March 10, 2025 • 27 mins
Unlock the transformative power of Large Language Models (LLMs) and witness their impact on public relations like never before. Peter Woolfolk talks with Frank Strong, the insightful founder of The Sword and The Script, to explore how these cutting-edge technologies are reshaping the PR landscape and altering our approach to information retrieval. We'll guide you through the evolution from traditional news outlets to Google searches, and now to the conversational prowess of LLMs. With practical examples, we reveal the benefits and the hurdles, providing an engaging and comprehensive look at the future of PR.
Discover the pivotal role LLMs are playing in reputation management, crisis communication, and consumer research—especially in sectors like automotive sales. Frank shares expert insights on the applications of AI in media monitoring, content brainstorming, and social media strategies. Yet, amidst this digital revolution, we underscore the irreplaceable touch of human creativity and judgment. Tune in to grasp the exciting potential of generative AI and how it's poised to redefine public relations, while always valuing the unique contributions only we humans can make.
Information on NEW podcast website.
Real Talk About MarketingAn Acxiom podcast where we discuss marketing made better, bringing you real...
Listen on: Apple Podcasts Spotify
Newsletter link:
https://www.publicrelationsreviewpodcast.com
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Announcer (00:03):
Welcome.
This is the Public RelationsReview Podcast, a program to
discuss the many facets ofpublic relations with seasoned
professionals, educators,authors and others.
Now here is your host, peterWoolfolk.
Peter Woolfolk (00:24):
Welcome to the Public Relations Review Podcast
and to our listeners all acrossAmerica and around the world.
Apple has ranked this podcastamong the top 1% of podcasts
worldwide, so thank you to allof our guests and listeners for
being the basis for this ranking.
Now, if you enjoy the podcast,we certainly would like to have
a review from you, and we'd liketo ask you to share this with
(00:46):
your friends.
Now question is Google stillthe leading search platform for
public relations practitioners?
Can anyone overtake Google.
Well, the new platform on theblock is called LLMs Large
Language Models and my guesttoday will expand on LLMs and
why they are more impactful.
(01:07):
So joining me today fromAtlanta, Georgia, is Frank
Strong.
Frank is the founder andpresident of the sword and the
script, a veteran-owned businessfocused on public relations,
content marketing and socialmarketing for the b2b
marketplace.
He has implemented programs andled teams for public relations,
content marketing and socialmedia on the in-house side of
(01:29):
the table with corporationsranging from startups to the
mid-market to globalorganizations.
He has also endured the rigorsof billable hours, having grown
up on the agency side, with gigsat PR firms small, large,
including the top global firm,hill Knowlton, now HK Strategies
(01:49):
.
So, Frank, welcome again to thepodcast.
Now give us an overview of LLMsand the major impact it is
having on public relations andsearching.
Frank Strong (02:00):
Hi Peter, thanks for having me back.
It's good to see you again.
Yeah, llm stands for LargeLanguage Model.
It is the model that driveskind of generative AI interface.
So if you go to a platform likeChatGPT, perplexity, google's
Gemini Claude, any of thoseplatforms, they're all being
(02:20):
driven by an LLM.
And, without getting into allthe ins and outs, an easy way to
think about an LLM is it's kindof like a very sophisticated
word suggest.
So if you go to a traditionalGoogle search engine and you
start typing in a search term,it's trying to guess what you're
going to ask.
You know you say red cars.
(02:40):
You could say red cars inKansas, red cars in Brazil.
You know fast red cars.
So it's going to give you allof these suggestions and so LLMs
kind of work the same way.
Peter Woolfolk (02:51):
Well, now that we've heard that sort of basic
explanation of it, what is itactually doing?
I mean, as people are using it,what are they getting?
What are they seeing that iscausing this rush over to LLMs?
Frank Strong (03:05):
Yeah, I think it would be good.
I'll just take a quick stepback and explain kind of the old
newspaper model and how thatwas displaced by Google and how
that's being displaced by LLMstoday, and the reason for that
is, you know, it's conceivable.
There are people today thatdidn't grow up seeing stacks of
newspapers by a vendor and so atraditional newspaper you wanted
(03:26):
to be what was called above thefold.
If you hold a newspaperstraight up and you fold it in
half, they would stack thenewspapers together.
It would look like a bale ofhay and that's how they shipped
them around.
That's how it was displayed onthe newsstands, and so the above
the fold front page of anewspaper was the most visible
place that you could possibly beFor years and years.
That's how people got most oftheir information until Google
(03:48):
came along.
Then it got a whole lot easier.
Instead of sorting through anewspaper when you're looking
for information or newspapers itcould be many or going to the
library and sorting throughmicrofilm of old newspapers, you
could just go to Google, ask aquestion and it would give a
list of sources, right, and youcould choose which source you
want to look at.
You could review which onesanswered your question best.
(04:09):
Those queries that answeredyour question best tended to
rank higher over time in thesearch engine.
That's all being changed byllms and the reason for that is
llms are able to presentinformation as just an answer.
It it's almost like you'retalking to a person and they're
typing up a response to yourquestion.
(04:30):
You ask it a question and itgives you like a primer.
That comes with some benefitsand drawbacks too.
It should be noted, like whenyou think.
For example, when you went anddid a traditional Google search,
it gave you however manyresults it could be 20, 30, 100.
And you got to choose whatinformation you thought best
answered your question.
Llms present it as a briefingand it presents it from a
(04:54):
position of authority.
Like here is the answer to yourquestion and that can have some
drawbacks, mainly because A it'sproviding an answer and that
may or may not be the correctone or the one that works best
for you.
And two, we know for a factthat generative AI programs do
hallucinate.
That means they just simplymake up the answer and they're
(05:17):
doing that because they'retrying to.
It's basically like an averageof words, so you have to be
careful what you believe.
Peter Woolfolk (05:23):
You know I have.
Once I read your article, Ilooked into some of the other
LLMs and I jumped onto aperplexity and I asked a very
basically a question such as youknow, how many listeners does
the Public Relations Reviewpodcast have?
And while it was very, veryquick with that coming back, I
was shocked at how muchinformation came back, and it
(05:45):
came back in a format as thoughwe were having had had a
conversation about this.
So I was, you know, pleasedabout it.
It did not give me a specificnumber, as you said before it.
Let you know, we went here andhere's what we're here, here's
what we got there and theseother sort of places.
So I was not only was I pleased, I was was just surprised at
the format that it came back in,and I think this might be
(06:07):
helpful to a lot of people whomight consider using these.
Frank Strong (06:11):
Yeah, I completely agree.
It's incredibly amazing and Ithink your example brings up a
good question, because you'vehad this podcast for a long time
.
So it's been indexed and theselarge language models have been
able to be trained on all kindsof content and obviously,
including your podcast.
That's how it's able to returnthe answer, and so the question
(06:32):
that it conjures is if LLMs aregoing to be the new front page,
and there's some evidencesuggesting that this is
happening.
Google's search traffic hasdropped for the first time since
2015.
And it was a small amount, butthere is data that suggests
referral traffic from LLMsPerplexity, chatgpt, claude
Gemini, google Gemini that'sgrown, so we know that this is
(06:56):
in transition.
So the question that PRprofessionals should be asking
themselves is how do we get intoan LLM?
You've gotten your podcast intoan LLM and that is one approach
, right.
You have to have content, and Ithink a lot of the answers to
the question of how do I pitchan LLM comes back to the
fundamentals that PR peoplealready know.
It's having a strong point ofview, it's being able to
(07:19):
articulate a story so that itgets picked up, it gets a
visibility and indexed.
It's answering specificquestions right.
When people go to a traditionalsearch engine and they type in a
search query, that is, bydefinition, an expression of
need I'm looking for information.
So answering specific questionsis a great way to get crawled
(07:40):
and indexed by an LLM to providethat response great way to get
crawled and indexed by an LLM toprovide that response.
And then, last but not least,the traditional aspects matter.
So traditional media is goingto be deemed highly credible by
LLMs and so it's going to usethat to train the data and
return answers.
All of the traditional thingsthat PR people did.
Press releases, for example.
(08:01):
There seems to be some evidencesuggesting that press releases
are being used, and that makes alot of sense to me, because
when you think about pressreleases relative to the writing
on the rest of the web and wecan make fun of you know press
releases are promotional andpeople do funky things with them
, but the writing is probably ahigher quality relative to the
rest of the web, and so it's areally good data set on which to
(08:25):
train these models, and I thinkthat's why it's happening.
If there's a caveat, it would bethat at some point this
happened with Google.
Press releases were a good wayto get into search engines.
Google started deprecating thelink value because they don't
want you to be able to buy apress release to get into their
search engine.
That's kind of gets spammy.
I suspect something like thatwill happen with LLMs at some
(08:46):
point.
Peter Woolfolk (08:47):
Where else do you see LLMs fitting in?
Because you know some of thethings we've talked about here
and I think I may have mentionedearlier that some organizations
where consumers could call intothe company and ask questions
about the company or companyproducts or those kinds of
things Would this be anappropriate use for LLMs?
Frank Strong (09:08):
Absolutely, and I think there's two ways to look
at that.
One would be to ask a question.
Just like you asked thequestion of perplexity about
your podcast, you can go to anypublic generative AI tool and
ask a question about a companyand it's going to return results
.
And that's the kind of mediarelations aspects that PR people
need to be worried about is howare we being presented in these
(09:29):
LLMs?
That's one use case.
The other use case along thoselines is companies can purchase
technologies, these generativeAI tools, and train them on
their internal documentation.
If you have a product, forexample, a technical manual,
that can consume the wholetechnical manual, probably do it
better than a human, and thenit's trained on the answers that
(09:52):
customers may have about yourproduct.
So come to a chat bot on yourwebsite and they ask a specific
question and that AI has beentrained on your product
documentation.
It can return a very goodanswer, which is a far it sounds
like a far better approach thansaying here's a bunch of help
desk articles.
Go ahead and search throughthese and see if you can find
your answer and then, if youcan't, submit a support ticket
(10:15):
and we'll get back to youwhenever we can and we'll
schedule a call and go throughyour issue, right.
So it's just improving theexperience, I think, from a
customer service perspective,and that's great use case,
that's that's being entered nowwell you know, I certainly see
that as a viable function.
Peter Woolfolk (10:29):
I'm sure there might be some companies using
that already.
As a matter of fact, I haven'tsee where they've got.
Some companies have a on videosthat can respond to consumer
questions, and so I'm sure thatI Other organizations are going
to use this as a PR tool because, yes, we can answer your
question right away.
You don't have to wait for areal person to show up.
(10:49):
You know, as soon as you callus, we can get back to you, or
we can answer you right awayeven with someone talking to you
, and in some cases, I wouldn'tbe surprised if they can take a
real human face and animate itso that it looks like it's
actually talking.
In some cases, I think that isbeing done.
Frank Strong (11:07):
Oh yeah, that's definitely happening.
Some of the visuals are reallycompelling.
I mean, they're able to makeyou know fully-fledged movies
based on some generative AIprompts.
It's actually crazy.
One of the ways that you cannotice, though, is that, for
whatever reason, the same waythat generative AI tends to
hallucinate with text-basedanswers, it makes some wonky
errors with visuals.
(11:27):
Like, if you ask it to create apicture and you look at the
hands it tends to have likeseven fingers on it.
It'll have like weird thingslike that.
So these are like telltalesigns and videos and images that
you're able to spot some ofthis.
Peter Woolfolk (11:40):
Well, what are some of the things, some of the
other things that you see in thefuture for public relations
with these LLMs?
Frank Strong (11:48):
Well, I think we're in the middle of a sea
change right now, like now isthe moment right If you were
around when the Google searchengine started becoming the
dominant source, the firstsource of place that people went
to for information.
We're seeing thattransformation happening now.
Google, the traditional searchengine, is very worried about
this by the way.
That's why they're investing inGoogle Gemini, and you may see
(12:10):
what they call AI artificialintelligence overviews in search
.
If you go to Google right nowand you type in a search, it's
going to give you an AI overview.
So they're working on this andthey're playing catch-up chat.
Gpt cause them by surprise, butI think this is an area that is
emerging rapidly.
We can see that it is takingover as a primary source because
(12:30):
it provides a very neat, clean,simple answer when people ask
questions, and so we got to beup on it.
It means we're going to have tomonitor our reputation there.
How are these LLMs presentingthe organizations that we
represent, whether it's anemployer or a client?
How do we correct things ifthey go awry?
Right, what happens whenthere's crisis communication?
If it's getting something wrongand it sends off a tizzy, how
(12:52):
are we going to go about fixingthat?
And then, generally, how do weget into the LLMs, like we've
described already in the podcast?
So this is emerging and it'sgoing to unfold over the next 12
to 18 months.
We're going to see an awful lotof change.
Peter Woolfolk (13:08):
And I also see it as a sales tool as well,
because obviously you've gotcars, tv sets or whatever else
you might be selling.
That probably is a lot easierfor in some cases LLMs and
whatever sort of visuals thatthey have with that can save a
lot of time and answer a lot ofquestions quickly to consumers.
Frank Strong (13:29):
Yeah, I completely agree.
I think this is an extension ofa trend that we've seen already
and that is in the old days,you know.
Let's say you had to buy a carIn the old days, you didn't know
much about cars.
You went to the car dealership.
The salesperson knew everythingabout the car.
Then, when the internet got big, you could do all kinds of
research and not from justofficial review companies, but
(13:51):
you could see individual userswould write reviews.
You could get a whole lot moreinformation and walk into a
dealership prepared, knowingwhat you want, what the specs
are, how it compares.
I think we're going to seesomething like that become even
more refined, more repolishedand better answers from
generative AI going forward.
That's going to be a lot easier.
(14:14):
Instead of spending hours andhours doing searches and trying
to pull information together,it's basically presenting you
with a dossier.
That's an answer to whateverquestion that you ask, which is
pretty incredible.
Again the drawbacks are you'vegot to make sure it's accurate.
So you mentioned perplexity Inmy observation.
They were one of the first LLMsto provide source links, so it
will tell you where it got theinformation.
(14:35):
You can click the link and gocheck out the source and see.
Is this credible?
Do?
I believe that, and all of the,as far as I can tell, all of
the other LLMs have sincefollowed suit, because you know,
we know, working in publicrelations, having you know
credible sources, citing yourdata, is important, and so we're
starting to see the LLMs dothat.
There's also one other emergingproperty that LLMs are the
(14:58):
companies that make them, areworking on to combat the
hallucination, right.
So, for whatever reason theydon't really understand why, but
you can ask a generative AItool a question and if it
doesn't know the answers, oncein a while this happens less
more often than not it gets itright, but once in a while it
doesn't know the answer and itjust makes one up.
It just gives you an answerthat's completely fictional and
(15:21):
it presents it with suchconfidence.
You're like you know it'd be,like somebody like, well, here's
the answer, I'm absolutelyconfident it is right.
And you don't know.
They don't really know whythat's happening, but one of the
things they're working on todevelop is this new development
called reasoning, and whatreasoning is doing is it's
showing you.
It's showing you the steps ofhow the LLM got to the answer
(15:43):
that it is, and I think all ofthese developments will make
LLMs more credible, moreconclusive.
They're going to get betterover time and deliver better
answers.
So it's something that the PRcommunity really needs to pay
attention to.
Peter Woolfolk (15:54):
You know, I would think so and I just keep
thinking of some other areasthat it can be used in.
I mean public relations, sales,you know, just general
information.
You know, at a conference ofsome kind, a lot of different
ways that or students can callup and or interact with it to
get responses to particularquestions, have to do classroom
(16:16):
materials and so forth, and itcomes through and I guess they
could add visuals to this thingas we move along yeah,
absolutely mean there's a coupleof very definitive applications
in public relations that arealready happening.
Frank Strong (16:29):
If you're using most of the big PR vendors, if
you're using MuckRack, Cisionjust added it.
Meltwater has got a huge AIteam.
I mean they've got scientiststhey've gone out and done
acquisitions just to hirescientists with PhDs to help
them work on this stuff.
All of the big ones have that,so they can do things like if
(16:49):
you're using media monitoring,it can better categorize, you
know placements and put tags andcategories of it.
Is it a story?
Is it a mention?
Is it a backlink?
Pr?
People are using it forbrainstorming, right Like you
just need to.
You know, sometimes looking ata blank page is the biggest
hurdle to getting started whenyou're doing a writing project.
Well, get into a generative AIsystem.
(17:10):
Ask it a bunch of questions.
Announcer (17:11):
It's going to give you a whole bunch of ideas.
Frank Strong (17:14):
Summarization Exceptional at summarization,
like here's a big report.
I don't have time to read thiswhole thing, but, you know, read
it for me and give me the fivekey points I need to understand.
Take me and give me the fivekey points I need to understand.
Take my blog post or pressrelease and give me 10 tweets
that I can then go ahead andschedule, or Facebook posts or
LinkedIn posts like a lot of usecases.
There's a few vendors that aredoing some interesting things
(17:36):
where it has prompts that willhelp you write a pitch or help
you write a press release.
I'm not big into that.
I really think.
I think it's good forbrainstorming, but I'd rather
have a human write.
I think humans write betterthan generative AI At least PR
people do.
But one of the interestingthings that one of these tools
does is it will look at thewords that you are using as
(17:57):
you're writing your pitch andthen it will go out and review
articles that are currentlypublished by reporters and come
back as you're writing yourpitch and saying you should
pitch these people becausethey've written these topics,
and here's the link to that.
That's a pretty coolapplication of generative AI.
Peter Woolfolk (18:12):
You know, one of the things that I have done
with things such as chat, gpt,for instance, that I might lay
out some things that I'd likefor you to write for me on this.
Then, once it comes out, then Iwill take it and make it my own
, in other words, put in thewords or other things, and that
makes it as if it was comingfrom me.
But it has saved me a lot oftime and you know and I'm just
(18:37):
tweaking it so that I'mcomfortable with what it says
rather than just ripping andrunning with it that could cause
some problems down the road.
Frank Strong (18:41):
Yeah, no, a hundred percent.
I mean, I think what you saidis really important, that if
you're using these tools, youknow, think about it as a first
draft and then edit it and makeit your own.
The problem is is I mean,generative AI is basically a.
These LLMs are giant models ofwords, and it understands how
(19:03):
people tend to put those wordsin a certain order.
Based on probability, it isprobable that a sentence is
going to read a certain way andhave a certain structure, and so
what that means is the answersthat generative AI is giving you
is, by definition, an average.
Here is the average languagethat most people use, and we
know in PR, we don't want to beaverage, we want to be different
(19:23):
.
We want to explain to peoplewhat's different about our
products what's better right.
So it's really important tomake sure you have that human
touch.
But I think your application ofthe tool is a good one.
I do that an awful lot where,if I don't like the way a
sentence is written, I'll ask it.
You know, rewrite this sentence, give me a better structure.
Is there an easier way to writeit?
Or another application is?
(19:45):
I don't really like thisheadline.
I'm using the sub headline here.
Can you give me fivealternatives?
Peter Woolfolk (19:51):
And it'll.
Frank Strong (19:51):
it'll belt them out and I can choose to use it.
I can rearrange, I can mix andmatch, and that's helping me to
make my content better.
Peter Woolfolk (19:58):
Well, you know, actually with this podcast now
it, it records everything andthen it cranks out not only the
transcript for me, but it alsosuggests titles for each episode
.
Now I'll read those.
Either I like them, or I don'tlike them, or I'll make
alterations to them, but itgives me all of that.
It gives me blurbs for Facebookand other social media posts and
(20:24):
those kind of things.
I read them.
If I like them, fine.
If not, I change them.
But these things are a greattime saver is what I'm really
seeing here.
There's a huge amount of timebeing saved and helping you get
the job done faster and, in somecases, maybe even more
accurately, because it may comeup with some information that
you had overlooked, because, asyou said, these things reach out
(20:46):
to a wide range of places topick up the information and add
it to what it's giving to you.
Frank Strong (20:51):
Yeah, that's right .
I mean, it's almost cliche tosay this at some point, but the
topic comes up over and over ispeople are like are we going to
be replaced by generative AI?
And I don't think we are goingto be replaced as individuals.
Generative ai is very good atspecific tasks, but it can't do
a job.
You can't say go be my prperson for me right, I can't do
that, but jet pr.
(21:12):
People need to pay attention tothis stuff because, while you
won't probably won't be replacedby generative ai, you could be
replaced by somebody that haslearned how to use it
effectively and been able to bemore productive.
That's a real risk.
Peter Woolfolk (21:25):
You know the other thing, when I look at a
lot of times about publicrelations, public relations a
lot of times is aproblem-solving exercise, and if
the platform doesn't know whatthe problems are, it can't offer
you any solutions to them.
Frank Strong (21:42):
So you know questions like how do we fix ABC
and D?
That's right, you can use it tobrainstorm possible solutions.
I completely identify with theproblem solving because I'm
obviously a consultant now, butI spent 10 years on the in-house
side and anytime somebody got aquestion that they didn't know
how to answer, they sent it toone of two places Corporate
communications or the legaldepartment.
(22:03):
So you're constantly gettingthese crazy wild requests that
you're like why am I getting?
Peter Woolfolk (22:07):
this.
I don't know what to do.
Frank Strong (22:09):
But if you have an LLM system that's trained on
your internal documentation,it's a great place to start.
Start brainstorming, I mean,you know, here's the problem,
how do I solve this right?
It's just a good way to showhow you can be more efficient
and effective and productiveusing these tools.
Peter Woolfolk (22:23):
Great Well, frank.
You've provided us with anawful lot of information really
on LLMs.
Are there any closing?
Frank Strong (22:39):
remarks you think that we need to cover so that
our listeners can be completelynot completely, but even more up
to speed on these LLMs.
You know, I think the key is toget out there and experiment
with them.
Make it a point to set sometime aside and try to use some
of these tools.
It doesn't matter, really,which one you use.
If you like ChatGPT, go for it.
If you like Perplexity, whichis one of the ones I really like
, go for it.
If you want to use GoogleGemini, go for it.
But take some time toexperiment with brainstorming,
(22:59):
with rewriting headlines, withhaving it giving you blurbs to
post on social media.
Give this stuff a try so thatyou get some experience with it
and you know what thepossibilities are, because the
changes are going to come fastand furious.
The development is happening ata pace that is just
unbelievably quick, and thesetools are getting better and
(23:19):
better.
I mean think about when thisthing, when ChatGPT launched
what two years ago?
Peter Woolfolk (23:23):
When it first started.
Frank Strong (23:25):
It was really a stunning event.
People were like, oh my gosh,amazing.
And it's come so far in such ashort time, absolutely.
I saw this, you know, it waskind of a meme not too long ago
and it showed like a chariotfrom thousands of years ago and
then it showed a horse con drawncarriage in like the 1800s.
It's like, look, this is 2000years, technology hasn't changed
(23:47):
much.
But then when you think aboutthe pace of change from the
right butter brothers launchedan aircraft to now where we're
putting robots- and helicopterson mars or landing these things
on, like we just took a samplefrom a comet in outer space.
We landed a device on a comet inouter space, took a sample and
brought it back to Earth.
Like the pace of change isremarkable.
(24:08):
That is going to happen inthese generative AI applications
tools.
So the pace of change hasincreased.
Peter Woolfolk (24:15):
Well, frank, let me say thank you once again.
You have provided us with somevery, very valuable information.
I've learned something, I'msure our listeners will learn
something, and I'm certainlyglad that you've been a guest on
the Public Relations Review fora second time, and perhaps down
the road, I believe, we'llprobably reach out to you again.
Frank Strong (24:34):
Thank you, I'll be here.
Thanks for having me on, peter.
Peter Woolfolk (24:36):
Okay, and to our listeners, thank you for
listening.
If you've enjoyed the PublicRelations Review Podcast, we'd
certainly like to get a reviewfrom you and also let your
friends know to listen to thenext edition of the Public
Relations Review Podcast.
Announcer (24:52):
This podcast is produced by Communication
Strategies, an award-winningpublic relations and public
affairs firm headquartered inNashville, Tennessee.
Thank you for joining us.
Welcome.
This is the Public RelationsReview Podcast, a program to
discuss the many facets ofpublic relations with seasoned
professionals, educators,authors and others.
Now here is your host, peterWoolfolk.
Peter Woolfolk (00:24):
Welcome to the Public Relations Review Podcast
and to our listeners all acrossAmerica and around the world.
Apple has ranked this podcastamong the top 1% of podcasts
worldwide, so thank you to allof our guests and listeners for
being the basis for this ranking.
Now, if you enjoy the podcast,we certainly would like to have
a review from you, and we'd liketo ask you to share this with
(00:46):
your friends.
Now question is Google stillthe leading search platform for
public relations practitioners?
Can anyone overtake Google.
Well, the new platform on theblock is called LLMs Large
Language Models and my guesttoday will expand on LLMs and
why they are more impactful.
(01:07):
So joining me today fromAtlanta, Georgia, is Frank
Strong.
Frank is the founder andpresident of the sword and the
script, a veteran-owned businessfocused on public relations,
content marketing and socialmarketing for the b2b
marketplace.
He has implemented programs andled teams for public relations,
content marketing and socialmedia on the in-house side of
(01:29):
the table with corporationsranging from startups to the
mid-market to globalorganizations.
He has also endured the rigorsof billable hours, having grown
up on the agency side, with gigsat PR firms small, large,
including the top global firm,hill Knowlton, now HK Strategies
(01:49):
.
So, Frank, welcome again to thepodcast.
Now give us an overview of LLMsand the major impact it is
having on public relations andsearching.
Frank Strong (02:00):
Hi Peter, thanks for having me back.
It's good to see you again.
Yeah, llm stands for LargeLanguage Model.
It is the model that driveskind of generative AI interface.
So if you go to a platform likeChatGPT, perplexity, google's
Gemini Claude, any of thoseplatforms, they're all being
(02:20):
driven by an LLM.
And, without getting into allthe ins and outs, an easy way to
think about an LLM is it's kindof like a very sophisticated
word suggest.
So if you go to a traditionalGoogle search engine and you
start typing in a search term,it's trying to guess what you're
going to ask.
You know you say red cars.
(02:40):
You could say red cars inKansas, red cars in Brazil.
You know fast red cars.
So it's going to give you allof these suggestions and so LLMs
kind of work the same way.
Peter Woolfolk (02:51):
Well, now that we've heard that sort of basic
explanation of it, what is itactually doing?
I mean, as people are using it,what are they getting?
What are they seeing that iscausing this rush over to LLMs?
Frank Strong (03:05):
Yeah, I think it would be good.
I'll just take a quick stepback and explain kind of the old
newspaper model and how thatwas displaced by Google and how
that's being displaced by LLMstoday, and the reason for that
is, you know, it's conceivable.
There are people today thatdidn't grow up seeing stacks of
newspapers by a vendor and so atraditional newspaper you wanted
(03:26):
to be what was called above thefold.
If you hold a newspaperstraight up and you fold it in
half, they would stack thenewspapers together.
It would look like a bale ofhay and that's how they shipped
them around.
That's how it was displayed onthe newsstands, and so the above
the fold front page of anewspaper was the most visible
place that you could possibly beFor years and years.
That's how people got most oftheir information until Google
(03:48):
came along.
Then it got a whole lot easier.
Instead of sorting through anewspaper when you're looking
for information or newspapers itcould be many or going to the
library and sorting throughmicrofilm of old newspapers, you
could just go to Google, ask aquestion and it would give a
list of sources, right, and youcould choose which source you
want to look at.
You could review which onesanswered your question best.
(04:09):
Those queries that answeredyour question best tended to
rank higher over time in thesearch engine.
That's all being changed byllms and the reason for that is
llms are able to presentinformation as just an answer.
It it's almost like you'retalking to a person and they're
typing up a response to yourquestion.
(04:30):
You ask it a question and itgives you like a primer.
That comes with some benefitsand drawbacks too.
It should be noted, like whenyou think.
For example, when you went anddid a traditional Google search,
it gave you however manyresults it could be 20, 30, 100.
And you got to choose whatinformation you thought best
answered your question.
Llms present it as a briefingand it presents it from a
(04:54):
position of authority.
Like here is the answer to yourquestion and that can have some
drawbacks, mainly because A it'sproviding an answer and that
may or may not be the correctone or the one that works best
for you.
And two, we know for a factthat generative AI programs do
hallucinate.
That means they just simplymake up the answer and they're
(05:17):
doing that because they'retrying to.
It's basically like an averageof words, so you have to be
careful what you believe.
Peter Woolfolk (05:23):
You know I have.
Once I read your article, Ilooked into some of the other
LLMs and I jumped onto aperplexity and I asked a very
basically a question such as youknow, how many listeners does
the Public Relations Reviewpodcast have?
And while it was very, veryquick with that coming back, I
was shocked at how muchinformation came back, and it
(05:45):
came back in a format as thoughwe were having had had a
conversation about this.
So I was, you know, pleasedabout it.
It did not give me a specificnumber, as you said before it.
Let you know, we went here andhere's what we're here, here's
what we got there and theseother sort of places.
So I was not only was I pleased, I was was just surprised at
the format that it came back in,and I think this might be
(06:07):
helpful to a lot of people whomight consider using these.
Frank Strong (06:11):
Yeah, I completely agree.
It's incredibly amazing and Ithink your example brings up a
good question, because you'vehad this podcast for a long time
.
So it's been indexed and theselarge language models have been
able to be trained on all kindsof content and obviously,
including your podcast.
That's how it's able to returnthe answer, and so the question
(06:32):
that it conjures is if LLMs aregoing to be the new front page,
and there's some evidencesuggesting that this is
happening.
Google's search traffic hasdropped for the first time since
2015.
And it was a small amount, butthere is data that suggests
referral traffic from LLMsPerplexity, chatgpt, claude
Gemini, google Gemini that'sgrown, so we know that this is
(06:56):
in transition.
So the question that PRprofessionals should be asking
themselves is how do we get intoan LLM?
You've gotten your podcast intoan LLM and that is one approach
, right.
You have to have content, and Ithink a lot of the answers to
the question of how do I pitchan LLM comes back to the
fundamentals that PR peoplealready know.
It's having a strong point ofview, it's being able to
(07:19):
articulate a story so that itgets picked up, it gets a
visibility and indexed.
It's answering specificquestions right.
When people go to a traditionalsearch engine and they type in a
search query, that is, bydefinition, an expression of
need I'm looking for information.
So answering specific questionsis a great way to get crawled
(07:40):
and indexed by an LLM to providethat response great way to get
crawled and indexed by an LLM toprovide that response.
And then, last but not least,the traditional aspects matter.
So traditional media is goingto be deemed highly credible by
LLMs and so it's going to usethat to train the data and
return answers.
All of the traditional thingsthat PR people did.
Press releases, for example.
(08:01):
There seems to be some evidencesuggesting that press releases
are being used, and that makes alot of sense to me, because
when you think about pressreleases relative to the writing
on the rest of the web and wecan make fun of you know press
releases are promotional andpeople do funky things with them
, but the writing is probably ahigher quality relative to the
rest of the web, and so it's areally good data set on which to
(08:25):
train these models, and I thinkthat's why it's happening.
If there's a caveat, it would bethat at some point this
happened with Google.
Press releases were a good wayto get into search engines.
Google started deprecating thelink value because they don't
want you to be able to buy apress release to get into their
search engine.
That's kind of gets spammy.
I suspect something like thatwill happen with LLMs at some
(08:46):
point.
Peter Woolfolk (08:47):
Where else do you see LLMs fitting in?
Because you know some of thethings we've talked about here
and I think I may have mentionedearlier that some organizations
where consumers could call intothe company and ask questions
about the company or companyproducts or those kinds of
things Would this be anappropriate use for LLMs?
Frank Strong (09:08):
Absolutely, and I think there's two ways to look
at that.
One would be to ask a question.
Just like you asked thequestion of perplexity about
your podcast, you can go to anypublic generative AI tool and
ask a question about a companyand it's going to return results
.
And that's the kind of mediarelations aspects that PR people
need to be worried about is howare we being presented in these
(09:29):
LLMs?
That's one use case.
The other use case along thoselines is companies can purchase
technologies, these generativeAI tools, and train them on
their internal documentation.
If you have a product, forexample, a technical manual,
that can consume the wholetechnical manual, probably do it
better than a human, and thenit's trained on the answers that
(09:52):
customers may have about yourproduct.
So come to a chat bot on yourwebsite and they ask a specific
question and that AI has beentrained on your product
documentation.
It can return a very goodanswer, which is a far it sounds
like a far better approach thansaying here's a bunch of help
desk articles.
Go ahead and search throughthese and see if you can find
your answer and then, if youcan't, submit a support ticket
(10:15):
and we'll get back to youwhenever we can and we'll
schedule a call and go throughyour issue, right.
So it's just improving theexperience, I think, from a
customer service perspective,and that's great use case,
that's that's being entered nowwell you know, I certainly see
that as a viable function.
Peter Woolfolk (10:29):
I'm sure there might be some companies using
that already.
As a matter of fact, I haven'tsee where they've got.
Some companies have a on videosthat can respond to consumer
questions, and so I'm sure thatI Other organizations are going
to use this as a PR tool because, yes, we can answer your
question right away.
You don't have to wait for areal person to show up.
(10:49):
You know, as soon as you callus, we can get back to you, or
we can answer you right awayeven with someone talking to you
, and in some cases, I wouldn'tbe surprised if they can take a
real human face and animate itso that it looks like it's
actually talking.
In some cases, I think that isbeing done.
Frank Strong (11:07):
Oh yeah, that's definitely happening.
Some of the visuals are reallycompelling.
I mean, they're able to makeyou know fully-fledged movies
based on some generative AIprompts.
It's actually crazy.
One of the ways that you cannotice, though, is that, for
whatever reason, the same waythat generative AI tends to
hallucinate with text-basedanswers, it makes some wonky
errors with visuals.
(11:27):
Like, if you ask it to create apicture and you look at the
hands it tends to have likeseven fingers on it.
It'll have like weird thingslike that.
So these are like telltalesigns and videos and images that
you're able to spot some ofthis.
Peter Woolfolk (11:40):
Well, what are some of the things, some of the
other things that you see in thefuture for public relations
with these LLMs?
Frank Strong (11:48):
Well, I think we're in the middle of a sea
change right now, like now isthe moment right If you were
around when the Google searchengine started becoming the
dominant source, the firstsource of place that people went
to for information.
We're seeing thattransformation happening now.
Google, the traditional searchengine, is very worried about
this by the way.
That's why they're investing inGoogle Gemini, and you may see
(12:10):
what they call AI artificialintelligence overviews in search
.
If you go to Google right nowand you type in a search, it's
going to give you an AI overview.
So they're working on this andthey're playing catch-up chat.
Gpt cause them by surprise, butI think this is an area that is
emerging rapidly.
We can see that it is takingover as a primary source because
(12:30):
it provides a very neat, clean,simple answer when people ask
questions, and so we got to beup on it.
It means we're going to have tomonitor our reputation there.
How are these LLMs presentingthe organizations that we
represent, whether it's anemployer or a client?
How do we correct things ifthey go awry?
Right, what happens whenthere's crisis communication?
If it's getting something wrongand it sends off a tizzy, how
(12:52):
are we going to go about fixingthat?
And then, generally, how do weget into the LLMs, like we've
described already in the podcast?
So this is emerging and it'sgoing to unfold over the next 12
to 18 months.
We're going to see an awful lotof change.
Peter Woolfolk (13:08):
And I also see it as a sales tool as well,
because obviously you've gotcars, tv sets or whatever else
you might be selling.
That probably is a lot easierfor in some cases LLMs and
whatever sort of visuals thatthey have with that can save a
lot of time and answer a lot ofquestions quickly to consumers.
Frank Strong (13:29):
Yeah, I completely agree.
I think this is an extension ofa trend that we've seen already
and that is in the old days,you know.
Let's say you had to buy a carIn the old days, you didn't know
much about cars.
You went to the car dealership.
The salesperson knew everythingabout the car.
Then, when the internet got big, you could do all kinds of
research and not from justofficial review companies, but
(13:51):
you could see individual userswould write reviews.
You could get a whole lot moreinformation and walk into a
dealership prepared, knowingwhat you want, what the specs
are, how it compares.
I think we're going to seesomething like that become even
more refined, more repolishedand better answers from
generative AI going forward.
That's going to be a lot easier.
(14:14):
Instead of spending hours andhours doing searches and trying
to pull information together,it's basically presenting you
with a dossier.
That's an answer to whateverquestion that you ask, which is
pretty incredible.
Again the drawbacks are you'vegot to make sure it's accurate.
So you mentioned perplexity Inmy observation.
They were one of the first LLMsto provide source links, so it
will tell you where it got theinformation.
(14:35):
You can click the link and gocheck out the source and see.
Is this credible?
Do?
I believe that, and all of the,as far as I can tell, all of
the other LLMs have sincefollowed suit, because you know,
we know, working in publicrelations, having you know
credible sources, citing yourdata, is important, and so we're
starting to see the LLMs dothat.
There's also one other emergingproperty that LLMs are the
(14:58):
companies that make them, areworking on to combat the
hallucination, right.
So, for whatever reason theydon't really understand why, but
you can ask a generative AItool a question and if it
doesn't know the answers, oncein a while this happens less
more often than not it gets itright, but once in a while it
doesn't know the answer and itjust makes one up.
It just gives you an answerthat's completely fictional and
(15:21):
it presents it with suchconfidence.
You're like you know it'd be,like somebody like, well, here's
the answer, I'm absolutelyconfident it is right.
And you don't know.
They don't really know whythat's happening, but one of the
things they're working on todevelop is this new development
called reasoning, and whatreasoning is doing is it's
showing you.
It's showing you the steps ofhow the LLM got to the answer
(15:43):
that it is, and I think all ofthese developments will make
LLMs more credible, moreconclusive.
They're going to get betterover time and deliver better
answers.
So it's something that the PRcommunity really needs to pay
attention to.
Peter Woolfolk (15:54):
You know, I would think so and I just keep
thinking of some other areasthat it can be used in.
I mean public relations, sales,you know, just general
information.
You know, at a conference ofsome kind, a lot of different
ways that or students can callup and or interact with it to
get responses to particularquestions, have to do classroom
(16:16):
materials and so forth, and itcomes through and I guess they
could add visuals to this thingas we move along yeah,
absolutely mean there's a coupleof very definitive applications
in public relations that arealready happening.
Frank Strong (16:29):
If you're using most of the big PR vendors, if
you're using MuckRack, Cisionjust added it.
Meltwater has got a huge AIteam.
I mean they've got scientiststhey've gone out and done
acquisitions just to hirescientists with PhDs to help
them work on this stuff.
All of the big ones have that,so they can do things like if
(16:49):
you're using media monitoring,it can better categorize, you
know placements and put tags andcategories of it.
Is it a story?
Is it a mention?
Is it a backlink?
Pr?
People are using it forbrainstorming, right Like you
just need to.
You know, sometimes looking ata blank page is the biggest
hurdle to getting started whenyou're doing a writing project.
Well, get into a generative AIsystem.
(17:10):
Ask it a bunch of questions.
Announcer (17:11):
It's going to give you a whole bunch of ideas.
Frank Strong (17:14):
Summarization Exceptional at summarization,
like here's a big report.
I don't have time to read thiswhole thing, but, you know, read
it for me and give me the fivekey points I need to understand.
Take me and give me the fivekey points I need to understand.
Take my blog post or pressrelease and give me 10 tweets
that I can then go ahead andschedule, or Facebook posts or
LinkedIn posts like a lot of usecases.
There's a few vendors that aredoing some interesting things
(17:36):
where it has prompts that willhelp you write a pitch or help
you write a press release.
I'm not big into that.
I really think.
I think it's good forbrainstorming, but I'd rather
have a human write.
I think humans write betterthan generative AI At least PR
people do.
But one of the interestingthings that one of these tools
does is it will look at thewords that you are using as
(17:57):
you're writing your pitch andthen it will go out and review
articles that are currentlypublished by reporters and come
back as you're writing yourpitch and saying you should
pitch these people becausethey've written these topics,
and here's the link to that.
That's a pretty coolapplication of generative AI.
Peter Woolfolk (18:12):
You know, one of the things that I have done
with things such as chat, gpt,for instance, that I might lay
out some things that I'd likefor you to write for me on this.
Then, once it comes out, then Iwill take it and make it my own
, in other words, put in thewords or other things, and that
makes it as if it was comingfrom me.
But it has saved me a lot oftime and you know and I'm just
(18:37):
tweaking it so that I'mcomfortable with what it says
rather than just ripping andrunning with it that could cause
some problems down the road.
Frank Strong (18:41):
Yeah, no, a hundred percent.
I mean, I think what you saidis really important, that if
you're using these tools, youknow, think about it as a first
draft and then edit it and makeit your own.
The problem is is I mean,generative AI is basically a.
These LLMs are giant models ofwords, and it understands how
(19:03):
people tend to put those wordsin a certain order.
Based on probability, it isprobable that a sentence is
going to read a certain way andhave a certain structure, and so
what that means is the answersthat generative AI is giving you
is, by definition, an average.
Here is the average languagethat most people use, and we
know in PR, we don't want to beaverage, we want to be different
(19:23):
.
We want to explain to peoplewhat's different about our
products what's better right.
So it's really important tomake sure you have that human
touch.
But I think your application ofthe tool is a good one.
I do that an awful lot where,if I don't like the way a
sentence is written, I'll ask it.
You know, rewrite this sentence, give me a better structure.
Is there an easier way to writeit?
Or another application is?
(19:45):
I don't really like thisheadline.
I'm using the sub headline here.
Can you give me fivealternatives?
Peter Woolfolk (19:51):
And it'll.
Frank Strong (19:51):
it'll belt them out and I can choose to use it.
I can rearrange, I can mix andmatch, and that's helping me to
make my content better.
Peter Woolfolk (19:58):
Well, you know, actually with this podcast now
it, it records everything andthen it cranks out not only the
transcript for me, but it alsosuggests titles for each episode
.
Now I'll read those.
Either I like them, or I don'tlike them, or I'll make
alterations to them, but itgives me all of that.
It gives me blurbs for Facebookand other social media posts and
(20:24):
those kind of things.
I read them.
If I like them, fine.
If not, I change them.
But these things are a greattime saver is what I'm really
seeing here.
There's a huge amount of timebeing saved and helping you get
the job done faster and, in somecases, maybe even more
accurately, because it may comeup with some information that
you had overlooked, because, asyou said, these things reach out
(20:46):
to a wide range of places topick up the information and add
it to what it's giving to you.
Frank Strong (20:51):
Yeah, that's right .
I mean, it's almost cliche tosay this at some point, but the
topic comes up over and over ispeople are like are we going to
be replaced by generative AI?
And I don't think we are goingto be replaced as individuals.
Generative ai is very good atspecific tasks, but it can't do
a job.
You can't say go be my prperson for me right, I can't do
that, but jet pr.
(21:12):
People need to pay attention tothis stuff because, while you
won't probably won't be replacedby generative ai, you could be
replaced by somebody that haslearned how to use it
effectively and been able to bemore productive.
That's a real risk.
Peter Woolfolk (21:25):
You know the other thing, when I look at a
lot of times about publicrelations, public relations a
lot of times is aproblem-solving exercise, and if
the platform doesn't know whatthe problems are, it can't offer
you any solutions to them.
Frank Strong (21:42):
So you know questions like how do we fix ABC
and D?
That's right, you can use it tobrainstorm possible solutions.
I completely identify with theproblem solving because I'm
obviously a consultant now, butI spent 10 years on the in-house
side and anytime somebody got aquestion that they didn't know
how to answer, they sent it toone of two places Corporate
communications or the legaldepartment.
(22:03):
So you're constantly gettingthese crazy wild requests that
you're like why am I getting?
Peter Woolfolk (22:07):
this.
I don't know what to do.
Frank Strong (22:09):
But if you have an LLM system that's trained on
your internal documentation,it's a great place to start.
Start brainstorming, I mean,you know, here's the problem,
how do I solve this right?
It's just a good way to showhow you can be more efficient
and effective and productiveusing these tools.
Peter Woolfolk (22:23):
Great Well, frank.
You've provided us with anawful lot of information really
on LLMs.
Are there any closing?
Frank Strong (22:39):
remarks you think that we need to cover so that
our listeners can be completelynot completely, but even more up
to speed on these LLMs.
You know, I think the key is toget out there and experiment
with them.
Make it a point to set sometime aside and try to use some
of these tools.
It doesn't matter, really,which one you use.
If you like ChatGPT, go for it.
If you like Perplexity, whichis one of the ones I really like
, go for it.
If you want to use GoogleGemini, go for it.
But take some time toexperiment with brainstorming,
(22:59):
with rewriting headlines, withhaving it giving you blurbs to
post on social media.
Give this stuff a try so thatyou get some experience with it
and you know what thepossibilities are, because the
changes are going to come fastand furious.
The development is happening ata pace that is just
unbelievably quick, and thesetools are getting better and
(23:19):
better.
I mean think about when thisthing, when ChatGPT launched
what two years ago?
Peter Woolfolk (23:23):
When it first started.
Frank Strong (23:25):
It was really a stunning event.
People were like, oh my gosh,amazing.
And it's come so far in such ashort time, absolutely.
I saw this, you know, it waskind of a meme not too long ago
and it showed like a chariotfrom thousands of years ago and
then it showed a horse con drawncarriage in like the 1800s.
It's like, look, this is 2000years, technology hasn't changed
(23:47):
much.
But then when you think aboutthe pace of change from the
right butter brothers launchedan aircraft to now where we're
putting robots- and helicopterson mars or landing these things
on, like we just took a samplefrom a comet in outer space.
We landed a device on a comet inouter space, took a sample and
brought it back to Earth.
Like the pace of change isremarkable.
(24:08):
That is going to happen inthese generative AI applications
tools.
So the pace of change hasincreased.
Peter Woolfolk (24:15):
Well, frank, let me say thank you once again.
You have provided us with somevery, very valuable information.
I've learned something, I'msure our listeners will learn
something, and I'm certainlyglad that you've been a guest on
the Public Relations Review fora second time, and perhaps down
the road, I believe, we'llprobably reach out to you again.
Frank Strong (24:34):
Thank you, I'll be here.
Thanks for having me on, peter.
Peter Woolfolk (24:36):
Okay, and to our listeners, thank you for
listening.
If you've enjoyed the PublicRelations Review Podcast, we'd
certainly like to get a reviewfrom you and also let your
friends know to listen to thenext edition of the Public
Relations Review Podcast.
Announcer (24:52):
This podcast is produced by Communication
Strategies, an award-winningpublic relations and public
affairs firm headquartered inNashville, Tennessee.
Thank you for joining us.
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.