IAB Italia this month published its White Paper on Artificial Intelligence, a document developed by the association's dedicated AI Working Group to map how artificial intelligence is being adopted across marketing and advertising in Italy. The publication, dated April 13, 2026, sets out both the current state of the sector and a detailed outlook extending toward 2030.

The paper positions 2026 as the year of production-level deployment. Not isolated experiments, not proof-of-concept pilots, but AI integrating stably into daily workflows across creative production, targeting, performance reporting and campaign measurement. According to the document, this shift has arrived rapidly: the paper notes that within the last three years, AI has moved from a specialist infrastructure topic to a central theme in virtually every business conversation.

Seven domains, one framework

The white paper structures its analysis across seven thematic areas: research, insight and trend identification; content production; automated reporting; targeting and audience profiling; performance analysis; creativity; and hyper-personalisation. Each section pairs operational guidance with real case studies drawn from IAB Italia member companies, making the document practical as well as descriptive.

Running through all seven domains is a single organising idea: AI does not replace human expertise, it amplifies it. According to the document, "le evidenze emergono un punto chiave: l'AI non sostituisce la competenza, ma la amplifica" - the key finding is that AI does not replace competence, it amplifies it, and the best results come when technology and people work in synergy.

Research and insight: from information overload to structured pipelines

The paper opens its substantive analysis with a detailed treatment of market research and insight generation. According to the document, 75% of knowledge workers already use generative AI, and nearly half adopted it within the six months prior to the paper being written. The volume of data sources available to marketing professionals - reports, white papers, social posts, first-party data, third-party signals - has expanded to a point where manual desk research can no longer keep pace with market velocity.

To address this, the paper proposes a six-phase framework called the FAI (Framework AI to Insight), covering objective definition, signal collection, data cleaning and enrichment, pattern analysis, insight narrative construction, and decision implementation with impact measurement. The framework is explicitly human-in-the-loop: AI accelerates and synthesises, but final selection and verification of insights remains with the team.

On the technical side, the paper identifies three categories of data relevant to trend analysis. Unstructured data - social posts, news articles, audio transcriptions, online reviews - sits alongside semi-structured data like RSS feeds and JSON from APIs, and structured data from databases, CRM and ERP systems. Making these interoperable is described as a significant technical challenge, one addressed by tools including Brandwatch (with integrated NLP), Diffbot (which transforms web content into structured data via AI-powered computer vision), and platforms such as Trifacta and KNIME for data preparation.

For trend identification specifically, the paper covers Natural Language Processing including BERT-based algorithms for sentiment analysis and named entity recognition, topic modelling techniques such as Latent Dirichlet Allocation, and time-series forecasting models including ARIMA, Meta's Prophet, and Recurrent Neural Networks. The intent is to move from keyword search to what the paper calls semantic discovery: identifying meaning rather than matching strings.

Content production and the legal landscape

The content production section takes a notably cautious tone, dedicating substantial space to legal and ethical constraints. The paper analyses EU Regulation 2024/1689, better known as the AI Act, and its implications for generative content. It distinguishes between inputs used to train GenAI systems and the outputs those systems produce, noting that the copyright status of each category remains contested.

According to the document, the EU AI Act requires transparency for AI-generated content: deployers must ensure outputs are labelled as machine-generated, with an exception only for AI that performs standard editing assistance functions. The boundary between assistance and autonomous generation is, the paper acknowledges, legally ambiguous - and that ambiguity has direct implications for advertising, where the use of synthetic content that is not correctly signalled could constitute misleading commercial practice under the Italian Consumer Code (D.Lgs. 206/2005).

On deepfakes, the paper is explicit. Italian law introduced Article 612-quater of the criminal code through the Legge IA, establishing penalties of one to five years imprisonment for anyone who, without consent of the person depicted, distributes images, video or audio falsified through AI systems in a manner designed to deceive. This provision sits alongside the AI Act's own deepfake disclosure obligations, which require visible labelling distinguishing AI-generated from authentic content - a stricter standard than the technical watermarking permitted for other synthetic content categories.

The paper also covers watermarking technology in depth. It cites Google's SynthID as an example of digital watermarking that embeds an imperceptible signal directly into image pixels, surviving subsequent modifications such as filters, colour changes and format conversions. This kind of invisible marking is permissible for most synthetic content but not sufficient for deepfakes, where the AI Act demands clearly visible labelling.

Two draft Codes of Practice on AI-generated content transparency - published by the European Commission on December 17, 2025, and March 5, 2026 - are referenced in footnotes, indicating the paper was finalised against an actively developing regulatory backdrop.

Practical case studies in the content section include AdHub Media's PetStory project, which used the proprietary Edi-Tree platform with OpenAI's 4o mini for text generation and FLUX 1 Schnell for image generation to launch a new pet-focused editorial property rapidly, and Taboola's GenAI Ad Maker embedded in its Realize platform.

IAB Italia had previously addressed adjacent regulatory territory. The association's September 2025 white paper on digital sustainability covered the environmental dimensions of digital advertising operations. That earlier document established a pattern of structured, working-group-driven analysis that the AI paper continues.

Targeting, identity and first-party data architecture

The targeting section opens with a concise account of the regulatory constraints now shaping audience infrastructure. It notes that third-party cookies are in decline and that GDPR, the ePrivacy Directive and the AI Act together impose a framework in which first-party data onboarding, marketing automation and AI-driven profiling must all comply with legal bases for processing.

Within that constraint, the paper identifies five technical domains for AI-assisted targeting. The first, Identity and Data Foundation, covers first-party data onboarding, CRM enrichment and the use of Customer Data Platforms. The second, Predictive Targeting and Scoring, describes models that score users for propensity to purchase, churn probability, or lifetime value - enabling dynamic micro-segmentation rather than the static rules-based segments that dominated earlier periods.

The third domain, Cross-Channel Audience Activation and Orchestration, addresses the technical challenge of coordinating audience signals across paid search, programmatic display, social and CTV environments. The fourth, Contextual Targeting, has gained renewed significance as identifier-based targeting has weakened; the paper treats it not as a fallback but as a genuinely independent capability, particularly for publisher inventory monetisation through private marketplace deals.

The fifth area, Conversational Profiling, is treated with explicit legal caution. The paper describes techniques that infer user characteristics from dialogue patterns - noting their effectiveness while highlighting the GDPR and AI Act implications of profiling based on conversational data, including potential restrictions under the AI Act's provisions on manipulation and subliminal influence.

A case study in this section describes a premium automotive brand that used what the paper calls Cognitive Brand Matching to achieve a threefold improvement in ROI by aligning audience targeting with brand-relevant contextual signals rather than relying solely on behavioural data.

The targeting content connects to a broader pattern documented across European markets. IAB Europe's September 2025 AI impact report, which surveyed 95 companies, found that 85% already deploy AI tools for marketing, with targeting and content generation leading at 64% and 61% respectively.

Performance analysis and reporting automation

The reporting section covers four capabilities the paper frames as a progression from descriptive to prescriptive. Automated insight generation and anomaly detection sits at the descriptive end; diagnostic analysis and root cause identification represents the next stage; predictive simulation and scenario modelling follows; and prescriptive recommendation with optimisation closes the sequence.

The practical case studies here are drawn from global brands operating in Italy. Heineken is cited for a programme combining AI-driven media optimisation with automated insight generation through Google's tools. Unilever appears twice - for work with IBM that applied AI to reduce media waste and improve quality metrics across campaigns. A third case study describes anomaly detection AI deployed for operational performance monitoring and cybersecurity.

The paper also covers centralisation of data in Google BigQuery for AI-driven cross-channel analysis, and frame-by-frame video analysis for brand safety and advanced targeting - a technique that allows contextual suitability scoring at the scene level rather than the content level.

Hyper-personalisation: predictive plus real-time

The hyper-personalisation section draws a clear boundary between classical personalisation - which is retrospective, segment-based and relatively static - and what the paper calls iper-personalizzazione: a combination of predictive analytics and real-time adaptation. The distinction, as the paper frames it, is between checking historical weather records to assess whether rain is likely, and having a system that sees clouds arriving, recalculates in real time, and suggests the best action at that moment.

Three application domains are described. Personalisation based on historical data - using RFM scoring, lookalike modelling and propensity models - represents the established baseline. Dynamic and conversational personalisation introduces next-best-action logic updated in real time based on cross-channel signals. The third domain covers the infrastructural shift needed to make advanced hyper-personalisation operational: specialised agents, orchestration frameworks and integrations connecting AI models to CMS, CRM, e-commerce and knowledge base systems.

The Salesforce Agentforce deployment on Salesforce.com is presented as a detailed case study. According to the paper, the agent handled more than 100,000 conversations after going live, qualified leads 40% faster than previous workflows, generated more than 30,000 new leads, and operated in 11 languages.

The future: Marketing WITH, OF and TO AI

The final third of the paper steps back from operational guidance to map three structural directions the paper calls Marketing WITH AI, Marketing OF AI, and Marketing TO AI.

Marketing WITH AI covers the use of AI as a creative and analytical partner: generative models for content ideation, predictive models for audience segmentation and scenario simulation, and automation of campaign delivery with continuous feedback loops. According to the document, by 2030 working in marketing will mean knowing how to direct, interrogate and orchestrate AI systems - not to be replaced by them, but to become more strategic through the combination.

Marketing OF AI addresses the communication challenge of marketing AI itself as a product, service or brand value. The paper notes that 76% of consumers, according to IBM's Global AI Adoption Index (2024), trust AI technologies more when the data source and algorithm are made explicit. Explainability, transparency policies in chatbots, and AI trust ratings are all cited as emerging communication practices.

Marketing TO AI is the most forward-looking of the three. The premise is that AI agents will increasingly intermediate purchasing decisions on behalf of human users - selecting products, filtering content, managing transactions. When this occurs, brands will need to optimise not for human perception but for machine readability. The paper introduces the concept of Agent Optimization as a successor framework to SEO: structuring feeds, APIs, JSON-LD markup and product data so that AI agents can correctly parse, evaluate and recommend a brand's offering.

According to the document, this will require intelligent metadata on all content, structured language formats readable by machine, cross-platform consistency (since agents will penalise brands for inconsistencies between what appears on e-commerce, social platforms, comparison sites and owned properties), and reputation management calibrated not for human emotional response but for statistical reliability as evaluated by algorithmic scoring systems.

The paper also identifies four new professional roles likely to emerge: the AI Prompt Designer, responsible for authoring the language-level instructions that govern machine outputs; the Creative Technologist, bridging artistic and developer capabilities; the Data Ethicist, governing the use of data in creative contexts; and the Agent Orchestrator, coordinating workflows that combine multiple AI agents operating in parallel.

Regulatory readiness and governance measures

Woven through the content production and targeting sections is a detailed governance framework. The paper recommends that organisations adopt written AI use policies covering prompt formulation guidelines - specifically instructing staff to verify that prompts do not contain personal data without legal basis, do not use third-party images or voices without consent, do not incorporate intellectually protected material, and do not embed biases that could lead to discriminatory outputs.

Staff training is framed as a legal obligation, not just a best practice. Article 4 of the AI Act requires deployers and providers of AI systems to ensure AI literacy among their employees. The paper references the Spanish company INECO's approach - beginning with a general onboarding session for all staff, then progressively specialised sessions differentiated by role and existing knowledge level.

The paper also describes the AI Pact, the European Commission's voluntary early-adoption programme. According to the document, roughly half of the organisations that have joined the AI Pact have committed to obligations beyond what the regulation strictly requires, particularly around human oversight, risk mitigation and transparency in AI-generated content management including deepfakes.

This regulatory dimension is not new to IAB Italia's European counterparts. IAB Europe's July 2025 whitepaperaddressed growth, guardrails and policy for the European digital advertising market, and IAB Spain's Top Digital Trends 2026 report identified the AI Act's enforcement timeline as creating immediate compliance pressures, with most obligations having taken effect on August 2, 2025. IAB Italia's white paper contributes the Italian national perspective to this broader European regulatory alignment. The EU Parliament committee's March 2026 position on the Digital Omnibus on AI, proposing fixed compliance deadlines of November 2, 2026, for AI-generated content marking, adds further urgency to the governance measures the paper describes.

Why this matters for the marketing community

The white paper arrives as Italian marketing professionals face a compressed timeline. The AI Act's general-purpose model provisions entered application on August 2, 2025. The proposed November 2026 deadline for AI-generated content labelling, if confirmed, gives organisations less than 18 months to implement compliant workflows. Meanwhile, as the IAB Italia Search Forward conference of March 2026 documented, 35.34% of Italian online adults already use generative AI tools to inform shopping decisions - nearly double the 19.6% recorded in February 2025.

The paper is structured to be useful at different levels of organisational maturity. For teams already deploying AI tools, the legal and governance sections provide a compliance framework. For teams earlier in adoption, the seven-domain structure and accompanying case studies map the available territory. The landscape of Italian AI providers categorised by specialisation area, included within the document, offers a practical starting point for vendor evaluation.

Timeline

Summary

Who: IAB Italia, the Italian association for interactive advertising, developed the white paper through its dedicated AI Working Group, drawing on the experience of member companies that develop, implement and use AI solutions across the Italian digital advertising supply chain.

What: A comprehensive white paper covering seven domains of AI application in marketing - research and insight, content production, reporting, targeting and profiling, performance analysis, creativity, and hyper-personalisation - together with a legal and governance framework covering the AI Act, copyright, deepfakes, and organisational readiness, and a forward-looking section mapping the Marketing WITH, OF and TO AI framework extending toward 2030.

When: Published on April 13, 2026, with regulatory references extending to the European Commission's second draft Code of Practice on AI-generated content published on March 5, 2026, and the proposed November 2, 2026, compliance deadline for AI-generated content labelling under the AI Act.

Where: The paper addresses the Italian digital advertising market specifically, with a landscape of Italian AI providers categorised by specialisation, while situating Italy within the broader European regulatory framework of the AI Act and related Commission guidance.

Why: The paper aims to help Italian marketing professionals and companies navigate the shift from experimental AI deployment to production-level integration, providing practical case studies, a structured governance framework, compliance guidance and a strategic outlook at a moment when AI tools are becoming standard workflow infrastructure rather than optional additions.

Share this article
The link has been copied!