AI and advertising

Messages
4,503
Name
Graham
Edit My Images
No
I've mentioned before in AI threads that anecdotally, high end marketing/advertising is "mainly" rejecting AI generated imagery, because customers are reacting badly to products advertised with fake images, but this is the first time I've seen some research about it

This is a news item on research funded by Icelandair on AI images used in advertising, and on the back of it they are actively promoting that no AI images are used in their advertising, and encouraging other companies to do the same.


The letter that Icelandair has been sending out


And, a summary of the results, most of which is in the news article. Not enough information to assess how credible the results might be, but nonetheless interesting.

 
Last edited:
I've mentioned before in AI threads that anecdotally, high end marketing/advertising is "mainly" rejecting AI generated imagery, because customers are reacting badly to products advertised with fake images, but this is the first time I've seen some research about it
Many decases ago, I was on a marketing course run by the Newpaper Society and the key speaker brought up this very subject, in relation to airbrushing photos.

"Nothing annoys a customer more than you taking them for a fool", was the last line in her presentation and I've never found anything to refute that bald statement.
 
Many decases ago, I was on a marketing course run by the Newpaper Society and the key speaker brought up this very subject, in relation to airbrushing photos.

"Nothing annoys a customer more than you taking them for a fool", was the last line in her presentation and I've never found anything to refute that bald statement.
Apparently, children are extremely adept at identifying AI images in advertising and dismissing them as fake, and not to be trusted.
 
Advertising in general is not to be trusted!
 
If AI trains itself on previous stuff, you have to wonder how much of that is rubbish, in which case it isn't going to be very clever.
 
Advertising/design is a funny old fish. As an industry, it tends to reflect society. In very few cases is it ahead of the curve.

I've been working as a copywriter in ad agencies since the mid 90's and as fashions come and go, the one thing that has remained is that ad agencies all want to be the first to do something.

Far from rejecting it, in my experience, every single Ad agency (and I have friends at various other agencies) is falling over itself to adopt AI.

Partly because it's the latest buzzword, so clients want their agencies to be up on the latest trends.

As I mentioned previously in the other thread, part of the reticence for AI imagery has been to do with copyright. As publicly available AI engines (Midjourney et al) scrape the internet for source imagery and, depending on where it comes from, leaves people (our clients) open to being sued for using imagery without permission.

But it's being used on far more than just imagery, drafting copy, certainly. We use it almost exclusively now for internal presentations to clients. Far cry from the days when I started when you'd hand-draw scamp ads and take them to show clients, now you can pretty much generate a storyboard of stills purely from AI. You can even use AI to convert those stills to video.

It's definitely coming. This is one the first ads to celebrate using AI throughout:

View: https://youtu.be/alBJplIu8dY?feature=shared


The one thing I'd add however, is that there there is also a feeling of being very much 'anti' stuff in advertising. Being deliberately contrarian. The over-used phrase of "When everyone zigs, you should zag".

We all joke about the Ferrero Rocher 'Ambassador' ads or the infamous 'Junedad' advert from Sunlife. But bad advertising that gets noticed is better than advertising that doesn't get noticed at all.

When typography was mostly done by hand, everyone strove for perfection in their letter forms. When Macs came in and started making perfect type every time, you started to see people going out of their way to make distressed type or make it obvious they were using hand lettering. Not being perfect made them stand apart from all the perfect type. So there will be a period of ads being 'real' but I imagine only until AI becomes totally accepted.
 
but I imagine only until AI becomes totally accepted.
I hope that such a day never comes.

The whole idea of relying on any software to implement the correct decisions, in every case, is insanity squared. Relying on a black box, which you can't even debug, is insanity cubed.
 
Advertising/design is a funny old fish. As an industry, it tends to reflect society. In very few cases is it ahead of the curve.

I've been working as a copywriter in ad agencies since the mid 90's and as fashions come and go, the one thing that has remained is that ad agencies all want to be the first to do something.

Far from rejecting it, in my experience, every single Ad agency (and I have friends at various other agencies) is falling over itself to adopt AI.

Partly because it's the latest buzzword, so clients want their agencies to be up on the latest trends.

As I mentioned previously in the other thread, part of the reticence for AI imagery has been to do with copyright. As publicly available AI engines (Midjourney et al) scrape the internet for source imagery and, depending on where it comes from, leaves people (our clients) open to being sued for using imagery without permission.

But it's being used on far more than just imagery, drafting copy, certainly. We use it almost exclusively now for internal presentations to clients. Far cry from the days when I started when you'd hand-draw scamp ads and take them to show clients, now you can pretty much generate a storyboard of stills purely from AI. You can even use AI to convert those stills to video.

It's definitely coming. This is one the first ads to celebrate using AI throughout:

View: https://youtu.be/alBJplIu8dY?feature=shared


The one thing I'd add however, is that there there is also a feeling of being very much 'anti' stuff in advertising. Being deliberately contrarian. The over-used phrase of "When everyone zigs, you should zag".

We all joke about the Ferrero Rocher 'Ambassador' ads or the infamous 'Junedad' advert from Sunlife. But bad advertising that gets noticed is better than advertising that doesn't get noticed at all.

When typography was mostly done by hand, everyone strove for perfection in their letter forms. When Macs came in and started making perfect type every time, you started to see people going out of their way to make distressed type or make it obvious they were using hand lettering. Not being perfect made them stand apart from all the perfect type. So there will be a period of ads being 'real' but I imagine only until AI becomes totally accepted.
Only based on what I've read as I have no personal experience this is how I see it as well, and I have made similar posts to yours in other thread about the multiple uses of AI, so I'm not disagreeing with t

However, I was specifically referring to two things in my OP. High end advertising photography where you might have up 50 people involved in the shoot, including the client and art directors from the client and ad agency asking for subtle changes during the shoot. Subtle changes that could be expensive and unsatisfactory to do in post, or with an AI generated image. It seems it’s a bit like CGI, which was meant to threaten high end advertising photography, but in practice turns out to be far too expensive, except for special purposes.

However, the research posted in my OP wasn't about advertising agencies rejecting AI generated images, it was about consumers rejecting it, which I find more interesting.
 
Only based on what I've read as I have no personal experience this is how I see it as well, and I have made similar posts to yours in other thread about the multiple uses of AI, so I'm not disagreeing with t

However, I was specifically referring to two things in my OP. High end advertising photography where you might have up 50 people involved in the shoot, including the client and art directors from the client and ad agency asking for subtle changes during the shoot. Subtle changes that could be expensive and unsatisfactory to do in post, or with an AI generated image. It seems it’s a bit like CGI, which was meant to threaten high end advertising photography, but in practice turns out to be far too expensive, except for special purposes.

However, the research posted in my OP wasn't about advertising agencies rejecting AI generated images, it was about consumers rejecting it, which I find more interesting.
That's a fair pushback.

However, given the accelerated rate at which our own in-house AI is improving means it will be indistinguishable from the real thing very quickly.

Our in-house version has gone from basically being unusable to incredible just this year alone.

I don't think it will be long before consumers (or anyone) can't tell the difference.
 
Last edited:
I wonder if the Uncanny Valley theory has anything to do with this. Even people who aren't adept at spotting AI imagery will have a little spark in their evolved brain telling them that something is "off" with the image.

Along with that, I'd guess with so many AI models being trained on the same datasets (usually stolen) their output is anything but unique and human.
 
Last edited:
I don't think it will be long before consumers (or anyone) can't tell the difference.
I'm not sure this is the point, but I agree AI is dramatically improving, and I can see it taking over a lot routine advertising work, But, even the obviously good stuff, looks too perfect, and somehow fake to me, but then maybe I'm simply not identifying the really good stuff as being AI.

The selling point of "photographs" is that people "believe" them to be real, If you take the tourism example from Icelandair example there is an expectation that a real person pointed a real camera at a real view, and this is the view the tourist expects to see. Once you start generating them with AI they lose credibility, even if they are indistinguishable from the real thing. I've suggested it before, but I suspect that we might see organisation like Icelandair making a point of giving bylines to their images and links to the photographers website so potential customers can check out the authenticity and integrity of the photographer.

In many, maybe all, situations the value of using a photograph is because the viewer knows it;'s something made by a person who was there. Albeit possibly manipulated.

I can understand why the Icelandair research showed the majority of people being unhappy with the idea of AI generated images being used for advertising.
 
I'm not sure this is the point, but I agree AI is dramatically improving, and I can see it taking over a lot routine advertising work, But, even the obviously good stuff, looks too perfect, and somehow fake to me, but then maybe I'm simply not identifying the really good stuff as being AI.

The selling point of "photographs" is that people "believe" them to be real, If you take the tourism example from Icelandair example there is an expectation that a real person pointed a real camera at a real view, and this is the view the tourist expects to see. Once you start generating them with AI they lose credibility, even if they are indistinguishable from the real thing. I've suggested it before, but I suspect that we might see organisation like Icelandair making a point of giving bylines to their images and links to the photographers website so potential customers can check out the authenticity and integrity of the photographer.

In many, maybe all, situations the value of using a photograph is because the viewer knows it;'s something made by a person who was there. Albeit possibly manipulated.

I can understand why the Icelandair research showed the majority of people being unhappy with the idea of AI generated images being used for advertising.

My point was more about how would people reject it if they don't spot it?

Tourism etc, I can totally see the point - and indeed the validity - of including a byline that says images are genuine. Especially having been to Iceland, you'd think that the whole country is unreal.

Don't get me wrong, I'm mostly against it. Having been through several technological 'advancements' the long-term effect is never that it makes the job any easier or better.

When we used to do photoshoots on film, you'd shoot in the morning, the photographer would send shots off to be processed and take you out for a lovely boozy lunch. By the time that was all over, the shots would be printed and you could go back and view them.

My first digital shoot was a huge disappointment on that front. Shots appeared immediately on a screen and then we were done.

The unfortunate side-effect (IMO) is that when things are quick and easy to do, they lose value. Anyone thinks they can do them. People are seduced by shiny things. At present, I'd say AI is capable of getting you to a point which is 85% of the way. Unfortunately, for many people that's enough.

But to compare it back to photography, it's possible to make a photograph that is technically perfect, but dull. Conversely, it's possible to produce something out of focus and over exposed that captures the soul. If AI ever figures out what it is that makes the difference, we're all knackered.
 
Last edited:
My point was more about how would people reject it if they don't spot it?
If you don't spot it, there would be nothing to reject. Unless there were consequences that came from it being AI, but that would also apply to manipulated "real" photographs.

Tourism etc, I can totally see the point - and indeed the validity - of including a byline that says images are genuine. Especially having been to Iceland, you'd think that the whole country is unreal.

Don't get me wrong, I'm mostly against it. Having been through several technological 'advancements' the long-term effect is never that it makes the job any easier or better.
I'm not against AI, it’s a potentially valuable tool, but it's not photography and it doesn't produce photographs, because a photograph comes with some assumptions
When we used to do photoshoots on film, you'd shoot in the morning, the photographer would send shots off to be processed and take you out for a lovely boozy lunch. By the time that was all over, the shots would be printed and you could go back and view them.

My first digital shoot was a huge disappointment on that front. Shots appeared immediately on a screen and then we were done.

The unfortunate side-effect (IMO) is that when things are quick and easy to do, they lose value. Anyone thinks they can do them. People are seduced by shiny things. At present, I'd say AI is capable of getting you to a point which is 85% of the way. Unfortunately, for many people that's enough.
And the extra 20% to 25% is where the real work is done
But to compare it back to photography, it's possible to make a photograph that is technically perfect, but dull. Conversely, it's possible to produce something out of focus and over exposed that captures the soul. If AI ever figures out what it is that makes the difference, we're all knackered.
I know what you mean, but people make photographs, and to return to my last post, this is what makes photographs interesting, seeing how other people see the world.
 
I’ve started using AI to test different image styles and layouts before picking one to shoot properly—saves loads of time messing around with setups that don’t work.
Yes, it's used a lot for testing and creating mockups, but it can also create ideas that are impossible to recreate in real life.
 
At work it is being used to add backgrounds to images of real items, e.g. item is an Xmas themed plate, add in an “Xmas dinner table with a red tartan tablecloth, pictured from above”.

A year ago many of the images looked awful to my eye, but the package being used has come on a lot in a short time, I personally still prefer a clean mono background to let the product sell itself, but our B2B customers prefer to see the items on socials (means they see it today rather than when placing next months order, if they bother to check the new item pages rather than skipping as they are in a rush and know what they want to order) and it’s within a budget that is reasonable for the price points and margins involved.

To do a range of 12 new items it’s now 60 minutes of photography on a white background and 30 in the AI package.
Previously it would have been 3 hours procuring materials (buying or borrowing from home etc) and setting the scene up, 3 hours photography with the scene being rearranged to suit each item, possibly multiple times, and then into photoshop for tweeking and resizing etc.

There was also the trap in that the better the photographer (in house staff rather than contractors) the longer they seemed to take chasing perfection vs that being sufficient quality for the intended purpose - the “worst” was the lady with a photography related degree who would happily spend a day on a single item if not given very firm dead-lines…

As someone commented above it’s probably only 85% of the old way, but it’s sufficient quality for the modern purpose, which doesn’t include any print media which previously was the main driver for highest quality images in our line of work.
 
At work it is being used to add backgrounds to images of real items, e.g. item is an Xmas themed plate, add in an “Xmas dinner table with a red tartan tablecloth, pictured from above”.

A year ago many of the images looked awful to my eye, but the package being used has come on a lot in a short time, I personally still prefer a clean mono background to let the product sell itself, but our B2B customers prefer to see the items on socials (means they see it today rather than when placing next months order, if they bother to check the new item pages rather than skipping as they are in a rush and know what they want to order) and it’s within a budget that is reasonable for the price points and margins involved.

To do a range of 12 new items it’s now 60 minutes of photography on a white background and 30 in the AI package.
Previously it would have been 3 hours procuring materials (buying or borrowing from home etc) and setting the scene up, 3 hours photography with the scene being rearranged to suit each item, possibly multiple times, and then into photoshop for tweeking and resizing etc.

There was also the trap in that the better the photographer (in house staff rather than contractors) the longer they seemed to take chasing perfection vs that being sufficient quality for the intended purpose - the “worst” was the lady with a photography related degree who would happily spend a day on a single item if not given very firm dead-lines…

As someone commented above it’s probably only 85% of the old way, but it’s sufficient quality for the modern purpose, which doesn’t include any print media which previously was the main driver for highest quality images in our line of work.
It's all about the "fit for purpose" argument.
 
I've already said this, but the iterations are improving all the time - at an alarming rate.

Our internal AI at work was basically unusable maybe 18 months ago. It would put random hands and objects in the middle of images for no reason. You'd occasionally get something good, but it took a lot of goes. But now...

We recently produced an internal video for a client where we took a still of their CEO and animated it, cloned his voice and had him 'present' a two-minute video to their shareholders. The point of the video was about using AI in their business which is why it was all done that way.

It was cut with AI B-roll that was indistinguishable from the real thing. We added (fake) news reports and generated motion GFX. The last time I visited this thread, none of that was possible (at least not commercially).

The whole things was very slick. And did its job - pulling in an additional £35 Million of investment.

It's scary how easy it was really. And easy to see how it could be misused.

The only thing it couldn't do at that time was the music. But anyone that's on Instagram may well have seen the recent spate of modern songs re-imagined as 60's Soul (for example). They're amazing.

It's not art though.

As a professional user of AI, I understand why people like it. I don't like using it though - it feels like I'm a turkey that's been convinced Christmas sounds nice.

As an amateur photographer, it's the challenge I enjoy. I've seen those AI batch editing programmes for photos and I can totally see them appealing to commercial photographers who might be editing thousands of photos from one session. But on a personal level, I like going through them and doing it one by one.

My biggest issue and I keep telling people this at work... is that it doesn't actually save 'time' in the long term. I mean it does, but as creatives we don't get that time. The new, shorter, faster, immediate way to produce things becomes the norm. And then that's what clients expect.

Where creativity often happens is in the down time, when you're waiting for a set to be redressed, or you've sent a storyboard off to an illustrator to be drawn up. Having time and giving things 'the overnight test' as we used to call it is invaluable. The process took longer and meant that adjustment were made along the way. Now it's so quick to do things, you only get the chance to think of a 'better' idea when it's done and dusted.

There's a great story about PIXAR that illustrates it perfectly.

THE “BLANK WALL PRINCIPLE”
How Pixar Discovered Their Most Profitable Fix by Staring at an Empty Wall
In the early days of Pixar, right after Toy Story became a hit, the team faced a strange creative slowdown.

Every movie idea felt forced.
Scripts stalled.
Storyboards dragged.
Characters felt flat.
Meetings went nowhere.

One afternoon, Andrew Stanton walked into the studio and noticed something unusual.
The giant wall where Pixar usually pinned storyboards, the wall that once overflowed with sketches and color, was completely blank.

Not because they had no ideas.
But because every idea was being polished inside laptops. No one wanted to pin anything on the wall until it looked perfect.
The obsession with perfection stopped the entire process.

Stanton created a new rule:
If an idea exists, it goes on the wall. No excuses.
The next day the blank wall was covered with:
Bad sketches
Half-formed concepts
Awkward drawings
Ugly drafts
Jokes that made no sense
It looked chaotic, messy, and unprofessional.
But within weeks, that wall produced something legendary.
Finding Nemo.

One of Pixar’s greatest films and their first nine hundred million dollar blockbuster.
All because the team stopped hiding imperfect ideas and put them into the light where creativity could breathe.

1f4a1.png
THE MARKETING LESSON

Your business probably has a blank wall problem.
Ideas stuck in drafts
Funnels you are overthinking
Offers you never launch
Content you keep tweaking
Stories you never post
Videos sitting in your camera roll
Landing pages waiting for perfection
Perfection is procrastination in disguise.
Put it out.
Get it visible.
Let momentum improve it.

1f9e0.png
THE AUTHOR TAKEAWAY

If your writing is struggling, it is not because you lack ideas or creativity.
It is because you’re judging before letting it exist on the wall of inspiration. You can seek the help of others to unravel tangles or make leaps with you.
Share your experience with supportive creatives.
Refine the story in motion by always pushing forward.
Success comes from movement, not perfection.
So let it be messy. That’s how it comes alive!
 
Last edited:
this is pretty brutal. But guess what, major corps will just keep ramming it down our throats until we just accept it. There is nothing glorious in having to watch / scroll through the ads in the first place, so does it even matter that they get AI slop treatment? I bet the likes of netflix are already queuing up to release live on demand AI movies and it is all the talk in the tech industry right now.

I have to admit google pro model is getting scarily good for basic design jobs. I had to make a product themed Christmas card and poster for a client who seems all too happy with AI output in less than 20min. Chat - pathetic. Google pro - it did 100% and I pushed it deep and far, including all secondary changes. I used stylized / drawing style rather than photorealism very intentionally here. It is crazy how it can decode complex fonts and rewrite text in that style, move elements, readjust composition for different aspect ratio all by simple text commands. It wasn't anything a painter and graphics designer couldn't do (and better), but 20min instead of 2 days for a short and basic application.
 
I don't object to it if I can't see the difference.

I don't like the articles like "This battery uses plain salt and can power a house for 50 years" and other totally ridiculous things that you see, complete with detailed technical drawings and text.
Most people on here would recognise it as rubbish, but a good many would not and the same would go for adverts.
It could be used to mislead or cheat people.

Very worrying is the use of AI for writing things for degrees etc!
 
I don't object to it if I can't see the difference.

I don't like the articles like "This battery uses plain salt and can power a house for 50 years" and other totally ridiculous things that you see, complete with detailed technical drawings and text.
Most people on here would recognise it as rubbish, but a good many would not and the same would go for adverts.
It could be used to mislead or cheat people.

Very worrying is the use of AI for writing things for degrees etc!
If the Google AI search summaries are anything to go by, the most worrying thing for degree students will be low grades.
 
Just a thought...

I wonder how many people, up to the modern period, would look at a painting or drawing and think that it was an accurate reflection of reality?
 
I don't object to it if I can't see the difference.

I don't like the articles like "This battery uses plain salt and can power a house for 50 years" and other totally ridiculous things that you see, complete with detailed technical drawings and text.
Most people on here would recognise it as rubbish

Salt-based sodium-ion batteries are a real thing that are already in commercial production, and they have a very long service life compared to Li-ion cells

Maintaining 70% of their original charge capacity at 20,000 cycles is not an unrealistic proposition, which translates to about 54 years if you are recharging them daily for domestic or grid energy storage.
 
Just a thought...

I wonder how many people, up to the modern period, would look at a painting or drawing and think that it was an accurate reflection of reality?

I would suspect very few - there is a clear gulf between a painting and the real world, though with some there is a greater sense of reality than others.

On a different note, the French supermarket chain Intermarche had a 2.5 minute hand-drawn cartoon advert made this Christmas.

View: https://youtu.be/Na9VmMNJvsA?
 
Just a thought...

I wonder how many people, up to the modern period, would look at a painting or drawing and think that it was an accurate reflection of reality?
An important job of portrait painters during the Renaissance period was to travel the world painting portraits of prospective royal brides and bring them back to the King so he could decide if he approved of their appearance before agreeing to a marriage.



I'm also fairly sure there are other examples I've heard of where people accepted paintings and drawings as being real (proof), when they were fictional. e.g. fantastical beasts from distant lands as well as "savage" customs and behaviours of people from distant lands and cultures.

I fear most people believe what they want to believe, regardless of how unreliable the evidence might be. Or is that me being too cynical :-(
 
Salt-based sodium-ion batteries are a real thing that are already in commercial production, and they have a very long service life compared to Li-ion cells

Maintaining 70% of their original charge capacity at 20,000 cycles is not an unrealistic proposition, which translates to about 54 years if you are recharging them daily for domestic or grid energy storage.
I thought they were still under development to make them useable for most applications.

But I still don't believe one the size of small microwave will power a house :)
 
Or is that me being too cynical :-(
Not in the least.

There are several instances of both brides and grooms being returned, when the reality didn't live up to advertising. There have even been conflicts caused by paintings, which presented certain claims that were strongly disputed by others present at the event illustrated.
 
Not in the least.

There are several instances of both brides and grooms being returned, when the reality didn't live up to advertising. There have even been conflicts caused by paintings, which presented certain claims that were strongly disputed by others present at the event illustrated.

Anne of Cleeves is a fairly well known example IIRC. The problem there was that the painter was meant to paint a likeness. :p
 
Anne of Cleeves is a fairly well known example.
Of course, unless someone disinters her and creates a face over her skull, we can never really know if Holbein overegged the pudding or Henry just wanted a different type of look.

He was pretty fickle when it came to women and Anne must have had something, because she outlived all her sister wives!
 
He was pretty fickle when it came to women and Anne must have had something, because she outlived all her sister wives!

She came from a powerful family that he didn't want to fall out with. She had no supporters here, didn't get pregnant, and so it was simpler just to divorce her and tuck her away so he could chase some tail.
 
I'm not against AI, I use it quite a lot, but we all need to understand that it's a useful tool that is currently far from perfect.
What is is currently extremely good at is producing something that seems to be a good result, with almost zero effort, and instantaneously, but with very severe (current) limitations.

Image creation is just one example, the internet is flooded with poor AI images and videos, things will get better and, as businesses generally are far more interested in cost and convenience than with quality and truth, AI is bound to almost completely take over soon, because some of the images are already up to the poor standards of the lower end of the pro photography professionals.

On the text side, AI always seems to work out what we want to hear and reinforces our opinion rather than challenge it, which is a major fault. And it is often factually wrong too, so every answer needs to be manually checked for accuracy. A solicitor friend tells me that he uses it a lot, especially for writing or refining skeleton arguments, it generally does a good job but nearly always quotes case law wrongly, and he needs to manually look up all of the references and check them.

And, of course, it doesn't have any actual intelligence, it just finds whatever has already been published and accepts that to be factual. I know someone who uses (or used) AI to create articles on photographic lighting, and the results are terrible. For example, it came out with the usual bland, vague and useless statements about lighting with 1, 2 or 3 lights, when it came to using 4, it said that 1 light goes into each room corner, all pointing at the subject, which of course is laughable nonsense, which it found somewhere on the net and accepted to be true.

AI is of course also now used by many to write sales listings, but always gets it very wrong, it's a lazy approach that simply doesn't work - yet. It also creates SEO, which is why a lot of product titles make no sense at all to the potential buyer.
 
Back
Top