Google Ads Copy Hack Using Google Natural Language Tool and ChatGPT

Glen: 00:00:11

Hi everyone, Glen from Solutions 8 here and here we are

2

: 00:00:13

in the Solutions 8 test kitchen.

3

: 00:00:15

It's actually physically my kitchen

that I've got my office set up in.

4

: 00:00:19

but this is where I'll be,

running tests, experimenting

5

: 00:00:23

with things, playing with things.

6

: 00:00:24

And what I wanted to do is shoot

you a quick video on how I use the

7

: 00:00:28

natural language API tool, from Google

and incorporate that, with ChatGPT.

8

: 00:00:33

With some other, tools, this is for copy

and in particular finding positive copy

9

: 00:00:38

that Google's going to look at favorably.

10

: 00:00:40

Now, when we are creating ad copy,

now there's generative AI tools

11

: 00:00:44

coming out, inside Google now.

12

: 00:00:46

at the time of this recording,

Gemini Ultra was just released today.

13

: 00:00:51

So that will be interesting to see how

all that pans out and how it compares

14

: 00:00:55

with GPT, but at this stage I'm using.

15

: 00:00:58

GPT four plus.

16

: 00:00:59

and in this example, I'm gonna show

you how to use the cloud natural

17

: 00:01:02

language tool, and chat GPT to

create more copy ideas, add ideas

18

: 00:01:07

so that you can have, with a view

of improving click through rate.

19

: 00:01:11

And conversion rate in your ad copy.

20

: 00:01:13

in this example, I'm just

going to use solutions eight.

21

: 00:01:17

I'm just going to grab all the copy.

22

: 00:01:18

That's on our homepage, come into

the natural language tool under demo.

23

: 00:01:22

You'll see here, it says, try the API.

24

: 00:01:24

Just paste everything in

there, hit it, run it.

25

: 00:01:26

It then outputs all this gobbledygook and.

26

: 00:01:31

sentiment.

27

: 00:01:31

I want to click on the sentiment tab

and you'll see here that anything

28

: 00:01:34

green has a positive sentiment, which

is Google's natural language tool.

29

: 00:01:39

it's API saying this

copy is deemed positive.

30

: 00:01:44

Google likes positive

sentiment in ad copy.

31

: 00:01:47

So I've gone through and I've just

copied everything that's green

32

: 00:01:50

and I have this document here now.

33

: 00:01:53

Then what I do is I've created a prompt.

34

: 00:01:55

I come over to chat GPT

and I go to GPT for now.

35

: 00:01:59

What I like to use is a GPT,

an actual GPT there's two.

36

: 00:02:04

I like to play with one

school professor synapse.

37

: 00:02:06

One's called super synapse.

38

: 00:02:07

Now it's made by, These guys over

at Synapse Labs, I actually like

39

: 00:02:12

using these GPTs because it's

acts like a bot or an agent.

40

: 00:02:18

I'm going to click start, fire

up the SuperSynapse what I've

41

: 00:02:22

done is I've created this prompt.

42

: 00:02:23

And, in this, I've got, I want you to

act as an expert Google as copywriter.

43

: 00:02:27

You have been conversing

with Google as 20 years.

44

: 00:02:30

You utilize the styles

of legendary copywriters.

45

: 00:02:32

chat GPT or any large language

model likes to use markdown.

46

: 00:02:37

Now, markdown is like a

simplified version of HTML.

47

: 00:02:40

And you'll see that headings are

determined with these hashtags.

48

: 00:02:44

Lists are done like

that, bullets like that.

49

: 00:02:45

And bolding, Is done like this.

50

: 00:02:48

So there's two stars

asterisks on the site.

51

: 00:02:51

So I bolded that to make ChatGPT

aware that this is important.

52

: 00:02:57

from a HTML perspective, if you wrap

some words in a HTML tag with bold, the

53

: 00:03:02

search engines determine based on the

tags attached to keywords if something's

54

: 00:03:07

important or you want to pay attention to.

55

: 00:03:08

Now, what I've done is I've got, 14.

56

: 00:03:11

Very famous copywriters that renowned.

57

: 00:03:14

and then I pumped it and said, I've

grabbed copy from this website.

58

: 00:03:18

I've told it, I've run the copy through

the natural language demo, looking

59

: 00:03:22

for highly rated, positive sentiment.

60

: 00:03:24

Usually I'm using the copyright on my

15 headlines, full descriptions for

61

: 00:03:27

each of the legendary copywriters.

62

: 00:03:30

then this means 15 headlines for

each for the first couple of when

63

: 00:03:34

asking to you moved to the next,

64

: 00:03:37

make sure you've brought 15 headlines,

four students for each legendary

65

: 00:03:40

corporate before we move on to

the next legendary copywriter.

66

: 00:03:42

Each headline, must be on its own line.

67

: 00:03:44

Make sure they were high, right?

68

: 00:03:45

High and seven, blah, blah.

69

: 00:03:46

Can you help me with the split?

70

: 00:03:47

So then I.

71

: 00:03:47

Again, I use this.

72

: 00:03:49

So this is a heading one tag.

73

: 00:03:51

I'm going to copy this, come over

to chat CPT, paste that there.

74

: 00:03:56

Now I'm going to grab all of this

positive sentiment copy from our

75

: 00:04:01

website, it'll do a bit of analysis.

76

: 00:04:03

This is how they've written this.

77

: 00:04:04

GPT.

78

: 00:04:05

It's giving you a quest.

79

: 00:04:06

We'll start with, for David Ogilvy.

80

: 00:04:09

So Ogilvy was, the first

legendary copywriter on the

81

: 00:04:12

list there, David Ogilvy.

82

: 00:04:14

So they've kicked off with

that, expand up the bar.

83

: 00:04:18

I like this process.

84

: 00:04:20

And I like to have, chat TPT go

through all of them because there's

85

: 00:04:23

all different types of headlines

that you can use in descriptions.

86

: 00:04:27

continue to the next one.

87

: 00:04:30

So now it's going to say, sure.

88

: 00:04:33

The next one should be Bill burn back.

89

: 00:04:36

So this now it's going through

each legendary copywriter and

90

: 00:04:39

it is, based on its knowledge.

91

: 00:04:42

It's tells you what his style

is like, and it will then start

92

: 00:04:46

writing copy around his style.

93

: 00:04:48

Okay.

94

: 00:04:49

and you essentially start getting a

whole heap of different headlines and

95

: 00:04:53

descriptions in different formats that

you probably wouldn't, Even consider now,

96

: 00:04:57

sometimes these are a little bit long.

97

: 00:04:59

The headlines you might just

need to tweak them and just them.

98

: 00:05:02

Same with descriptions.

99

: 00:05:03

They can go over 90 characters,

going to grab another text document.

100

: 00:05:06

And I'm going to grab all these headlines

101

: 00:05:09

and I'm going to grab these descriptions.

102

: 00:05:11

I've only done two, but ideally for

this exercise, I would do all of

103

: 00:05:15

the, legendary copywriters just so

I can get different styles of copy

104

: 00:05:20

and I can go through the headlines.

105

: 00:05:22

I can go through descriptions

and go, I like that.

106

: 00:05:24

I like how that's worded or that really

encapsulates what I'm trying to get in

107

: 00:05:29

ads and you essentially can have a massive

repository of copy that you can refer

108

: 00:05:33

to when you split testing ads and copy.

109

: 00:05:36

So I'll just copied.

110

: 00:05:37

all that into a text doc, what I'm

going to do is, copy all of that.

111

: 00:05:42

Now come into another version of here.

112

: 00:05:47

I'm just going to paste all

of that I've just copied.

113

: 00:05:50

Click analyze.

114

: 00:05:51

Now this is probably what you'll get.

115

: 00:05:52

Satellizing.

116

: 00:05:54

And what should happen is because

we've given positive sentiment to chat

117

: 00:05:59

GPT and it's created different styles

of copy around these copywriters,

118

: 00:06:03

we theoretically should get positive

sentiment back from our copy.

119

: 00:06:07

So that's why I'm running this test

start again that paste in my copy.

120

: 00:06:12

from chat GPT, I'm not a robot

121

: 00:06:14

go and we want to get the sentiment

we got one there, one yellow one,

122

: 00:06:18

but pretty well, most of the copy

that we've got here is all green.

123

: 00:06:22

So that's how you can do it with chat GPT.

124

: 00:06:25

Now to use GPT 3.

125

: 00:06:27

5, if you got the free version, I

find it's not as good, but if I'm

126

: 00:06:31

using GPT, And I'll put these links.

127

: 00:06:34

There's the GitHub prompt, so if

you come over to SuperSynapse,

128

: 00:06:39

this prompt is too long.

129

: 00:06:41

If you click on the prompt here, if

I copy this and I come into ChatGPT,

130

: 00:06:46

what I want to do is add a link.

131

: 00:06:48

Now the problem with this is you can

only have up to 1, 500 characters

132

: 00:06:53

and the super synapse is 3, 975.

133

: 00:06:56

So that's why you can

always use the synapse COR.

134

: 00:07:01

If I copy that prompt.

135

: 00:07:02

I can paste that in there and

that's actually too long itself.

136

: 00:07:07

grab the prompt, and just paste it in

there and you're more or less done.

137

: 00:07:10

Here it is.

138

: 00:07:11

It starts.

139

: 00:07:11

So you would do the same process.

140

: 00:07:14

it's hit and miss.

141

: 00:07:15

It's very spasmodic.

142

: 00:07:16

You don't get the same consistency with 3.

143

: 00:07:18

5 as you do with four.

144

: 00:07:20

but you can still get through it.

145

: 00:07:22

and it's faster, it's way faster.

146

: 00:07:24

So if I was to Copy that, prompt, and

then I'll put in the same copy from

147

: 00:07:29

our Solutions 8 web page homepage.

148

: 00:07:32

You can see it's way quicker at

doing the headlines, but like all the

149

: 00:07:36

descriptions, yes, continue to the next.

150

: 00:07:39

I hope you've enjoyed the video.

151

: 00:07:40

We'll put links to, in the description

below for the cloud natural language.

152

: 00:07:45

Like I said, it's been around

for a little while now.

153

: 00:07:47

It's nothing new, but for those of

you that don't, or weren't aware

154

: 00:07:52

of it, it's a great tool to use.

155

: 00:07:54

Ideally, what I'd like to do is create

our own, or external API tool that we

156

: 00:08:00

can use plugging into vertex, which

is Google's natural language tool.

157

: 00:08:04

In light of the fact that Bard Ultra

was released today, I think, apparently

158

: 00:08:09

it's supposed to be a GPT4 killer.

159

: 00:08:11

We'll be testing that

just to see how that goes.

160

: 00:08:14

But if this has been valuable,

please, Leave a like and comment in

161

: 00:08:19

below and if you've got any further

questions about this tool or anything

162

: 00:08:22

else, feel free to pop them through.

163

: 00:08:25

Until next time, have a great day.

164

: 00:08:28

Bye.