AI's effect on digital marketing extends beyond individual tools. Information discovery, advertising platforms, content economics and decision-making speed are changing together. McKinsey's State of AI 2025 reports that 88% of organisations use AI in at least one function, with marketing and sales among common applications. This article examines five measurable structural changes and the responsibilities that remain human.
Summary: Five shifts stand out: search is moving towards direct answers, with Pew observing traditional-result clicks of 8% when an AI summary appeared versus 15% without one; advertising automation moves competition towards data and creative quality; cheaper content production makes editorial quality more important; accessible analysis speeds decisions; and personalisation becomes more individual. Strategy, brand, trust and creative direction still require people.
Five Changes at a Glance
| Area | Earlier model | Emerging model | Business implication |
|---|---|---|---|
| Search | Link-based results and clicks | AI summaries and conversational answers | Being a useful cited source matters alongside rankings |
| Advertising | Manual bidding and targeting | PMax and Advantage+ automation | Signal quality and creative variety become key inputs |
| Content | Expensive, slower production | Faster, cheaper drafting | Unreviewed scale creates risk; editorial control differentiates |
| Data | Reports waiting for analysts | Natural-language access to insights | Decision speed becomes an advantage |
| Experience | Segment-based campaigns | Individual, real-time adaptation | First-party data infrastructure becomes essential |
Change 1: Search Behaviour and Answers Without Clicks
Users increasingly receive answers on the results page or from an assistant. A website click is no longer the assumed outcome of every search. This challenges the traditional expectation that strong rankings necessarily produce proportionate traffic.
Pew Research Center followed approximately 900 US adults' search behaviour in March 2025. Users clicked a traditional result in 8% of searches with an AI summary, compared with 15% without one. Clicks on sources inside summaries were around 1%, while 26% of visits to pages with summaries ended the browsing session. These are observations from one month and one country, not universal rates for every market or query. The study does not establish whether ending a session means the user received a satisfactory answer.
In February 2024, Gartner predicted that traditional search-engine volume would fall 25% by 2026 because of chatbots and virtual agents. This was a forecast; we do not have definitive evidence here confirming that outcome. What can be observed is the spread of discovery across Google AI Overviews, ChatGPT, Perplexity and Gemini.
Visibility therefore has a broader meaning. Rankings remain important, while being cited in AI answers becomes another area to assess. Our GEO guide explains the framework, and our AI Overviews content optimisation article covers the Google-specific application.
Change 2: Advertising Platforms Automate and Agency Roles Evolve
Google Performance Max and Meta Advantage+ delegate much of targeting, bidding and placement to platform AI. Advertiser control increasingly concerns the inputs: conversion signals, creative and measurement.
Meta's fourth-quarter 2024 results reported an annualised revenue run rate above USD 20 billion for Advantage+ shopping campaigns, with approximately 70% annual growth. More than 4 million advertisers were using its generative advertising tools. These figures illustrate the scale of adoption.
When many advertisers use similar automation, differentiation comes from three areas:
- Signal quality: accurate conversion tracking, offline sales information and first-party integrations give algorithms better evidence.
- Creative variety: meaningful variants provide useful material for testing as targeting becomes automated.
- Measurement architecture: independent evaluation helps distinguish incremental outcomes from conversions that would have happened anyway.
The agency's value shifts towards designing these inputs and reviewing automated outcomes, rather than simply operating campaign controls.
Change 3: Content Economics and the Value of Quality
Generative AI reduces drafting time and cost, excluding the separate work of source checking and editorial review. This creates both a flood of low-value material and greater relative value for supervised production.
Google's March 2024 scaled content abuse policy addresses mass content created primarily to manipulate rankings, regardless of whether people or AI produced it. Google initially targeted a 40% reduction in low-quality, unoriginal results through this update and earlier work, later reporting 45% after the rollout completed on 19 April (Google Search announcement).
AI use itself is not prohibited. The risk concerns low-value production at scale for manipulation. The operating model matters:
- Unreviewed production: prompt directly to publication, without source checking, an editor or original analysis. Even if it brings short-term traffic, manipulative scaled publishing creates policy risk.
- Human-supervised production: AI supplies drafting capacity while people verify facts, add real experience and perform editorial review.
As production becomes cheaper, the ability to generate text stops being a sufficient advantage. The differentiator is a process that combines capacity with dependable quality.
Change 4: Analysis Becomes More Accessible
Questions once dependent on lengthy reporting cycles can increasingly be asked in natural language: which campaign is losing money in a segment, or where did conversions begin declining? Analytical capabilities become more accessible to smaller teams.
The commercial effect is decision speed. Finding waste today rather than at next week's meeting can matter financially. Anomaly detection, forecasting and alerts support earlier intervention instead of purely retrospective reporting.
However, McKinsey reports that only a minority attribute enterprise-wide EBIT impact to AI despite widespread adoption. Acquiring tools and creating value are different achievements. Process integration matters: faster insights accomplish little unless someone is responsible for acting on them at the right time. Businesses need to connect outputs with decisions, not merely generate reports more quickly.
Change 5: Customer Experience Moves from Segments towards Individuals
Personalisation traditionally grouped people by age, interests or purchasing history. AI can support finer adaptation, such as different product ordering, content and timing for individual visitors.
Conversational assistants are a visible example. Compared with rigid scripted bots, language-model assistants can follow context and support handover. Properly configured, they can provide initial contact and qualification outside working hours. They can still answer incorrectly or misdirect users, so clear handover thresholds are necessary. Contact details and behavioural information are personal data; privacy information, a valid legal basis, minimisation and security belong in the initial design.
The less visible prerequisite is first-party data infrastructure. A business with fragmented CRM records and disconnected conversion data lacks the foundation for dependable personalisation. Improving customer experience is often a data-organisation project before it becomes an AI project.
What Remains: Strategy, Brand, Trust and Creativity
AI accelerates execution but does not independently determine the right direction. Choosing markets, positioning and offers still requires people who understand customers and competition.
When many businesses can produce similar materials at similar speed, brand history, voice and earned trust remain difficult to copy. As content volume grows, being a reliable source becomes more valuable.
AI is useful for variants, while original concepts and creative direction still need human judgement. Since advertising automation places greater weight on creative inputs, good ideas remain valuable. The distribution of work changes, but strategic responsibility does not disappear.
Five Practical Steps for Businesses
A large transformation programme is not required to begin. Most businesses can start with concrete steps in an appropriate order:
- Map processes: identify repetitive, data-intensive work such as reporting, drafting and monitoring. These offer practical starting points.
- Broaden visibility measurement: assess citations and mentions in AI Overviews, ChatGPT and Perplexity alongside Google rankings.
- Establish editorial standards: make fact checking, source verification and human approval mandatory before scaling content.
- Improve advertising signals: repair conversion tracking, first-party flows and independent measurement before expecting automation to compensate for weak data.
- Pilot, measure and expand: choose one process and a measurable goal. Demonstrate value before wider deployment rather than equating adoption with profit.
For everyday applications rather than the broader transformation, see how AI is used in digital marketing.
References
- McKinsey: The State of AI in 2025: Agents, Innovation, and Transformation, November 2025
- Pew Research Center: Google users are less likely to click on links when an AI summary appears in the results, 22 July 2025
- Google Search Central: March 2024 core update and new spam policies
- Google Search: New ways we're tackling spam and content quality, March 2024, including the 40% target, 45% reported reduction and 19 April completion note
- Gartner: Search volume forecast for 2026, 19 February 2024
- AdExchanger: Q4: Meta Minted Money And Improved Its Monetization, covering Meta's Q4 2024 results
DijitalPi combines AI-assisted execution with experienced marketing oversight. Let's plan your AI marketing infrastructure.
Cookie preferences