AI marketing rules
Book a conversationA practical guide to making better AI decisions before marketing and creative work goes live
TL;DR
For marketing teams using AI, the rules are changing quickly and unevenly. New laws are taking effect across U.S. states and the EU, platforms have their own requirements, and existing advertising laws still apply.
Small differences in how AI is used, how it’s set up and where the work runs can change which requirements apply. There are practical steps you can take now: understand your team’s current uses, identify what needs attention and build clear policies, responsibilities and checks into how the work gets made, approved and released.
Putting those pieces in place early can help avoid expensive fixes to campaigns and processes already underway. This guide explains the landscape and offers a starting point.
This guide helps leaders understand:
For one industry reference, see the IAB AI Transparency and Disclosure Framework V2, a voluntary advertising-specific guidance framework, not a law.
This is a selective operating guide for marketing and creative work. It is not a comprehensive guide to AI law, privacy, security, or regulation.
Relevant provisions are already active in the European Union, New York, and California. Pennsylvania enacted Act 99 on October 22, 2026, while additional California and Utah provisions approach on January 1, 2027. Which requirements matter depends on the actor, market, medium, and use. There is no single deadline for every organization or campaign.
Requirements can also come from different places:
When these decisions are disconnected from production, creative can reach final approval before anyone knows whether it can run as planned. Depending on the campaign and applicable requirements, possible consequences include:
These outcomes are not inevitable. Their likelihood depends on the campaign, markets, agreements, platforms, and controls already in place.
An agency delivers image, video, and audio variations for several platforms and markets. Some assets use generative fill. One includes a synthetic spokesperson. Another adapts a customer quote. The campaign is creatively approved and ready to traffic.
Then the questions begin:
Together, these questions show whether reviewers have the facts they need. The advertiser, agency, publisher, and platform may each hold part of the answer or assume someone else handled it. Creative approval alone does not establish that the released version is ready.
TikTok documents one narrow example. For eligible non-Spark ads using its Ads Manager AI-generated-content disclaimer feature, creating a new campaign, including duplicating one, resets the disclosure toggle even when an asset was previously marked.
Supporting detail. The feature is not available for every ad or account, and Spark Ads follow a different process. This example reflects documented platform instructions, not account testing or a universal TikTok rule. The practical point is that approving an asset once does not prove its current release settings are correct.
TikTok Ads Manager disclaimer instructions
For publishers and ad sales leaders, the broader question is direct: What can we accept, promise, and deliver?
Several ideas are often compressed into a single question about whether AI was used. They do different jobs.
uses AI somewhere in the process while a person retains meaningful control over the final work. Grammar cleanup, research support, background removal, or options that a person substantially reviews and revises may fit this working description. It is not a universal legal category or an automatic exemption from disclosure.
includes an image, video, audio, text, or interaction created or materially changed by an AI system. A generated product scene, substantially altered demonstration, cloned voice, or realistic synthetic video can raise different questions from routine editing.
are not the same. An invented human-like performer is different from a digital replica of an identifiable person. Legal definitions vary, so these working descriptions do not settle which rule applies. Prompts, reference assets, voice, or a source performance may still involve a real person's rights or contractual protections even when the final image looks invented.
tells a person something relevant about AI's role in the content or interaction. Its wording, timing, placement, and accessibility may depend on the law, platform, format, and context.
can carry information about an asset's origin or edit history in a machine-readable form. It does not prove that a claim is true, that rights were cleared, or that the content is lawful. It can also be incomplete or removed during editing and distribution.
A visible label, platform declaration, technical credential, rights clearance, and internal approval record may all be relevant. They are not interchangeable. A label also does not cure an unsupported advertising claim, invented customer experience, missing talent consent, or other underlying problem.
Begin by understanding how teams currently use AI across marketing and creative work. Map where it enters the process, who is involved, and which uses need closer attention. Then apply the following steps to a consequential campaign or content process.
Identify the business purpose, audiences, markets, people represented, channels, formats, intended reuse, and where AI is involved.
Review claims, evidence, source assets, talent and customer permissions, applicable contracts, personal information, and questions requiring a specialist.
Determine any audience-facing notice, platform declaration, accessibility requirement, or machine-readable provenance. Record why an exception or voluntary choice applies.
Name the creative, brand, agency, legal, technical, distribution, and final release owners. Establish who handles routine work, exceptions, and escalation.
Check the final crop, edit, export, localization, upload, sponsorship setting, campaign duplication, and actual channel display.
Retain the permissions, claim support, decisions, and approvals needed for the work under an agreed privacy and retention approach. Internal records do not all belong in public metadata.
The policies, responsibilities, and checks should match the organization's real production process. A high-profile synthetic spokesperson campaign may require more review than routine editing support. Teams and partners should know what to provide, where decisions happen, what to verify before release, and when to stop and ask for help.
Sources were reviewed on October 21 and October 23, 2026. These are research review dates, not legal approval or continuous monitoring. This map is selective and does not determine how a requirement applies to a particular organization or campaign.
Advertising claims still need support. Endorsements, testimonials, reviews, native advertising, and material connections still need appropriate treatment. Disclosing AI use does not cure a deceptive claim or invented customer experience.
FTC advertising and marketing guidance (guidance not statute)It distinguishes providers from deployers and includes different requirements for direct AI interactions, machine-readable marking, deepfakes, and some public-interest text. It also contains exceptions and qualifications. Certain systems placed on the market before August 2 have until December 2, 2026 for the provider marking duty under Article 50(2). That transition does not postpone every Article 50 duty.
EU AI Act Article 50It contains specific exclusions, including audio advertising and certain promotion of expressive works. The responsible actor, medium, knowledge, and exception need to be evaluated for the actual campaign.
General Business Law § 396-bSeparate platform and model-hosting provisions begin January 1, 2027. The roles and thresholds matter. A brand using an outside tool should not assume that every provider or platform duty applies directly to the brand.
California AI Transparency Act, BPC Chapter 25The official history confirms enactment. The final linked act text was not available during this review, so this guide does not state an operative date. The latest available bill text addresses specified knowing, reckless, or willful use of synthetic advertising content that creates a false consumer perception about goods or services, with stated exceptions. Recheck the final act text before applying it.
SB 806 official history and Act 99 statusIt covers a defined synthetic performer prominently included in an advertisement and includes specified exceptions. It does not apply to every AI-assisted ad.
SB 1050 chaptered textIt covers specified providers and platforms using thresholds and scope that differ from California. Exclusively internal systems that are not publicly accessible are expressly excluded from the covered provider definition.
Digital Content Provenance Standards Act, compiled Chapter 72c (PDF)The IAB AI Transparency and Disclosure Framework V2, published in August 2026, offers voluntary advertising-specific guidance based on risk and materiality. It includes an operational framework and implementation roadmap. It is not a law or universal safe harbor. Organizations still need to apply relevant requirements to their own people, partners, and processes.
IAB AI Transparency and Disclosure Framework V2
Platform rules may require a label or setting even when a cited statute does not. Platform acceptance does not establish legal clearance. Talent, agency, vendor, publisher, and union agreements can add obligations involving consent, compensation, security, reuse, warranties, evidence, and removal. The actual agreements need to be examined rather than inferred from the final asset.
The useful starting point is often an existing campaign or content process, not a new policy in isolation.
Jim Marsh can work with leaders and the appropriate legal, compliance, privacy, labor, security, and technical specialists to:
Jim spent 15 years at HBO helping establish and scale social media, digital marketing, data-driven marketing, and emerging technology practices while maintaining high audience, brand, cultural, and creative standards. Through JMC Intelligence, he now applies that experience to AI strategy, workflows, governance, implementation, enablement, and handoff.
This work does not provide legal advice, regulatory certification, formal audit, or a guarantee of compliance or business performance. The goal is clearer decisions, identified gaps, named owners, and a process the organization can use.
If you have a campaign or process where AI requirements are becoming difficult to interpret or manage, contact Jim Marsh at jim@jmcintel.com.
This guide is intended to help leaders ask better operating questions and work effectively with qualified specialists. It does not cover every jurisdiction, regulated sector, privacy or biometric rule, election ad, talent agreement, platform, or active legal proceeding.