Campaign optimisation is becoming agentic
Smart bidding was the first step: you set a target, the platform set the bids. What is different now is scope. Budget shifts between campaigns, audience expansion, creative selection and even asset generation increasingly happen without a human in the loop for each decision.
That makes the quality of your conversion signal the single biggest lever you control. An agentic system optimises faithfully toward whatever you told it to value — so if you count every form fill as a win, it will buy you an enormous number of unusable form fills, faster than any human media buyer could.
This is why I moved the accounts I run onto qualified cost per lead, with the qualification verdict fed back from the CRM. The machine does the buying; my job is making sure it is being paid to chase the right outcome.
AI video changed the economics of creative testing
Creative used to be the bottleneck. A shoot meant budget approval, a crew and two weeks. Testing four hooks was a quarterly event.
With AI-assisted scripting, generation, voice and editing, producing several variants of a vertical ad is now a same-week activity, and it is what makes a weekly creative rotation realistic for a small in-house team. The constraint has moved from production capacity to judgement: which hook is worth testing, and which result is real rather than noise.
Attribution is turning predictive
Third-party cookies, consent modes and cross-device behaviour broke deterministic tracking, and platforms filled the gap with modelled conversions. That is useful but not neutral — every platform models in its own favour.
The practical answer has not changed: own a measurement layer the platforms do not control. A clean GA4 implementation, one versioned tagging layer, a strict UTM taxonomy, and CRM outcomes blended into Looker Studio. Modelled numbers are then a signal you compare against your own source of truth rather than the only number in the room.
Answer engines are a new discovery surface
A growing share of research now ends inside ChatGPT, Perplexity, Claude or Gemini — with an answer and a handful of citations, not ten blue links. If your pages are not structured to be quotable, you are invisible on that surface even while ranking well on Google.
- Give the direct answer in the first 40 to 60 words of the page.
- Use clear question-shaped headings and a real FAQ block.
- Mark up the entity behind the content with Person, Service and Article schema.
- Keep names, numbers and claims consistent everywhere they appear.
- Publish an llms.txt and allow the AI crawlers you want citations from.
What has not changed
Offer, positioning and proof still decide whether a campaign works. AI compresses execution time; it does not make a weak offer profitable, and it does not decide what your business should be known for.
The teams pulling ahead in 2026 are not the ones using the most tools. They are the ones with clean data, a clear definition of a good lead, and a weekly rhythm of decisions on top of it.
Frequently asked questions
Will AI replace performance marketers?
It replaces manual execution — bid adjustments, variant production, reporting pulls. It does not replace the person defining the objective, the conversion signal and the creative angle, and those decisions now carry more weight, not less.
Should I let the platforms fully automate my campaigns?
Only once your conversion tracking reflects genuine business outcomes. Automation on a bad signal scales waste. Fix the signal first, then loosen the controls.
Is AI-generated creative penalised by the ad platforms?
No. Platforms judge creative on policy compliance and performance. What gets punished is low-quality, low-relevance output, whoever made it.
Want this applied to your account?
I run this playbook on live budgets. See the case studies or get in touch.
Start a conversation