The big idea

A platform built ON generative models, not just serving them

Generating one nice-looking ad image is a solved problem — plug into any image model. What's genuinely hard is everything AROUND that: generating dozens of testable variants, making absolutely sure none of them violates ad-platform policy (a violation can get the whole account suspended), tracking commercial usage rights, and actually learning from real performance data instead of guessing.

We'll build the product layer: programmatic variant generation, a fail-closed compliance gate, rights tracking, dynamic budget allocation across variants, and a real performance feedback loop.

How to read this: Each step opens with a real design decision — make the call before I show you what ships. Watch the diagram grow, hover the boxes, and at the end kill the compliance gate and flag a winning-but-manipulative variant to see the safety nets engage. Hit Begin.

Step 1 · The baseline

One hand-designed creative per campaign

Simplest version: a designer manually produces one ad creative per campaign. What's the real limitation?

Design decision: One manually-designed creative per campaign. What's the core limitation for a real advertiser?

The call: It can't be tested against alternatives at any real scale — there's no way to discover which headline, image, or CTA actually performs best without generating and comparing many variants. — Advertisers fundamentally need to TEST — run multiple variants against real audiences and let actual performance data reveal what works, which a single hand-made creative structurally can't provide.

A single hand-designed creative is an untested guess — real ad performance depends heavily on wording, imagery, and CTA choices that are hard to predict without actually testing alternatives against real audiences. The platform's value is enabling that testing at scale.

Advertising is fundamentally an empirical, testable domain: Unlike some creative work where "good" is mostly subjective, ad performance is directly measurable (clicks, conversions) — which means the real opportunity is generating enough tested alternatives to let data, not intuition alone, determine what actually works.

Step 2 · Programmatic variant generation

Many variants, not one "best" guess

Should the platform try to generate the single best creative from a brief, or something else entirely?

Design decision: Given a creative brief, should the platform aim to generate ONE optimal creative, or many?

The call: Many — programmatically generate a range of variants (different headlines, images, CTAs) from one brief, specifically to enable real A/B testing across combinations. — Generating many genuinely different variants gives advertisers real alternatives to test against actual audiences — the platform's job isn't to guess the single winner, it's to produce a good pool of candidates and let real performance data determine which one(s) actually win.

Build a Variant Generator that programmatically produces many creative variants from a single brief — different headlines, images, and calls-to-action — explicitly to enable real testing, not to guess a single "best" answer up front.

Generate for testing, not for a single verdict: When ground truth (what actually converts) can only be determined empirically, the generation system's job is producing a good, diverse CANDIDATE POOL — the determination of "best" belongs to real-world testing, not the generation step itself.

Step 3 · The compliance gate fails closed

One violation can suspend the whole account

Every generated variant needs to be checked against brand guidelines and ad-platform policy before going live. If that check is temporarily unavailable, should variants publish anyway?

Design decision: The compliance check is briefly unavailable. Should variant publishing continue without it?

The call: No — the platform should fail closed: no new variant reaches a live ad platform while the compliance gate is unavailable, even though generation itself can continue. — Given the asymmetric stakes (a brief delay vs. a potential account suspension affecting every campaign that account runs), blocking publication until the gate is healthy again is the only defensible default — this is the same fail-closed principle used for spam filters and model-eval gates elsewhere, applied here because the downside of failing open is severe.

Make the Brand-Safety Gate fail closed: while unavailable, generation and drafting continue, but no variant can be published to a live ad platform. The stakes — a single violation risking suspension of the entire ad account — make this the only defensible default, regardless of the inconvenience of a delay.

Match the failure mode to the asymmetry of the stakes: The same fail-closed principle appears across very different domains in this series (spam filtering, model promotion) precisely because the underlying reasoning is domain-independent: when the cost of failing open vastly exceeds the cost of a brief delay, fail closed.

Step 4 · Rights and commercial licensing

Generated doesn't automatically mean usable

A generated image might resemble a real person, use elements that echo a trademark, or come from a model whose outputs have ambiguous commercial usage rights. Is passing brand/policy checks alone sufficient to publish?

Design decision: A variant passes brand-safety and platform-policy checks. Is that alone sufficient to clear it for a paid ad?

The call: No — separately track whether each generated asset is licensed for commercial ad use, and flag likeness/trademark risk (a generated face too closely resembling a real, identifiable person; imagery echoing a protected trademark) before it can be published. — Rights and licensing is a genuinely distinct concern from content-policy compliance — a perfectly brand-appropriate, policy-compliant image can still carry real legal risk if it wasn't actually licensed for commercial use, or if it incidentally resembles a real person's likeness without consent.

Maintain a Rights & Licensing tracker separate from content-policy checks: it verifies a generated asset is actually licensed for commercial ad use and flags likeness or trademark risk before publishing — a distinct legal concern from whether the content is appropriately worded or policy-compliant.

Content appropriateness and usage rights are different questions: "Is this content okay to show" (brand-safety/policy) and "are we actually allowed to use this specific asset commercially" (rights/licensing) are separate axes of risk — a system that checks only one leaves a real exposure gap on the other.

Step 5 · Dynamic budget allocation

Beyond a fixed A/B split

Many approved variants are ready to run. Should ad spend be split evenly across all of them for a fixed test period, then pick a winner?

Design decision: Several approved variants are ready. Split budget evenly across all of them for a fixed period?

The call: Use a multi-armed-bandit-style allocator that dynamically shifts more budget toward variants showing stronger real-time performance, and less toward weaker ones — continuously, not just at the end of a fixed test window. — A bandit allocator treats each variant as an "arm" and continuously reallocates spend toward whichever is currently performing best, based on accumulating real data — this captures much more value from the SAME total budget than a fixed even split that only reacts at the end of the test period.

Use a Bandit Allocator: rather than a fixed even split, continuously shift spend toward variants demonstrating stronger real performance and away from weaker ones, using accumulating data throughout the test rather than only reacting at a fixed test's conclusion.

Exploit what you're learning while you're still learning it: A traditional fixed A/B test separates "exploration" (testing) from "exploitation" (spending on the winner) into two sequential phases. A bandit approach blends them continuously, capturing more value from the same budget by not waiting until the very end to start favoring what's already winning.

Step 6 · Close the performance feedback loop

Learn from what actually converted

Performance data accumulates on every variant. Should that just sit in a dashboard, or feed back into anything?

Feed Performance Metrics back into future Variant Generation: not retraining the underlying generative model per campaign, but using observed patterns (which styles, tones, or structures tended to convert) to bias what kinds of variants get generated for future briefs — the system gets measurably better at proposing promising variants over time, informed by real outcomes rather than generating blind every time.

A feedback loop turns a one-shot tool into a compounding one: Without feeding real outcomes back into generation, every new campaign starts from the same blind baseline. Closing the loop means the platform's output quality compounds with usage, rather than staying static regardless of how much data it has accumulated.

Step 7 · Platform-specific formatting

One approved creative, many required shapes

An approved creative needs to run on Instagram Stories (vertical), a Google Display banner (wide), and a feed post (square) — each with different required dimensions and format constraints.

A Platform Formatter automatically adapts approved creative into each target platform's required aspect ratios and format specs — generating the necessary crops/resizes rather than requiring the advertiser to manually reformat one creative for every channel it needs to run on.

Approval and formatting are separate concerns: Content approval (Steps 3-4) should happen once per creative concept, not once per platform-specific format — separating "is this approved" from "what shape does each platform need it in" avoids re-running expensive compliance checks for what's fundamentally the same approved content in a different aspect ratio.

Step 8 · The sharp edges

Winning isn't the same as safe, and generation isn't free

Two real pressures: a variant that's genuinely winning on clicks but reads as manipulative once more people see it, and the real cost of generating dozens of variants, especially video, before knowing which will actually perform.

Run ongoing compliance re-checks on live, high-performing variants, not just a one-time gate at generation — a variant can start looking manipulative or attract complaints only once it's seen by a wider audience, and a metric win doesn't override a compliance failure discovered later. For cost, use tiered generation: generate cheap image/copy variants first, gather early performance signal cheaply, and only generate expensive formats (full video) for the variants that have already shown promise — rather than producing every format for every variant up front.

Design for the unhappy path: Compliance gate down → fail closed, nothing unvalidated publishes. A winning variant → still subject to ongoing re-checks, not exempted by performance. Generation cost → tiered, cheap-first generation avoids wasting expensive formats on unproven variants. A platform that only works for well-behaved, uniformly-safe, cheap variants is a demo; one that stays safe and cost-efficient at real scale is a product.

You did it

You just designed an AI ad creative platform.

• A —
• V — a
• T — h
• R — i
• A —
• A —
• A — p
• H — i