AI changed the cost of making things. It did not change what is worth making.
Most of the AI conversation in marketing is about tools. The more useful conversation is about economics: producing a variation of something now costs almost nothing, while producing something worth varying costs exactly what it always did.
That single shift explains almost everything happening in the industry right now — including why so much output has become simultaneously more polished and less memorable.
Where AI genuinely helps
Variation and volume
Thirty versions of an ad for testing. Twelve caption options. Every asset resized for every placement. This is real, immediate leverage and there is no reason to do it by hand.
The blank page
Getting from nothing to a rough draft is the slowest part of most creative processes. AI is very good at producing a mediocre first version quickly, which is worth more than it sounds — editing beats originating for most teams.
Analysis at a scale nobody was doing manually
Reading two thousand reviews and telling you what customers actually complain about. Clustering search queries. Summarising a quarter of comments. This work was previously skipped, not done badly — which means the gain is genuinely new.
The unglamorous middle
Transcripts, alt text, meta descriptions, translation drafts, reformatting, tagging. Deeply boring, genuinely time-consuming, and low risk if it is imperfect.
Where it quietly destroys value
Strategy
AI is trained on what already exists, which makes it structurally biased toward the consensus. Ask it for positioning and you will get the average of your category — the exact thing positioning is supposed to escape.
Voice
The reason your brand sounds like something is that a person made a series of specific choices. Generated copy regresses toward a competent, agreeable middle. Across a year, that is how a brand stops being recognisable without anyone noticing the moment it happened.
Anything claiming to be experience
Reviews, testimonials, case studies, first-person accounts. Generating these is not efficiency, it is fabrication, and one discovery costs more than the entire saving.
Judgment about what matters
AI will happily produce a perfectly structured campaign for the wrong objective. It has no way to know that your real problem is retention, not awareness.
What good practice actually looks like
- Automate the middle, not the ends. The thinking at the start and the judgment at the end stay human. The production in between is where the leverage is.
- Keep a written voice guide and check against it. If you cannot tell generated copy from yours, your voice was not distinctive enough to begin with.
- Never let it touch a factual claim unverified. Numbers, dates, laws, product specifications, competitor statements. Check every one.
- Disclose generated imagery of people. Our full position on that is here.
- Spend the time you saved on the part that cannot be automated. If AI halves your production time and you produce twice as much average work, you have gained nothing.
The competitive reality
When everyone can produce competent content instantly, competent content stops being a differentiator. The things that become scarcer — and therefore more valuable — are a genuine point of view, proprietary information, real access, and craft that is obviously expensive to produce.
Which is the opposite of the conclusion most brands drew, which was to make more.
The bottom line
Use AI for volume, variation, analysis, and the boring middle. Keep strategy, voice, judgment, and anything resembling a factual claim firmly in human hands. The brands losing ground to AI are not the ones that failed to adopt it — they are the ones that used it to make more of what nobody wanted.
Want an honest read on where AI fits in your marketing? Let's talk.