The Algorithm Is the Media Buyer Now. Creative Is How You Talk to It

For years, paid advertising rewarded people who knew where to find the right buttons.

A good media buyer could spend hours inside an ad account building audiences. Interests were layered. Lookalikes were tested at different percentages. Age brackets were split into separate ad sets. Exclusions were carefully maintained.

The creative mattered, of course. But targeting was where much of the technical advantage lived.

That advantage is disappearing.

Google, Meta and TikTok have spent the past few years steadily taking audience selection out of advertisers’ hands. Not completely, and not in every campaign, but enough to change what good performance marketing looks like.

The platforms now want fewer restrictions and more data. Give the system a goal, supply enough creative and let its models find the people most likely to respond.

This has created a slightly uncomfortable situation for performance marketers.

The machines are getting better at the part of the job we used to call media buying.

The part they still need help with is the advertising.

Meta doesn’t need your 17-interest ad set

Meta has been fairly open about where it wants advertisers to go.

Its Advantage+ audience guidance describes traditional audience inputs such as interests, gender and custom audiences as “suggestions”. The system can look beyond those suggestions when it believes doing so will improve performance.

Meta’s broader Performance 5 framework recommends account simplification, automation and creative diversification.

There is a reason for this.

Behind Meta’s ad delivery system sits technology designed to make sense of a vast and growing pool of ads. In 2024, Meta published details of Andromeda, its personalised ad retrieval engine.

Andromeda was built to select relevant ads from an enormous number of possible candidates. Meta specifically pointed to the explosion in creative volume caused by automated advertising tools and generative AI as one of the challenges the system needed to solve.

That is an important clue about modern advertising.

Meta is not simply asking, “Does this person belong to the audience the advertiser selected?”

Its systems are trying to understand the relationship between a person and an ad.

Those are two very different ways to buy media.

Consider a mortgage broker:

One ad talks about buying your first home with a small deposit. Another focuses on remortgaging after a fixed rate ends. A third speaks to self-employed business owners struggling to prove their income.

The advertiser could spend days trying to manually construct three perfect audiences.

Or the ads themselves can attract different patterns of attention, clicks and conversions, giving the delivery system information about who responds to each message.

The creative becomes a signal.

Google is reading more than your keywords

Search advertising used to feel like the exception.

Someone types a keyword. You bid on the keyword. Your ad appears.

Simple enough.

Google’s AI products are making that description increasingly incomplete.

With AI Max for Search campaigns, Google says its search-term matching technology can use an advertiser’s keywords, creative assets, and landing page content to identify relevant searches.

Read that again.

Your creative and website are helping Google’s systems decide where your advertising may be relevant.

The keyword list has not vanished. But it is no longer the only language advertisers use to communicate intent to the platform.

TikTok is following a similar path. Its Smart+ targeting documentation explains that automatic targeting lets the system identify and optimise towards audiences expected to perform best. TikTok also warns that switching to more restrictive custom targeting may limit delivery.

Across three very different advertising businesses, the pattern is remarkably similar.

Give the algorithm room.

Give it a conversion goal.

Give it better creative.

This changes what “bad creative” means

A forgettable ad has always been expensive. Now it may be expensive for another reason.

Imagine an advertiser uploads six videos. Every video uses the same spokesperson, makes the same broad promise and opens with a variation of “Are you ready to transform your business?”

Technically, there are six ads.

Strategically, there is one idea.

The platform has very little variety to work with.

Compare that with six ads built around six distinct reasons someone might buy.

One speaks to price. One demonstrates the product. One answers a common objection. One tells a customer’s story. One compares the old way with the new way. One focuses on a specific situation where the problem becomes urgent.

Those ads do not simply look different; they create different opportunities for relevance.

This is why creative testing needs to move beyond changing headlines, button colours and opening captions. A proper creative test should ask whether a completely different message can find a completely different pocket of demand.

That is a much bigger question than, “Should the text be yellow or white?”

More creative does not mean more AI slop

There is an obvious problem with all of this.

The platforms want creative volume at exactly the same time generative AI has made it incredibly easy to produce creative volume. That combination could result in a mountain of very average advertising.

Fifty variations of the same generic concept are not a creative strategy. They are clutter.

The competitive advantage is likely to come from finding more genuine reasons for customers to care.

Talk to the sales team and find the objection that appears on every second call.

Read customer reviews and notice the phrase buyers keep repeating.

Watch someone use the product for the first time.

Ask why a customer bought today instead of six months ago.

Find the detail your competitors have ignored because it is too specific for their brand guidelines.

Then make an ad about that.

Automation makes this kind of work more important, not less. When everyone has access to similar campaign tools, similar bidding systems and increasingly similar AI production software, the unusual insight becomes more valuable. The media buying advantage gets commoditised, and the customer insight does not.

The best ad might look like the cheapest one you made

Advertisers also need to reconsider what “good” creative looks like.

A polished commercial can still work brilliantly. There is no universal law saying phone-shot videos beat professional production.

But platform-native creative has a structural advantage when it fits the way people actually consume content.

TikTok’s own creative guidance and automation tools are built around testing multiple videos and responding to creative fatigue. Meta similarly encourages advertisers to diversify creative so its systems can match relevant messages with different people.

This explains why a founder talking into an iPhone can sometimes outperform a campaign that required a lighting crew and three rounds of storyboards.

The phone is not the secret.

The specificity is:

“This serum is amazing for your skin” tells us almost nothing.

“I started using this because foundation kept separating around my nose by 2pm” gives us a person, a problem and a moment.

A human understands the difference immediately.

Increasingly, an advertising system can learn from the difference too.

Let’ look at what this means for paid ads performance and targeting today vs yesterday:

The shift is subtle, but it changes where advertisers should spend their time.

The new media buying skill is feeding the machine better information

There will still be talented media buyers in five years.

But their value is unlikely to come from knowing an obscure targeting trick hidden three menus deep in Ads Manager.

The platforms have every incentive to automate those tricks.

The valuable marketer will understand what information the algorithm needs and how to turn customer insight into distinct creative concepts.

That means looking at creative as a portfolio:

• Do we have an ad for the customer who cares about price?
• Do we have one for the person who does not believe the product works?
• Do we show the product being used?
• Have we explained why the problem happens?
• Is there an ad for someone who has never heard of us?
• Are all our videos effectively saying the same thing with different people holding the product?

These are targeting questions now.

Not because age, location, customer lists or audience controls have stopped mattering. They haven’t.

But the centre of gravity has moved.

In the old model, the media buyer chose a group of people and decided what message to show them.

In the emerging model, you give the platform a collection of messages and its systems learn which people respond to each one.

The audience is no longer defined only in a settings panel.

It is discovered through the ad.

That makes the creative department part of media buying. It makes customer research part of targeting. And it makes every bland, interchangeable ad a bigger performance problem than it used to be.

The algorithm can find your customer.

The harder question is whether you have made anything that customer cares about.

rohit singh

See Full Bio