It has never been easier to make an ad.
Write the copy in ChatGPT. Build the artwork in Canva. Load it into Meta or Google. Let the platform choose audiences, placements and bids.
A job that once needed several people can now be underway before lunch.
That is genuinely useful.
It is also where the problem starts.
AI has dramatically reduced the friction involved in doing marketing. It has not removed the need to decide what the marketing should actually achieve.
And when those two things get confused, businesses can end up producing more creative, launching more campaigns and automating more decisions without becoming any clearer on whether the activity is making them money.
AI is not necessarily making advertising worse.
It is making weak marketing easier to scale.
The platforms are getting better at execution
There is no sensible argument for pretending the technology does not work.
Google Performance Max now automates bidding, budget optimisation, audience discovery, creative combinations and delivery across Google's advertising inventory. Google specifically recommends providing strong conversion data, high-quality creative and audience signals to help the system optimise effectively. Google Help
Meta is moving in the same direction. Advantage+ tools automate elements such as placements and creative optimisation, with the platform encouraging advertisers to supply creative appropriate for different environments such as Reels. Facebook
This automation can do things humans are simply not capable of doing manually at the same speed.
It can evaluate huge numbers of signals, adjust bids auction by auction, test creative combinations and find buyers outside the audiences you might have chosen yourself.
That part is not the problem.
The problem is assuming the machine also understands the economics sitting behind the campaign.
It doesn't have your P&L open in another tab.
AI will optimise the target you give it
Suppose your campaign is optimising for purchases.
The platform can become very good at finding people likely to purchase.
But there are still questions sitting outside that objective.
Which products did they buy?
What margin did those products carry?
Was a discount required?
Was this a new customer or someone already looking for you?
Will that customer ever purchase again?
Did fulfilment costs wipe out most of the remaining contribution?
Are you acquiring customers you actually want more of?
Unless your measurement and campaign setup give the platform useful signals about value, it cannot magically infer all of that from the fact that a transaction occurred.
This is why a campaign can look increasingly sophisticated inside the advertising account while the commercial result remains ordinary.
A 4x ROAS can still be a poor result
Take a deliberately simple example.
Two groups of products each produce $10,000 in attributed revenue from $2,500 in advertising spend.
Both report a 4x ROAS.
Looks equal.
But Product Group A has a 25% gross margin.
$10,000 revenue creates $2,500 of gross margin. After the $2,500 advertising cost, there is nothing left before fulfilment, transaction fees or fixed overheads.
Product Group B has a 50% gross margin.
The same $10,000 creates $5,000 of gross margin. After advertising spend, $2,500 remains before those other costs.
Same dashboard result.
Very different business result.
This is why Ledger keeps coming back to a fairly unsexy point:
Revenue tells you what sold. Margin tells you whether it was worth selling.
ROAS is useful. It just isn't your profit and loss statement.
Cheap creative has another hidden cost
Generative AI has also made producing advertising creative incredibly cheap.
Again, that is useful.
A small retailer no longer needs a large production budget to test multiple messages, layouts or product angles.
But cheaper production also makes it easier to publish work that has never really been thought through.
CreativeX analysed almost 900,000 digital ads and found substantial media spend being placed behind creative that did not follow basic digital suitability principles. Its research found 35% of ads lacked appropriate branding, 50% lacked a call to action and 12% used the wrong dimensions. CreativeX extrapolated its broader findings against projected 2023 digital spend and estimated $233.4 billion could be at risk from digitally unsuitable advertising. That number is an extrapolation, not a measured 2026 loss, so it should be treated accordingly. CreativeX
The useful lesson is simpler.
Making ten ads instead of two is not an advantage if all ten are generic, poorly branded or built around an offer nobody particularly wants.
More output is not the same thing as better marketing.
This matters more as Australian advertising spend grows
Australian businesses are putting more money into digital media, not less.
IAB Australia reported that internet advertising reached $19.8 billion in FY26, up 14% year on year. Search accounted for $8.6 billion and video reached $5.9 billion. IAB Australia
At the same time, AI is moving into everyday Australian business use. National AI Centre data reported that 43% of Australian SMEs had adopted AI to some degree across the December 2025 to February 2026 quarter. That figure covers business AI adoption broadly rather than marketing specifically, but the direction is clear: these tools are becoming normal operating equipment. National AI Centre
So the question is no longer whether businesses will use automation.
They will.
The better question is what happens before the automation gets switched on.
Strategy still comes before tactics
Mark Ritson's marketing planning model puts diagnosis first, strategy second and tactics third.
That order becomes more important as tactics become easier to execute.
Before worrying about another AI tool, campaign type or automatic setting, an owner should be able to answer a few much more boring questions:
What are we actually trying to sell?
Where does the margin sit?
Which customers are valuable?
What buying situation are we trying to influence?
Why would someone choose us?
What happens after the first purchase?
What are we measuring as success?
Only then does the platform become the execution layer.
Ritson's wider point is that tactical capability should follow diagnosis and strategy rather than substitute for them. Marketing Week
That is the distinction AI has made easier to forget.
Do not overcorrect and start micromanaging everything
There is an equally bad conclusion available here:
“AI isn't strategic, therefore I should manually control every audience, placement and bid.”
Not necessarily.
Broad targeting and machine-led delivery can expose your brand to buyers you may never have selected manually.
For many businesses, the algorithm probably is better at finding likely converters than someone spending Friday afternoon building another interest audience.
The point is not to fight the machine.
It is to give the machine a better job.
Good customer data can improve audience signals.
Good creative gives the algorithm something worth distributing.
Accurate conversion tracking tells it what happened.
Useful conversion values help distinguish outcomes.
Strong branding makes sure the customer remembers who actually ran the ad.
And commercial reporting determines whether the activity deserves more investment.
Distinctive brand assets matter here too. The Ehrenberg-Bass Institute describes colours, logos, characters, fonts and other distinctive assets as long-term contributors to mental and physical availability. AI-generated creative may speed up production, but the business still has to ensure its own distinctive assets remain present and recognisable. Marketing Science
Before you increase the budget, check the inputs
If you are already using Meta or Google automation, you do not need to rip everything apart.
Start by checking five things.
1. The business objective
Are you trying to maximise any sale, acquire new customers, sell a priority category, move stock or generate repeat purchase?
Those are not the same job.
2. The conversion signal
Is the platform optimising toward something that genuinely matters, or have low-value actions accidentally become campaign goals?
3. The customer data
Can you distinguish a valuable customer from a cheap conversion?
4. The creative
Is the ad distinctive, suited to the platform and clear about why someone should care?
5. The commercial result
Look past revenue and ROAS. Review available margin, CAC, product mix, new versus returning customers and repeat behaviour.
You do not need perfect data before advertising.
You do need to know what the data cannot tell you.
What AI can — and cannot — tell you
AI advertising platforms can become extraordinarily good at predicting behaviour within the information available to them.
That is valuable.
But an attribution dashboard cannot always tell you whether the advertising created genuinely incremental demand.
A purchase recorded after an ad interaction does not automatically prove the sale would never have happened otherwise.
Likewise, a high ROAS does not automatically prove good profitability, and a low CPA does not automatically mean you acquired a valuable customer.
Those questions require wider business data, sensible testing and some human judgement.
Which brings us back to the main point.
AI has made the tactical part of advertising easier.
Good, Use it.
Just do not outsource the commercial thinking along with it.
Automate the execution. Keep the judgement.
If your Google or Meta campaigns are producing plenty of numbers but you are still unsure what the spend is contributing commercially, Ledger's free marketing audit looks across the ads, customer journey, tracking and reporting to identify where money may be leaking and what is worth fixing first.

