Another week, another post about AI. Sigh.
Companies developing AI and their investors are banking on AI taking off, and so once again, it’s being crammed into products that arguably don’t need it, and will never need it.
I’ve written about the use of AI in Google Ads previously, but recently it’s taken things to the next level by giving you the capability to create ads that are almost entirely AI-generated in Performance Max campaigns. This includes headlines, descriptions, and images.
A few of my clients were lacking images for me to add to their Performance Max campaigns, so I thought there was no harm in trying it out.
Here’s my experience of using GenAI in Google Ads.
Don’t even bother if you need to include either of these two things
One of my clients operates in the FinTech space.
The images it generated were completely unusable for two reasons:
- The GenAI cannot produce readable text
- The GenAI still doesn’t know what a human hand looks like
I asked it to generate alternatives using different prompts, but it quickly became apparent that the industry they operate in is too complex to create AI images that fit the bill.
Not bad, but some bonkers suggestions
Another client I tested the GenAI on was a healthcare provider.
I’ll be honest – it didn’t do too badly here.
I asked it to create some images of a private doctor’s waiting room and consultation room, and as long as you don’t look too closely, you probably wouldn’t know the images were AI-generated. I still can’t use them seeing as this isn’t what the inside of their clinic looks like, but it was still interesting to see what it produced.
There were a few issues with the GenAI:
- Created a description that exceeded its own character limit
- Created an example prompt for the AI which it then said it couldn’t do
What else can Google Ads GenAI produce?
Intrigued by the failures and successes of the tool, I decided to explore further.
As well as being able to create AI images from scratch, Google has been showing off its editing features. An example that Google gave was adding a party hat to an image of a cat.
First of all, I was interested to find out how well the AI could create images of cats. Images and videos of cats are some of the most viewed content on the internet, so I figured it would be pretty easy for the AI to produce some very cute and realistic-looking cats.
A cat cafe has recently opened near me, and I thought about how they could’ve created ads for their business prior to its opening to generate some buzz. I prompted the AI to create a cat cafe on a sunny day.
Hoo boy. It produced nightmare fuel.
I mean, good Lord.
Obviously, nobody in their right mind would use these images in their advertising campaigns.
But working with this base, I went all in to test out the GenAI editing features.
It could indeed, add a pretty convincing party hat.
It did less well with adding a delicious plate of birds and mice in place of a cake (there aren’t even any mice).
When I asked it to expand the image, again, it did a pretty good job.
But removing the background was an absolute fail.
And it did alright with removing the cat’s arse from the picture, providing you don’t look too closely.
After seeing what it could do to edit an AI-generated image, I uploaded a stock image to see how well you could make edits. And, knowing that it struggles with people, I thought that would be a bit more of a challenge for it.
‘Expand image’ worked surprisingly well here. If you really zoom in, her chin and hairline don’t quite match, but it’s passable.
I was quite surprised to see that despite the issues GenAI has with facial features, it did edit some lippy and teeth onto the image when prompted. It doesn’t look brilliant, but better than anticipated.
At this point, I began to see just how much I could change using all of the AI editing tools available.
And this was the end result. Absolutely dreadful.
So, should you start using GenAI in your Google Ads campaigns?
In general, I don’t think that the GenAI images are good enough to use in ad campaigns. Most businesses will have some budget to produce their own bank of images, and for those that don’t, stock images can help fill the gap.
Plus, the time you spend faffing about with prompts to get something remotely usable, you could have just gone on a free stock image website or knocked up some images in Canva.
While I was experimenting, I kept thinking, who asked for these features? Did anyone? I definitely can’t imagine big brands (especially visual ones) using this tool, but I can’t think of ways smaller advertisers can utilise this, either.
Not only that, I felt incredibly guilty smashing in these prompts which I know are using up masses of resources just to generate these sub-par images.
Unless this tool improves significantly – and quickly – I can’t see this feature being used by advertisers often, or at all. But with the amount Google has riding on AI being the next big thing, I can’t see them admitting that it’s a failure any time soon.
