The Hypothesis

Can an AI generate platform-native ad creative across Google, Meta, and TikTok from a single plain-English brief — and keep it coherent with the landing page?

The Concept

Most optimisation stacks focus on what happens after the click. But the thing that generates the click — the ad itself — is still largely manual. Marketers write Google Ads headlines, craft Meta copy, and script TikTok hooks in isolation from the landing page they send traffic to.

This experiment closes that gap.

Describe the campaign in plain English and the system generates a full set of ad variants — formatted natively for each platform, labelled with hypotheses, and message-matched to the landing page being served.

When the landing page evolves, the ad creative evolves with it. The pre-click and post-click experience become a single system instead of two disconnected workflows.

Ad creative and copy generator

The hypothesis

Can an AI generate platform-native ad creative across Google, Meta, and TikTok from a single plain-English brief — and keep it coherent with the landing page?

The concept

Most optimisation stacks focus on what happens after the click. But the thing that generates the click — the ad itself — is still largely manual. Marketers write Google Ads headlines, craft Meta copy, and script TikTok hooks in isolation from the landing page they send traffic to.

This experiment closes that gap.

Describe the campaign in plain English and the system generates a full set of ad variants — formatted natively for each platform, labelled with hypotheses, and message-matched to the landing page being served.

When the landing page evolves, the ad creative evolves with it. The pre-click and post-click experience become a single system instead of two disconnected workflows.

How it works

1. Write a plain-English brief — describing the campaign objective
2. Add persona context — brand voice and product signals
3. Generate platform-native variants — for each ad channel
4. Label each variant — with a clear testing hypothesis
5. Match ad messaging — to the landing page being served
6. Feed performance back — into the optimisation loop

What it explores

• Can a single brief generate effective creative across fundamentally different platforms?
• Does platform-native formatting hold up for Google, Meta, and TikTok?
• Can ads and pages co-evolve without human coordination?
• Do hypothesis-labelled variants deliver more actionable insights than traditional A/B testing?

What we found

• Message-matched campaigns outperformed mismatched ones by 23% in conversion rate
• Platform-native formatting was reliable for Google and Meta, consistently meeting character limits and structural requirements
• TikTok hook scripts required human review to capture platform tone and cultural nuance
• Co-evolution model worked well for copy-led landing page changes but failed when the page shifted visual direction — the ad system couldn’t adapt imagery without human input

Learnings

• Pre-click and post-click alignment is a conversion lever — when the ad promise doesn’t match the landing page, users bounce.

• Platform-native formatting is solvable but requires ongoing maintenance as ad platforms regularly change formats, limits, and specifications.

• Co-evolution works best when ad and page optimise for the same business outcome — separate channel KPIs cause drift.

• TikTok is a cultural problem, not a formatting one — scripted hooks feel inauthentic without human polish, and may need a fundamentally different approach.

• A single-brief approach compresses creative production time but requires strong persona and brand context to avoid generic output.

Where it goes next

The message-match finding is strong enough to build on. We’re testing whether the co-evolution model can handle visual changes — not just copy — by connecting to a tagged asset library. The TikTok gap is worth exploring separately: what makes a hook feel native to that platform, and can that be systemised?