AI in marketing: what is proven, what is promising and what is still noise
Updated: Nov 3
First published 24th June 2025. Updated 25th October 2026 with current evidence on AI in marketing.
Most AI commentary in marketing falls into two camps: everything has changed, or nothing has. Neither helps a business decide where to spend. The more useful question is narrower. Where does AI already change commercial outcomes, where is it promising but unproven and where is the claim running ahead of the evidence?
What is established
Platform automation is now the default. Google's Performance Max and AI Max and Meta's Advantage+ campaigns choose bids, placements and audiences with machine learning. That is no longer optional or experimental. The advertiser's lever has moved from manual targeting to the quality of the signals fed in: accurate conversion tracking, conversion values that reflect margin rather than revenue and first-party audience data. Automation trained on weak signals optimises efficiently towards the wrong outcome.
AI answers are changing how people search. Google reported more than two billion monthly users of AI Overviews in July 2025 and OpenAI reported 900 million weekly ChatGPT users in February 2026. Pew Research Center found that US users clicked a traditional result on 8% of visits to a results page that showed an AI summary, against 15% when there was none. Discovery is spreading across more surfaces and fewer of them send a click.
Productivity gains are real but uneven. In a controlled study with Boston Consulting Group, consultants using GPT-4 completed tasks inside the model's capability around 25% faster and at higher quality, but performed worse on a task outside it (Dell'Acqua et al., Harvard Business School, 2023). Our experience matches: AI shortens first drafts, research synthesis and data preparation. It does not replace judgement on what matters.
What is promising but not yet proven
AI-generated creative at scale. Volume is now cheap. The evidence that more variants alone improve business outcomes is thin and creative quality remains one of the largest drivers of advertising return
Ads inside AI answers. Google has begun placing ads in AI Overviews and AI Mode. Pricing, measurement and performance benchmarks are still forming, so early budgets should be treated as tests with a control
Agents that research and buy on a customer's behalf. The capability is advancing quickly, but adoption for considered purchases is early. Clear product data and consistent information across the web are sensible preparation regardless
What is still noise
Guaranteed AI visibility. No one controls how a model answers. Anyone promising a fixed placement in ChatGPT or Gemini is selling certainty that does not exist
"SEO is dead." Google says its AI features draw on the same index as Search. Crawlable, accurate and well-structured content remains the foundation for both
Tools as strategy. Buying an AI platform does not answer who the customer is, what the business needs to sell or what it can afford to pay for a customer
The risks that deserve attention
AI raises practical obligations as well as opportunities. Personal data entered into third-party tools must be handled under the privacy law of each market a business operates in, including Australia's Privacy Act, the GDPR in the UK and EU and a growing set of US state laws. The EU AI Act is phasing in obligations from 2025, including transparency duties for general-purpose models. Generated content also needs checking for accuracy, intellectual property and brand consistency before it is published.
A simple test before adopting anything
Which commercial measure should this improve: revenue, margin, cost per acquisition or time saved?
Can it be tested against a control, so the effect is measured rather than assumed?
What data, brand or legal risk does it introduce and who owns that risk?
If it works, can it be run reliably every week, not just in a pilot?
AI is already part of how good marketing is planned and run. The advantage comes from applying it where the evidence supports it and measuring it as carefully as any other investment.
Sources: Pew Research Center, 22nd July 2025; TechCrunch, 27th February 2026; Alphabet Q2 2025 earnings call; Dell'Acqua et al., Navigating the Jagged Technological Frontier, Harvard Business School Working Paper 24-013, 2023; Google Search Central, AI features and your website
