize this.
And then be like, now write something
using only this information.
Otherwise, it's gonna just grab
stuff from its training data.
Alright, that's, this is very interesting.
Let's keep with the chronology.
So you've used Notebook to kind of
summarize all of the stuff the Grow and
Convert has said on, on these topics,
and then you fed that back into Claude.
Yes.
Yep.
So that's what this is here that I
pasted that in, if I remember right.
I didn't use just the raw outline,
I guess that Notebook gave me,
I did edit it some and I picked
and choose some things and then.
You know, uh, put that in here,
so it wasn't just straight AI tool
to AI tool, if that makes sense.
If you weren't tasked here, if we
hadn't said, hey, Caitlin, help us
do an experiment of writing this with
AI, would you, was the notebook part
where it, like, summarized all of
this past kind of content from Grow
and Convert into, like, an outline?
Like, was that even helpful or because
you are a growing convert, you know,
team member and you know this, you
know, information in your own head by,
by hand about what category keywords
are, how we use them, blah, blah,
blah, because you do this David,
would you have just done it yourself
and you created the outline yourself?
Yeah.
I think you hit it on the head.
There is that since I'm in this every
day, I know this one well enough that.
I would have just done it myself.
I can see it being super helpful for Um
onboarding a client or working with a
new account or something like that where
I don't know Anything about the industry
or the product because there's a lot of
information to keep straight there um But
yeah with grow and convert it's especially
this topic Um is one that i'm pretty
familiar with that's actually really
interesting um benji is like is there
a tool that's just like I don't know,
for marketers or for anyone, where it's
like, just input a ton of the companies.
Information.
And it's just like a chat tool that,
that you can just like query it.
And it, it, the tool basically would
be like, I'm the company expert on
the product or the brand or something.
I'm sure this exists.
Yeah.
I don't know.
But it made me think like, even for
like our onboarding call today with a
new client, put, put, putting that into
notebook and just getting a summary of
all the key points about the client.
I was thinking the exact same thing.
Cause we had that call this
morning is like, it just, you
can just create this and then.
You know, that content strategist
would have, you can just
start querying this thing.
Anyway, that's an interesting aside.
So, I've actually done like a
manual version of that for a few
clients, where I'll have what I
call like a master list or a cliff
notes version of, um, Everything
that we've learned from the client.
And it'll be like, you know, these
are our main audiences that we usually
write for here at each of their
products and how we usually talk
about them, the different, you know,
sometimes we talk about it like this.
Sometimes we talk about it like that.
Yeah.
So.
Yeah, I've seen those
like what you're saying.
You've just created a Google
Doc where you've just started
summarizing and being like, this
is how we talk about everything.
Yeah, and AI would be a good use case.
Okay, so let's see what Claude
wrote after all of this work
you've done with multiple tools.
And then, and then some examples
of kind of what you edited.
Yeah, like, let's start with this intro.
And so notably here, like even this edit
that I'm replacing some of this text with.
It is still written by AI, it was
just I took this sentence and asked
it to rewrite it in one of the tools,
most likely Clod, or expand on it in
some way, and then I liked what it
gave better on one of those tools.
Variations or iterations.
Yeah, that to me still sounds
like some artifact of, of it,
this being kind of like in, Let's
try to write with AI experiment.
That probably like, if you were
doing this naturally and you were
just like, I just want to save
some time and have AI help me.
If it, if it gave you an intro sentence
and you were like, Ah, I kind of
want to say it in a different way.
You would just rewrite it, right?
You wouldn't then go back and
be like, actually do this.
I don't know, maybe you would.
No, no, definitely not.
There was a lot more.
Reprompting that I was doing because
I was trying to leave as much of it in
pure AI as I could and Essentially here.
I was just choosing the best of
the AI that I was able to prompt.
Okay, and then Feel better because there
are some parts that we'll talk about in
a second where I said, oh finally there's
some like good Caitlyn level writing like
look to jump to the end then after You did
this and you created this draft, which you
both edited yourself and kind of like used
AI to edit as you're saying now, which is
like, you don't like the sentence because
we were doing this kind of AI experiment
instead of just doing what you naturally
would have done, which is rewrite it.
You then asked Claude to rewrite it
several times and you took the best
version and then you sent it to me.
I'll say my initial reaction was
like, if my first, first reaction
was like, Oh, this is pretty good.
And then as I started
editing and reading it.
Me knowing that Caitlyn normally is
like a great writer, I started to get
annoyed where I was just like, Man, I
gotta like edit like all this stuff.
And if Caitlyn had just written the
piece, I wouldn't have to edit all this.
So I felt like a noticeable, a
very noticeable decrease, but what
Caitlin's kind of smiling about
is in some parts, I was like, Ooh,
finally, like a Caitlin sentence.
It's much better.
And she was like, actually, I wrote that.
But what I feel better about is
what you're saying right now, which
is you're saying, Hey, I wrote it,
but it's because Much of this, you
asked it to write over and over again
because you were doing this experiment
and you just took its best response.
So that to me is like half
AI written, half Caitlin
massaging AI to write properly.
Yes, definitely.
Um, and if we get into some specifics
here, like as I was writing the article,
I did start at the top and I went down
through the sections, um, using AI.
So towards the end, I was getting
better at using AI and I was learning.
Like, when it was useful to use Notebook
versus Clod, and how to ask it for what
I wanted, and just how to use the tools.
. And so it was getting a lot easier
and I was getting better at using ai.
And that's where, let's see, where is it?
You'd left a comment where I said,
oh, this is, this is refreshing
to see Kaitlyn writing or Yeah.
You had said, I don't know if
what it is, but this seems better.
Yeah.
So you said, I'm curious now to look
at your videos in the other doc and see
how much AI was used in this section.
It could be that I'm mentally
exhausted by this point, but
it feels better than the rest.
Hmm.
I actually didn't know this.
We hadn't, or maybe you had told me
and I'd forgotten, but I feel like I'm
learning this for the first time now.
That comment was on October 31st.
Ooh, I left that comment on Halloween.
But I'm learning this now on January 30th.
So, you're saying the reason I felt
like, oh, it's getting much better was
because you were just getting better
at this whole thing of using AI.
Yes, and so, the other piece that's
interesting is that I was doing Less, like
I was giving it smaller chunks to work on.
So I would prompt something, uh, give it
a prompt in Notebook, and then a prompt
in Clod, I'd go back and forth, and then
I had, like, all of these versions, and I
would essentially, like, take a sentence
here, take half a sentence there, and I
was like, Frankensteining it together,
you know, like just quilting it together.
And so did that save you time
you think no Goodness, no.
Okay, so you just did this because
we this was our like experiment our
project Yeah, it was it was interesting.
I was interested to see if I went through
this work of like just pushing through
and figuring out how to use this new tool
if it could save me time in the future.
I think it's gonna take a lot more.
I almost feel like it's
almost like starting over.
Wait, wait, explain that.
What do you mean?
What is like starting over?
Like it reminded me of being a
new writer at Grow and Convert.
Where I was just like trying to wrap
my head around everything and like
produce what I needed to produce and
Understand all of the edits I was
being given and all of that like it's
you know different it's Different
mental muscles that you're using here.
So if we said hey Kaitlyn for Client
work, which I know you Kaitlyn doesn't
do client work Her role now is more
coaching other strategists to do client
work but let's just pretend that you were
doing you were just like a Strategist
doing and you had like three or four
clients At Grow and Convert and we just
said, you know, grow and convert policy.
We're, we're, we're using, you guys, you
guys can just use AI as much or as little
as you need from just this one experiment.
Would you, would you use it, do
you think it would save you time
just left to your own devices to do
growing, convert level client work?
I would like to experiment with
it a little bit further, but it
would be in a much different way
than I was doing with this article.
What way?
I think that would be useful to hear.
So I think the things that I would use
it for are, you know, sometimes I'm
debating between how I want to arrange
the arguments, if that makes sense.
Like, what's the most logical flow?
Do I want to talk about this first?
Or do I want to talk about that first?
Does it, does this argument
make sense in this section?
Or a different section?
And so, just feeding the tools
an outline with, like, both.
Flows of logic, or, flow, uh.
Sets of arguments, if that makes
sense, um, and seeing when it shoots
out helps me make that decision.
As to, like, how do I want to arrange
these arguments, because even if it's not
written well, I can kind of get a feel
for, oh, I like the flow of this better.
And then I dis would disregard
everything like it had written,
and I would go and write it myself.
But sometimes just having that,
like, bouncing board, I guess, for
how do I want to arrange this, it
can be interesting in that way.
You wouldn't get that feeling
from just the outline itself,
like seeing the outline.
You're saying turning the outline into
written prose gives you more clarity
of how you should structure a piece.
Even if that written prose is so
bad, you would rewrite it anyways.
Sometimes.
Yeah.
And not all the time.
That's really interesting.
Sometimes.
Yeah.
And I saw that in the survey
responses, too, of different writers.
One thing I learned from the survey
responses is I used to think of it kind
of like Benji, how he intro'd this episode
of just like, we wrote an article with AI.
I used to think of it like
that, like, is this AI written?
Did AI write this?
And I realized that it's like,
it's not so cut and dry like that.
It's like people were just like using
A. I. in different ways in their
process, and it was like hard to say
whether it was something was quote
unquote A. I. written some use it
for research, like we said, like, I
could see that being really helpful.
Like, imagine doing a competitor
post and you're using it to just pull
information about your competitors.
Like who are the top 10
competitors in the space?
And I actually have done that.
Yeah.
And historically, you
would have had to go.
Google those people, find out which
ones are the most common ones.
So like that, that is how I could see
this being really helpful is to help
with specific pieces of the process.
Whether it's to help you do the outline
in the very beginning by giving it sources
and then coming up with an outline.
Or if you need a list of different
companies in the space and you just
want to use AI to do that portion
of it, that could be really helpful.
I can see like almost like workflows
for different types of blog posts.
Like, whether it's a list post
or something like that, there's
different ways that you can use
AI and parts of the process.
It doesn't necessarily mean you need
to write the entire blog post with AI.
Or any of it.
Yeah.
True.
The other thing I use it for
is to like, polish sentences.
Honestly, I still don't even
use what it suggests to me.
I'll, you know, take a sentence where I'm
like, Uh, this sounds a little bit funny.
It's not quite how I want to word it.
But, you know, I can't think of
another way to say it right now.
And so I'll feed it to it.
give it a little bit of a prompt about,
like, the style I'm looking for, and
ask it to spit out half a dozen options.
And usually, again, it's this
Frankenstein version where I'm taking
half of this sentence, half of that
sentence, a word here, a word there,
and building an entirely new sentence.
So it can be helpful in that way,
to just kind of break you out
of the usual Way that you write
something if that makes sense.
Yeah, I I've used I've used it for that
kind of copywriting help I think Chachi
Bouti and Claude they're great thesauruses
in that respect You're like I'm trying to
say this thing the words not coming to me
or like what are like five Options or ten
options for this headline and it'll just
do it instantly and and I agree completely
with what you're saying is I often don't
use any of them verbatim, but it says
something that I'm like, Oh, yeah, it was
something like that that I was kind of
searching for in my head and the research
gives you inspiration and ideas, right?
You can go off of the research one.
You said Benji is interesting in
that what's useful about these chat
tools is it's like Google plus.
Google, you like search something
and then you get like, what other,
you know, pages on the web have
said, and then you have to click into
each page and you need to read it.
And sometimes those pages
just have a bunch of nonsense
that you don't need anyways.
Right.
Um, and like the famous one that everyone
complains about is like recipe blogs.
Everyone just wants the recipe, but
there's always that stuff above the recipe
that everyone has to scroll through.
Right.
And then like, whatever, and that's
Google's fault because Google's
somehow using that to rank.
And, and you can be like, just give
me the recipe and it'll do that.
That's the same thing with
the competitor research.
It's like a research tool plus
you can kind of chat with it.
And it's really, it's like
smart in that respect.
I would love to show a few
examples of kind of writing.
And what I edited when I received it.
And what I started to kind of annoy me.
And I think this will be a good
lesson to maybe close this with.
In that, I think some of these
are lessons for human writers.
In the sense that we see some of
these, I'm just going to call them
mistakes or issues, just like not
great writing in writers that apply,
in new writers that we work with.
And I'm just going to create a blanket
name for all of this, and I'm going
to call it kind of filler language.
Like it's just, it's, it's language
that it sounds, the English is perfect.
There's nothing
grammatically wrong with it.
And then it is beyond that.
Just like we said, mirage content
sounds like it should be good.
This is kind of like the micro mirage.
It's like mirage writing.
It's like at the writing, at the
sentence level, you're like, yeah,
yeah, it sounds like it makes sense.
And then you're like, what
does that actually say?
And you're like, I don't, I don't know.
Um, so for example, um, it, it, we're
in this section about category keywords
that, and, and, and Caitlin, you
have this section talking about like.
Um, synonyms, so if you're selling
accounting software, best accounting
software is kind of an obvious, um, uh,
category keyword, but there may be just
synonyms that Google kind of ranks the
same pieces for because they're the same,
like, and I can't think of good ones.
So example would be best accounting
tools will probably have the same
results as best accounting software.
Don't hold me to that, but
like, that's the idea, right?
And, um.
It, it, there's a couple of sentences
that I was like, what the heck?
So at the end, after explaining this,
it said, it says by targeting a variety
of synonyms, you can capture a wider
audience and increase your chances
of ranking higher in search results.
And then it continued.
I don't know if it again, Caitlin,
maybe this is you, but I just,
you don't write like this.
And then it'd be like, while
targeting synonyms can increase
conversions, it's important to note
that you can't always target all the
variations with one piece of content.
Blah, blah, blah.
But let's look at that first sentence.
By targeting a variety of synonyms,
you can capture a wider audience
and increase your chances of
ranking higher in search results.
Sounds fine, but I was
like, that's not true.
By targeting a variety of synonyms,
you don't capture a wider audience.
We're talking about synonyms.
Like, literally what we're saying is just
remember that there's multiple phrasings.
That mean the same thing, Google's
smart enough to know it and it's
ranking the same pieces for that.
So it's just basically
like the same keyword.
You need to think of
them as the same keyword.
So you, it doesn't capture a wider
audience and then to say, and increase
your chances of ranking higher in search
results is just, it's not true and it's
very odd phrasing that like, it doesn't
increase your chances of ranking in
search results, higher in search results,
we're talking about different keywords.
It increases your chances of
ranking for different variants.
Maybe, but even that you wouldn't
really say, and that's just
something to watch out for.
I have seen this in human writing
of people who apply, but it's,
it's very common in AI writing.
Um, and that's this micro example
over and over and over again.
In my experiment of trying to
write the analysis piece of this AI
writing survey of 50 plus marketers
who are using AI assistants.
I just gave up because that
is a very specific piece.
It's not an SEO piece.
It's a, it's like a traditional grow
and convert brand data analysis piece.
We have strong, very specific arguments
and like the takeaways were really
interesting and I just couldn't at
this, at some point, I just couldn't
take letting AI kind of do this kind
of kind of nonsense writing on it.
Um, and I gave up because it was
like a death by a thousand cuts.
You read it.
And you're like, yeah, yeah, yeah,
this is, this is a good point.
And then after a while you're like,
well, that doesn't quite make sense.
You need to pay, if you want to really
be like producing good writing, whether
you're doing it or AI is doing it, you
need to be very careful to stop when
you notice things like that and fix it.
And I think that summarizes pretty
well my, I have a reluctance to
use AI because like, I, I submitted
this article to you, right?
So like I, I worked with AI a lot.
I edited it myself and then I sent
it to you and then I got it back
from you with all of your edits.
And I read through it again
with fresh eyes and I was like,
this is embarrassing, David.
This isn't me.
Please.
Like, you know what I mean?
See, so you . Yeah.
But like when I was writing, like going,
writing it with ai, a lot of that I missed
because it was like so much better than
what AI had originally given me that
I was like, oh great, this is amazing.
And then I come back with fresh
eyes and I'm like, no, it's not.
So it's like.
I can't trust it.
You know what I mean?
And I feel like it It dumbs me down
when I'm using it too extensively.
Oh, like it affected your taste?
Yeah.
Yeah.
That, because you kept reading it,
you start to be like, this is good.
Like this is good.
And then you, and then you get
like human writing and then, and
then you're like, wait, actually
this isn't, this isn't good.
Yeah.
You, it, it's because you're, you're, when
you read it, it, it sounds good and you're
not, you haven't thought through all the
arguments that you wanna say yourself.
And so when you see it on paper,
you're like, oh, this is pretty good.
But then when you think about
how you would say it to someone
else, that's when it changes.
It is somewhat deceiving.
I've noticed that myself, was when
I ask it for something, at first
glance it always looks pretty good.
Because I haven't really thought
through myself how I would say it.
And what I've noticed is, when I write,
that's my way of clarifying my thoughts.
And so, seeing someone else's
writing before you've had a chance
to clarify your thoughts, It seems
a lot better than it actually is.
I wanna, instead of ending kind of
trashing AI writing a little bit like
we're doing right now, which is, which
is an important side of the discussion,
the one optimistic side that I'll say
is there was a response in the survey,
and we highlighted this I think in our
first video this year, where we said
Um, we're kind of exploring AI from
different angles from the search angle.
We talked about with you on the
last video, Caitlin and now on the
writing side, um, there was one
respondent that said a lot of clients
don't care about the writing at this
level, they can't see the difference.
And they don't, and that's not really
what they're hiring us or me for.
They just want this piece on this.
And so I had a lot of empathy for that.
And I thought we have seen this kind of
what we've called we, like we wrote Mirage
content, Benji, like seven years ago.
So we were complaining about blog
content being a mirage where way
before ChatGBT was a thing, right?
And we, there was the same idea that we
called it mirage content because you first
read it and you're like, this sounds good.
And then when you stop and think
about what is it actually saying
argumentation wise, you're like,
this isn't saying anything.
Yeah, it's a function of like
things, things were ranking
back then this, this was like.
When you could rank with, uh, the
roundup post where you get a bunch
of people's opinions and compile them
in a piece, like that kind of stuff.
And like the title looks good and you
like read it and you kind of have some
understanding of the topic and you're
like, Oh, this is decent, but you
actually get into the argumentation or
like the key points and you're like,
this actually doesn't make any sense.
And we meant it, I think, at a higher
level, like we were saying, like,
a lot of it is just kind of obvious
information and this and that.
This AI thing is mirage y, that's
why I'm saying it's almost not
mirage content, it's mirage writing.
It's at the granular level
that sentences themselves.
You're like, wait a minute,
like, it's just not true even.
Like, you, if you're ranking for
synonyms, it doesn't help you reach
a wider audience or whatever that
second half of that sentence was.
It doesn't, it doesn't,
it doesn't make sense.
But, we have a certain
standard at Grow and Convert.
Like, like I said at the beginning,
I'm very aware that not everyone does.
And so you can use AI in different
capacities and maybe it helps
you consolidate your thoughts.
Maybe it helps you do research
and consolidate that research.
And it saves time versus opening
10 million Google results tabs.
Maybe, um, it, you can give it a really
detailed outline where you're like,
do not mention any, I've done this.
I've tried this and it's like
better, you know, it still
annoys me, but it's still better.
It's like, don't make a single argument.
That's not in my outline.
Like just turn this
into complete sentences.
Like it's these AI tools
are very good at English.
You know what I mean?
Like writing.
And so there's different ways to use it.
And, and.
And content writing, there's like
a bajillion people in the world.
They're all doing it for different things.
So there's, there's ways where
you can kind of make it work.
But, um, at the highest levels where
the ver of marketing writing where
the phrasing and the positioning
and the messaging about your product
and its features and exactly how
the feature is differentiated
versus status quo or competitors,
when that really, really matters.
It, it, it breaks down that
like human writing is important.
And even within humans, we see clients
debate about exactly how to phrase that.
And so that that's, we debate it too.
Like in, when we're doing grow
and convert posts, it's, there's a
lot of back and forth in terms of
how we want to say something and
how we want to position something.
And even with this project itself, that
was why we wanted to deal with grow,
grow and convert piece first, because
to us, that's the, like the highest.
The hardest thing that we're going to
be able to produce if we're okay to
publish AI on grow and convert, then
it's likely we can do it for clients
too, because the amount of editing
that we have and the amount of back and
forth is much greater than any client.
Yeah.
So like, even if we're being a little
bit snooty about it, that's fine.
And if you're like, well, my clients
don't care, I'm going to use it.
I think like.
Go, go for it.
Like, if it, if it's helping
you out, like, go for it.
Um, but I will say, like, Caitlin,
I feel kind of what you were getting
at, but maybe not saying directly,
is I enjoy writing like I think a lot
of people also are annoyed by it and
so you see like people that are not
freelance writers using AI for writing
assistance because they're, they have
different positions and they're just like
doing writing and it just is stressful
for them or they are annoyed by it.
That's fine.
But for me, like.
There's like a craft, there's like
a, uh, like a fun, a struggling fun
of trying to figure out the puzzle of
like I want to say it in this way and
what are different phrasings and, and,
and it definitely takes from that.
But, this was super interesting,
thanks, thanks for sharing all this.
If you like this video,
don't forget to subscribe.
You can also get the audio
only versions of these shows
wherever you get your podcasts.
And, you can follow us at growandconvert.
com slash newsletter for any articles and
updates for when these videos come out.