For Marketing and Advertising

Protecting Creative Judgement: A Guide for Marketing Teams Using AI

Your media efficiency metrics are improving while your brand distinctiveness is declining. This is the trap of using AI to optimise what you already know instead of using it to answer what only you can judge. The tools you use every day (ChatGPT for copy, Midjourney for visuals, Performance Max for media buying) are exceptionally good at finding the middle ground. Your job is to keep them from getting in the way of your craft.

These are suggestions. Your situation will differ. Use what is useful.

Download printable PDF

Recognise When AI Is Replacing Strategy Instead of Serving It

When your creative team moves straight from brief to ChatGPT prompt, you are no longer doing strategy work. You are outsourcing the moments where you decide what your brand stands for, what your audience actually cares about, and why they should believe you. AI tools compress these decisions into efficiency. The risk is not that the output is bad. The risk is that it is competent enough that nobody asks whether it is right. Your account leaders need to protect the thinking that comes before the prompt.

›Before you write a prompt, write one paragraph explaining what insight about your audience or brand this campaign depends on. If you cannot write that paragraph without AI help, you are not ready to brief the AI tool.
›When a copywriter or designer hands you three variations from ChatGPT or Midjourney, ask which one breaks convention and why. If none of them do, send it back. Convention is where AI lives.
›Track which campaigns came from real strategic thinking versus which ones came from prompt engineering. After six months, compare their memorability scores and brand lift. You will see the difference.

Use AI to Expand Your Options, Not to Replace Your Taste

Midjourney and Adobe Firefly generate dozens of visual directions in minutes. This is useful only if you already know what you are looking for and what you want to reject. Many teams use AI to skip the hard work of deciding what makes their brand recognisable. They ask the tool to generate options and then pick the one that looks best. This is how brands become unmemorable. Your job is to use the speed of these tools to test more versions of your core idea, not to escape having a core idea.

›Start with a mood board or a visual principle (not an AI tool) that clarifies what your brand looks like. Then use Midjourney to generate variations within that constraint, not to generate the constraint itself.
›When you run Performance Max or other automated media buying, treat the algorithm's audience expansion as an option to test against your core target, not as a replacement for it. Measure brand perception in both audiences.
›Set a rule: if your agency could not explain why an AI output is the right choice in language that a client's CMO would understand, you do not use it. The tool should amplify your thinking, not hide it.

Keep Craft Knowledge in the Room

The account leader who understood what made the client's category work, what competitors were doing, and what had worked for this brand five years ago is the person who is hardest to replace with prompts. Many organisations are reducing these roles in favour of prompt engineers who can iterate faster. This is a mistake. The person who knows the category is the person who can tell whether an AI output is clever or just different. Without them, you will optimise toward what the algorithm rewards, not what builds a brand.

›When you hire or promote, prioritise people who can explain why a piece of work succeeds or fails in the market, not just people who can use the tools. These skills are what protect your work from becoming AI-averaged.
›Create a role or a regular practice where your most experienced strategist reviews every campaign before it goes live and asks: What does this assume about our audience? What does it assume about our brand? Would this work if a competitor used the exact same prompt? If you cannot answer these questions, the work is not ready.
›Train your creative team on why certain conventions matter in your category (tonality, visual style, audience targeting) before they start using AI tools. The tool is faster. Your judgement is rarer.

Measure What Matters, Not What Is Easy to Measure

Performance Max will tell you your cost per acquisition is down. It will not tell you whether people remember your brand next month or whether they chose you over a competitor because of what you said. Many teams optimise toward the metrics the tool gives them and assume that brand building is happening in the background. It is not. AI excels at optimising measurable outcomes in the short term. Your job is to decide whether those outcomes are building a brand or just converting traffic. If your memorability is flat while your efficiency is up, you are losing.

›Run a memorability or brand lift study every quarter on campaigns that have used significant AI generation. Compare them to campaigns where human creative judgement drove the concept. Track the difference in unaided brand recall and emotional response.
›When you run Claude or ChatGPT for email or social copy, split test AI-generated versions against human-written versions in the same audience. Measure open rates, click rates, and conversion rates. Many teams find no difference. Some find the human-written version outperforms.
›Set a rule that at least 20 percent of your media budget goes to campaigns where the creative idea came from human strategy and craft, not from AI generation. Measure these separately. Use them as your baseline for what brand building actually does.

Protect the Moments Where Your Judgement Matters Most

Not every decision deserves human attention. Bulk copy variation, asset sizing, and audience segment testing are places where AI tools do real work. But the moments where you decide what your brand actually stands for, what you want people to feel, and why your audience should believe you are not places where speed should win. These are the moments where your team's experience and taste matter. Your systems should protect time and space for this kind of thinking. When everything moves to AI, these moments disappear.

›Create a clear rule about which decisions require human sign-off and which can be delegated to AI. The rule might be: all core strategic choices, all changes to brand voice or visual identity, all campaign premises. Everything else can be iterated by the tool.
›Schedule a weekly or fortnightly conversation where your creative and strategy leads talk through what the AI tools are generating and what it is missing. Make this a formal practice, not something that happens if someone has time.
›When you train new team members, teach them what your brand should sound like and look like before you teach them how to prompt the tools. The tool amplifies your choices. It cannot replace them.

Key principles

1.Use AI to speed up execution of your strategy, not to replace the strategic thinking that makes your brand distinct.
2.The metrics AI tools optimise for (efficiency, conversion, reach) are not the same as the outcomes that build brands (memorability, distinctiveness, emotional connection).
3.Your craft knowledge about what works in your category, what your audience believes, and what your brand stands for is what protects you from becoming average.
4.Protect the moments where only human judgement matters: deciding what your brand means, why your audience should care, and whether an idea is truly memorable or just AI-plausible.
5.Measure both short-term performance and long-term brand health separately, and assume they are not the same until your data proves otherwise.

Key reminders

Before you prompt Claude or ChatGPT, write one sentence that explains the strategic insight this piece of copy depends on. If you cannot write that sentence without AI help, you are skipping the thinking that makes advertising work.
When Midjourney generates ten variations of your visual idea, ask yourself which one challenges convention in your category. If none of them do, the tool has found the middle ground. That is the opposite of what you want.
Run a six-month audit: which campaigns drove real brand lift and memorability? Which ones optimised efficiency but did not move brand perception? Look for the pattern. It will tell you where your human judgement is necessary.
Train your account leads to be able to explain every campaign decision in plain language to a client CMO. If the decision disappears when you remove the AI tool, it was not a real decision. It was convenience.
Set aside 20 percent of your time and budget for work where you skip the AI tools entirely. Use it to remind your team what human creativity actually produces. Use it as your baseline for measuring what AI does and does not do.