#5 Ingenious GPT Use Cases for Smarter PPC

Google has so far languished behind in the recent AI revolution.

It may well catch up, but it has some ground to regain, both on the competition, and from its very publicly lacklustre efforts with Bard.

In the sphere of Google Ads, the ‘Conversational Experience’ is Google’s best offering yet – but there’s nothing fundamentally new about its output (we’ve learnt well to take Google’s recommendations with various doses of salt… and this update simply puts those recommendations on steroids)

So for now, the game-changing applications of AI for PPC are coming from outside of Google. Primarily, from ChatGPT.

1 Thematic Search Term Filtering

A few weeks ago, I was looking to gather all of the ‘location-based’ search terms from a very generic ad group.

This would usually be quite a manual process of looking through the long list, and pulling out those that contained the name of a town/county/ postcode etc…

But this is just the kind of thing that ChatGPT can do quickly.

Prompt: “From the list below, print only the items that contain a place name [pasted list of search terms]”

This works by simply pasting the search terms into the prompt (up to a few hundred of them) – but with the ability (in ChatGPT Plus) to upload files along with your prompt, the limit is far higher…

And of course inputting a table with search term data allows you to ask not just for the list, but for aggregate or average metrics relating to whatever subset of search terms you are interested in.

Without a specialist tool, it is arduous to carry out this kind of ‘thematic filtering’ where the common thread is a certain ‘idea’ rather than an identifiable string. This kind of filtering requires some sort of ‘intelligence’. Enter AI.

2 Investigations

A client was recently concerned about net profit on a particular arm of their business, which is fed predominantly by PPC.

They gave me detailed sales data (#units sold of different types, various costs, revenue, gross and net profit by month, stretching back for years…) and asked me to dig into what we could do on the PPC side, to move the needle on that metric.

So – I distilled this to specific questions that I could then ask ChatGPT to explore from the bank of data.


1 Is volume or CPA the prime driver (do fixed costs cause profitability to increase with scale, or do we see better Net Profit even in lower-volume months when CPA is low)?

2 Of the two strands of activity – residential and commercial services – does the latter (with its higher CPA and higher revenue, but unknown ultimate profitability) – do more harm than good?

It’s a lot to dig into, but straightforward for ChatGPT.

All it takes is a prompt to run the correlations between the key metrics, and you’ll see which of the suspect variables really are associated with lower/higher net profit.

Quick correlation analysis is just one of the myriad functions ChatGPT can bring to your analysis toolkit – but a particularly helpful one when it comes to identifying points of leverage in the digital marketing ecosystem.

Another neat example of analysis capabilities comes from this post by Mike Rhodes , on using ChatGPT to cut through the hazy data presented in the interface (which has become Mike Rhodes’ thing, much to all of our benefit).

Read first Mike’s explanation of why this ‘incremental ROAS’ calculation is needed*, then the ingenious prompt to reach an answer in seconds.

Three key notes here:

• Pay attention to incremental ROAS in general (* for more on its importance in principle, see my blog post on incremental ROAS)

• ChatGPT can do these kinds of calculations for you

• You can use screenshots as your inputs when you use GPT4-V


3 Data Visualisations

The Google Ads Report Editor is great (see my blog post on the same for a deeper look) but it has limitations.

For example, in a chart with landing page as the primary dimension, there’s no way to filter by landing page, to narrow down the selection shown:

In cases like this, where the Report Editor can’t produce the visualisation you want ChatGPT can do the following:

Read the table attached from row three [this is usually worth stipulating with data downloaded from Google Ads to avoid occasional confusion. ed]. Please create from this a time series chart showing the conversion rate by day for the two landing pages [...] and [...]. The other landing pages and their data can be fully ignored.

4 Extracting Benefits

Stepping out of the Google Ads account itself, a good way to identify key benefits worth emphasising in our messaging, is to look through reviews for the site we’re advertising, and pick out recurring themes showing what customers like, and value, about the product.

This is another potentially arduous task made quick and easy with ChatGPT.

Paste in all of the review text, and ask for an analysis of the key benefits people are writing about.

If you like, add to this prompt a request to write, e.g., 10 30-character headlines calling out those key benefits!

5 Ad text

It was – above all – ChatGPT’s extraordinary ability to write coherently across a range of topics, tones and tasks that first ushered in the new AI era in which we find ourselves.

So of course, ChatGPT can write good ad text.

While there are many ways to make use that ability… this excerpt from a prompt shared by Erik Van Der Veen highlights two useful aspects of ChatGPT’s ad writing skills.

“…make a third table with the name “Ads” with these column headers…

Headline1 | an advertisement text of up to 30 characters without exclamation marks that fits the keyword of that row exactly
Headline2 | like Headline1 but using the Cialdini principle of social proof, without using an exclamation mark
Headline3 | like Headline1 but using the Cialdini principle of scarcity, without using an exclamation mark
Description1 | an advertisement text of up to 90 characters with an optional maximum of 1 exclamation mark that fits the keyword of that row – using the Cialdini principle of sympathy
Description2 | similar to Description1 but using the Cialdini principle of authority…”

Behold the output:

The two notable features…

• ChatGPT is a great mimic. If you want ad text in a certain style, with a certain emphasis, you can have it… and Cialdini’s principle’s – as used in this example – are a good choice (see my vid on the use of those principles in ads and landing page text for a deeper look).

• It can output in Google Ads Editor ready format… either as a table in the interface – as shown above – or as a ready-formatted csv to download > upload.


A few weeks ago Open AI released the GPT store, which is now filling up fast with custom GPTs, pre designed to excel at specific tasks.

So, how can you seek out the best GPT uses for PPC, and keep track of them as the field develops?

Well, that sounds like a good question for a GPT specifically designed to advise on PPC.

Here’s mine…

Google Ads Upleveller

Share this post

Lastest Posts

Old dog or master? Mixed messages from Marketing Live.

“It’s like trying to get an old dog to get up off the sofa” That…

Google Ads Updates - Q1 2024 Roundup

No slowdown in the pace of change with Google Ads updates in Q1 2024… Here’s…