Can you use AI to differentiate a brand without everything ending up sounding the same? Yes, but only if you change what you use it for. Ask it to write the message and it will hand you your category's average, admittedly well written. Use it to map what every competitor already says and to pinpoint the territory nobody owns, and it becomes a seriously powerful differentiation tool.
The diagnosis arrives with a distinguished signature. In an interview with Marketing Week, Sir John Hegarty, co-founder of BBH, describes much of today's marketing as "repetitive, cliché, boring", and blames brands and agencies staffed by "left brain thinkers" who understand a spreadsheet but don't really get an idea. His favourite argument: "something like 95% of new product launches fail", a rate he says has barely moved in 20 years, despite all the data now available. He concedes the exact figure is debatable, but the substance of his critique stands.
Why does so much AI-made advertising sound the same?
Because of how the models work. A language model predicts the most likely continuation of a piece of text, and the most likely is, by definition, whatever most resembles what already exists. Ask it for a line for a sports brand and it will talk about pushing through, effort and breaking limits. Ask it on behalf of your competitor and it will give them the same thing in different words.
The result is a quiet convergence. Each piece, on its own, looks perfectly fine. Together they form a category where nobody stands out and consumers can't remember who said what.
The tool isn't to blame. The blame lies with asking an averaging machine to do the job of an idea.
What Hegarty gets right, and where AI genuinely helps
Hegarty doesn't talk about artificial intelligence in the interview. He doesn't need to. His line "Data doesn't have the answer" rests on a simple point: data describes what has already happened, and marketing should be looking ahead. An AI trained on what already exists inherits exactly that limit.
But there's a nuance worth drawing out. If data describes the past, it also describes, with real precision, where everybody is standing. And knowing where the herd is happens to be exactly what you need in order to leave it.
Hegarty's advice fits this approach: write a manifesto that makes clear what the brand believes in, keep good campaigns running for longer instead of changing them out of boredom, and listen to non-buyers too, because a brand isn't built by its customers alone.
Why do brands follow the herd?
Herd behaviour theory explains that, under uncertainty, we tend to copy what others are doing. If everyone is heading one way, there must be a reason. And if it goes wrong, at least nobody was wrong on their own.
In advertising, that instinct creates category codes everyone copies: slow motion and sweat in sport, the smiling family in food, the floating dashboard in tech. Following them feels safe in front of an approvals committee. In reality it's expensive, because you pay the same media investment for a message the audience could attribute to anyone.
Badly used, AI amplifies the herd: if the whole category works with similar models and similar briefs, convergence is all but inevitable. To truly connect, you have to dare to diverge from the trending path. The good news is that the same technology lets you do it with method rather than blindfolded.
How to differentiate a brand's message with AI, step by step
- Map the whole category. Analyse the ads and organic content of every competitor, not just the three the client names in the brief.
- Group by territory. Identify which promises, tones, formats and visual codes keep repeating. That's your map of what's crowded.
- Look for unclaimed tensions. Cross-reference that map with the audience's real conversation: what worries or frustrates people that no brand is answering.
- Write the point of view before the tagline. This is where human judgement rules. A short manifesto, in the spirit of what Hegarty proposes: what the brand believes in and what it rebels against.
- Use AI as a sparring partner, not an author. Ask it which competitor each idea resembles and which cliché it repeats. If it can't find one, you're on the right track.
- Test concepts before committing media spend. Score their predicted clickability and keep the ones that are different and also look likely to work.
- Give the idea time. Differentiation is built through consistency. Changing the message every quarter sends the brand straight back into the noise.
Generic AI versus strategic AI
| Stage | Use that leads to generic | Use that differentiates |
|---|---|---|
| Research | Asking a chatbot for "industry trends" | Analysing the real ads and content of the whole category |
| Insight | Summarising what everyone says | Finding audience tensions nobody is addressing |
| Message | Generating fifty taglines and picking the one that sounds best | Writing from a brand point of view and using AI to stress-test it |
| Creative | Copying this month's viral format | Building distinctive, recognisable codes of your own |
| Validation | Launching and waiting to see what happens | Scoring concepts before the media investment |
How do you know a different idea will work before you invest?
This is the real handbrake. Standing out is scary because different has no track record, and no marketing director wants to defend an unproven bet in front of their board. That's where AI genuinely reduces risk instead of increasing sameness.
On the Defy platform, Defy Monitor scores the predicted clickability of each concept before any media spend, using a model validated against real ad performance. The idea is simple: know whether a creative will work before you pay for it. And because it analyses a whole category's ads in minutes, it also shows whether that idea looks too much like what competitors are already doing.
A warning, so you don't fall into the opposite trap: a high score doesn't make an idea good if it looks like all the others. The filter has to work both ways. Is it different? And is it likely to work? If it only passes one of those tests, it isn't ready yet.
Judgement is still human
AI can save an agency hours of competitor audits, summarise thousands of conversations and anticipate which concept has more pull. What it can't do is decide what a brand believes in. That remains the job of people with taste, judgement and a measure of courage.
Hegarty is right: data doesn't have the answer. But used well, it tells you which answers are already taken.
