In her latest column, ad tech expert Shirley Marschall takes a look at the old promise of perfect targeting, and why it really needs to be done better...

Digital advertising has always chased the promise of perfect targeting. First came cookies: easy to drop, hard to explain, and eventually impossible to justify. Then came universal IDs: hashed emails, first-party graphs, interoperable identity layers.

Most of them simply tried to rebuild the same flawed system. New plumbing, same logic. Stitching together every last signal in hopes regulators wouldn’t notice.

They noticed.

Legal challenges are piling up. Two lawsuits accuse The Trade Desk of secretly violating privacy laws. LiveRamp could be next. Regulatory pressure is growing. Consumer trust is shrinking. First-party data? Scarce. Consent? Questionable. Relevance? Still elusive.

The rise of AI

Just as the old system starts to collapse, a shiny new promise takes the stage:

AI-powered hyper-targeting. Hyper-personalisation. Hyper-everything.

The message? "Don’t worry! We can predict behaviour better than ever."

And what should you do when someone tells you not to worry? Start worrying. Because if the last 20 years have taught us anything, it’s this: The data is failing us. We’re optimising for an illusion. An identity that’s stitched, guessed, inferred. Not understood.

So what if…this time, before we hit accelerate, we hit pause? If identity-based targeting got us here, maybe identity was never the answer.

From identity to intention

People don’t behave like segments. They behave like… people. 

Some are visual, others auditory. Some prefer summaries. Others want depth. The same person even behaves differently depending on the moment:

On a crowded train? No sound. Late at night with headphones? Bring on the audio. Multitasking on a laptop? Summarised and skimmable, please. Lean-back TV mode? Full visual experience.

These aren’t quirks. They’re behavioural signals, often more stable than any identity graph.

So why are we still targeting static profiles when context is right in front of us?

Don’t personalise the person. Personalise the experience.

Personalisation used to mean tracking people across the web, collecting behavioural breadcrumbs, and serving ads they never asked for. We confused recognition with relevance.

But just because you know someone’s name doesn’t mean you know how they want to be approached. And most "personalised" ads? Still feel generic.

What people want isn’t creepy accuracy. It’s comfort, control, and context.

So instead of asking: Right person, right message?

Ask: Right format, right moment, right context?

Call it format-aware personalisation:

• News reader? Skip the autoplay, offer a summary.
• Headphones on? Serve up audio.
• Scrolling LinkedIn at the airport? Captions on, sound off.
• Cooking dinner? Try a podcast-style ad, not a banner.

Same message. Different experience.

This isn’t identity-driven personalisation. It’s situational relevance, adapting to how people prefer to engage in the moment.

From ID Tech to UX Tech

Ad tech has optimised for delivery, tracking, and attribution, but not for experience.

That’s the gap.

People don’t care if an ad "knows them." They care if it fits their moment - their screen, their setting, their headspace. This isn’t hypothetical. It’s already happening:

• Subtitles on by default.
• Podcasts at 1.5x speed.
• "Listen" buttons in newsletters.
• YouTube transcripts summarised by AI.

People aren’t avoiding content. They’re shaping it to fit their flow. And if an ad doesn’t fit that flow, it gets ignored. It’s not about getting in their face. It’s about fitting into their frame.

Smart ads know when to shut up

This is where GenAI flips the script - not by knowing who we are, but by understanding what we need, when we need it. Forget static segments. Forget stitched-together ID graphs.

What matters now is:

• Device behaviour (TV, mobile, desktop).
• Contextual cues (location, time of day, sound settings).
• Format preferences (text, audio, summary, full-length).

These signals are real-time, fluid, and user-chosen… not scraped, not stitched, and that changes the rules.

Because AI gives people the power to redirect, compress, or even outsource their attention entirely. So ads can’t just interrupt anymore. They have to adapt.

Quiet formats in public spaces. Rich formats in lean-back moments. Skimmable formats when time is tight. And sometimes, the smartest move isn’t to speak louder. It’s to know when not to speak at all.

Design for Preference. Deliver for the Moment.

What if we stopped defaulting to a single format and gave people choice - based on the moment, not the media plan?

• "Read" or "Listen" buttons.
• Short or long-form options.
• Audio, video, or text.

And design ad formats that flex:

• A podcast ad becomes a visual carousel for users who mute audio.
• A product demo turns into an AI-narrated summary for commuters.
• A long-form explainer shrinks into snackable highlights on mobile, then expands to an interactive deep dive on desktop.

Same message. Smarter delivery.

Let the startups sort the stack

No, we don’t need to reinvent the stack overnight.

That’s what the next generation of startups are doing…right?

• Format-flexible ad platforms.
• Context-aware personalisation layers.
• Privacy-first signals that adapt in real time.

Your job isn’t to fix the plumbing.

It’s to change the direction of the pipes.

The Takeaway

This isn’t anti-personalisation. It’s post-identity personalisation.

It’s not about targeting people. It’s about targeting moments.

Let the user be the user. Let the ad meet them there.

No cookies. No ID graphs.

Just better advertising.