Online advertising in the AI era

2025 should be viewed alongside 1999 (the web) and 2008 (mobile).

Not because everything suddenly changed overnight, but because the underlying rules did.

What was impractical a year ago is now possible. What used to require human judgment is increasingly automated. And the gap between brands with strong foundations and those without is widening fast.

Generative AI sits at the top of every CEO’s agenda for good reason. Just as every business needed an internet strategy in 1999, every brand now needs a clear view on how AI affects how they acquire and retain customers.

So the real question for brand leaders isn’t whether AI matters.

It’s where it actually changes outcomes.

What hasn’t changed

Unlike previous platform shifts, AI is not disrupting where consumers spend their time online.

Google (Search and YouTube), Meta (Facebook, Instagram, WhatsApp), TikTok, Amazon, Netflix and X still account for the vast majority of attention. If anything, AI is entrenching their positions by:

• Optimising content relevance at the individual level
• Making it faster and cheaper to produce high-quality content at scale

ChatGPT is the breakout consumer product of this wave, but even that is likely years away from overtaking Google Search. And there are no serious AI-native challengers to YouTube, Instagram or TikTok as entertainment platforms.

For brands, this means something counterintuitive:

There are fewer viable advertising channels, not more.

Big tech’s ad revenues reflect this reality. And brands that want access to consumer attention will continue to spend heavily on Google Ads, Meta Ads and TikTok Ads.

What has changed

What AI has transformed is how ads are prioritised and served behind the scenes.

Instead of marketers choosing audiences and formats based on intuition, increasingly powerful algorithms now make those decisions in real time. In most cases, they outperform humans.

That’s why budgets keep flowing into Performance Max and Advantage Plus campaigns.

There’s also a less generous interpretation. By removing granular audience controls, platforms encourage broader spend across more inventory, while still delivering “acceptable” ROAS. Hyper-targeting a small group of high-intent users may benefit brands, but it doesn’t always maximise platform revenue.

Either way, the direction of travel is clear:

brands have less control over targeting logic, not more.

So how do you improve acquisition when your choice of platform is constrained and your ability to steer the algorithm is shrinking?

Three levers brands still control

There are three ways AI can materially improve advertising performance today. Littledata helps with two of them.

1. Creative quality and velocity

AI makes it easier to test more variations, learn faster, and tailor creative to different products and contexts. This is already table stakes.

2. Signal richness

What large language models have shown is that intelligence improves dramatically as you expand the number of inputs they can learn from.

A ChatGPT-style model may work with hundreds of billions of dimensions. By contrast, a basic Meta Pixel pageview contains fewer than a dozen.

Ad platforms know this. That’s why Meta keeps pushing more data through Conversions API and Enhanced Conversions. The models can now handle far more context than they could a few years ago.

This is where Littledata’s product direction is focused.

We already send accurate Shopify revenue and on-site behaviour server-side. Increasingly, we’re adding richer touchpoints such as subscriptions, loyalty tiers, and repeat purchase behaviour, so ad platforms have a better chance of serving the right ad to the right person.

Today, platforms like Meta still limit how many standard dimensions brands can send. But pilots with larger advertisers suggest this will expand. One existing but underused dimension is predicted lifetime value, inferred from storefront behaviour and purchase history.

As models get better, dimension widening becomes the real advantage.

3. Event-based audience definition

The third lever is how you define who gets retargeted.

For years, brands have relied on simple events like “viewed product” or “added to cart.” The problem is that many of these users are bots or low-intent traffic that will never convert.

By defining stronger mid-funnel signals and filtering out noise, brands can dramatically improve retargeting and lookalike performance.

Examples include:

• Users who viewed the same product multiple times across sessions
• Add-to-cart users who are not already captured via email or loyalty
• Purchases above a minimum revenue threshold
• High-value customers in specific loyalty tiers

These event-based audiences are different from list-based audiences synced from tools like Klaviyo. Lists are limited to known customers and typically suffer from ~50% match rates against ad platform device graphs.

Cookie-based audiences, especially when enriched with first-party identity data from Shopify, are both broader and more flexible.

Littledata’s Event Customizer is designed for exactly this: allowing brands to create higher-quality signals and synthetic events that better reflect real buying intent.

The real opportunity

AI won’t magically create new routes to your customers.

What it will do is reward brands that feed existing channels with cleaner, richer, more meaningful signals.

The winners won’t be the brands chasing the latest AI tool.

They’ll be the ones with the strongest data foundations underneath their marketing.

At Littledata, our job is to make sending high-quality Shopify data into those channels simple, stable, and server-side by default.

We’re riding the AI wave with a solid board of first-party customer data.

That’s the only way to stay upright as the current gets stronger.