AI in Marketing: Profiling Rules in Canada and California

How Quebec's profiling rule, federal behavioral advertising guidelines, and California's ADMT regulations treat AI-driven marketing personalization.

Guide: AI Tools and IndustryLast verified: October 20266 min read

Most AI-driven marketing personalization rests on profiling, meaning automated processing of personal data to infer a person's interests or traits, and Canadian and California rules treat that step differently. Quebec requires profiling functions to stay off until the person switches them on, federal guidance in Canada allows opt-out consent for behavioral advertising only under set conditions, and California's automated decisionmaking rules, in force since January 1, 2026, expressly leave advertising outside their definition of a significant decision.

Key takeaways

  • Quebec's private sector privacy Act (section 8.1) bars switching on identification, location, or profiling functions by default and requires notice of how to activate them.
  • The federal Privacy Commissioner's guidelines treat opt-out consent for online behavioral advertising as reasonable only with clear notice, an immediate and lasting opt-out, and mostly non-sensitive data.
  • California's ADMT regulations list five kinds of significant decisions (finance, housing, education, employment, healthcare) and state that advertising to a consumer is not one of them.
  • Ad targeting is not unregulated in California: separate CCPA rules on selling or sharing data for cross-context behavioral advertising still apply.
  • Which rule governs a campaign depends on where the audience lives, so one workflow can face three different standards.

What profiling means in marketing rules

Marketing systems that recommend products, pick email send times, or choose which ad a person sees usually build a profile first. Quebec's Commission d'accès à l'information (CAI) quotes the statute's definition on its business guidance page: profiling is the collection and use of personal information to evaluate certain characteristics of a person, with work performance, economic situation, health, personal preferences, interests, and behavior given as examples. California's regulation text defines profiling as automated processing of personal information to evaluate personal aspects of a natural person and to analyze or predict them.

The two definitions read alike, yet the legal consequences diverge. In Quebec the definition triggers a notice and activation step at the point of data collection. In California the word appears in the regulations, yet the new automated decisionmaking duties attach to a narrower set of outcomes, as a later section shows.

Quebec: profiling stays off until the person acts

The CAI's page on collecting personal information explains that a business using technology with identification, location, or profiling functions must say so from the outset and must describe the means available to activate those functions. The page states that such functions cannot be activated by default, and that the person concerned must be able to turn them on voluntarily. The CAI's privacy policy guide cites section 8.1 of the private sector Act as the source of this rule and gives a personalized purchase-profile recommendation as a profiling example.

Automated decisions are handled separately. The same CAI page says that when personal information is collected to make decisions entirely by automated processing, the business owes additional information to the person, who may ask for human intervention. A marketing model that only ranks offers may fall outside that provision, a point for legal counsel to decide.

Federal guidance on behavioral advertising

The Office of the Privacy Commissioner of Canada (OPC) publishes guidelines on privacy and online behavioral advertising (page modified August 11, 2025). They say opt-out consent could be considered reasonable only when several conditions hold at once:

  • individuals are told the purposes in understandable terms, not buried in a privacy policy;
  • notice comes at or before collection and names the parties involved;
  • the opt-out is easy to use, ideally at or before collection;
  • the opt-out takes effect immediately and persists;
  • the data used is limited, as far as practicable, to non-sensitive information; and
  • collected data is destroyed or effectively de-identified as soon as possible.

Sensitive information in the guidelines includes health, financial, genetic, and biometric data, ethnic and racial origins, political opinions, sex life or sexual orientation, and religious or philosophical beliefs. For children, the OPC says meaningful consent is difficult to obtain and that organizations should avoid tracking children and tracking on sites aimed at children.

California: the ADMT rules stop short of advertising

The California Privacy Protection Agency announced on September 23, 2025 that the Office of Administrative Law approved regulations on automated decisionmaking technology (ADMT), risk assessments, and cybersecurity audits. They took effect on January 1, 2026, and businesses using ADMT for significant decisions must comply starting January 1, 2027. In the regulation text, ADMT means technology that processes personal information and uses computation to replace or substantially replace human decisionmaking. The regulations name five areas of significant decisions: healthcare services, housing, financial or lending services, education enrollment or opportunities, and employment, contracting opportunities, or compensation, where the decision grants or denies access. Subsection (ddd)(6) of the definitions section adds that a significant decision does not include advertising to a consumer.

That does not free ad targeting from California privacy law. The same regulation text says a service provider cannot contract with a business to deliver cross-context behavioral advertising, and a company that does so is treated as a third party. Selling or sharing data for that purpose therefore stays subject to the CCPA's opt-out machinery.

Side-by-side comparison

The table condenses the three sources above. It describes what each document says about marketing profiling and is not a compliance test.

QuestionQuebec (CAI)Federal Canada (OPC)California (CPPA)
Is there a default setting rule?Profiling functions cannot be on by defaultNo default rule, opt-out consent allowed under conditionsNot addressed in the ADMT definition of significant decisions
Is advertising profiling covered?Yes, the definition is not limited to any purposeYes, through the behavioral advertising guidelinesExcluded from ADMT significant decisions, covered by sale and sharing rules
Sensitive dataHandled by general consent rulesOpt-out unsuitable; examples listedHandled elsewhere in the CCPA, not examined here
Human review of automated decisionsPerson may ask for human intervention in fully automated decisionsNot part of the advertising guidelinesADMT rights cover significant decisions starting January 1, 2027

Questions for a campaign review

  • Which provinces or states hold the people being profiled?
  • Does the tool infer sensitive traits such as health or religion from browsing or purchases?
  • Is each profiling feature off until a person turns it on, where Quebec residents are involved?
  • Does the same model also feed decisions on credit, housing, jobs, or healthcare, which would bring California's ADMT rules into play?
  • Can a person opt out in one step, and does that choice reach every vendor holding the data?

This article gives general information from published regulatory sources. It is not legal advice, and statutes and guidance change, so a privacy lawyer or the regulators themselves should confirm how a rule applies to a specific campaign.

FAQ

Does Quebec's rule apply to a business outside Quebec?

The CAI's guidance addresses businesses collecting personal information in the private sector in Quebec. Whether it reaches an out-of-province business marketing to Quebec residents is a legal question that the sources reviewed here do not settle.

Is opt-out consent ever enough for AI-based ad targeting in Canada?

The federal guidelines say it can be reasonable for online behavioral advertising when all listed conditions are met, including limiting the data to non-sensitive information. Sensitive data and children's data are the areas the OPC flags as problematic.

Do California's ADMT rules ban targeted ads?

No. They define significant decisions in five areas and exclude advertising, so ad targeting falls under other CCPA provisions rather than the ADMT duties.

Is there a Canadian federal law written for AI in marketing?

None was identified in the sources checked in October 2026. Existing privacy statutes and regulator guidance do the work.

Last verified: October 2026

Sources

  • Collecte de renseignements personnels (business guidance), Commission d'accès à l'information du Québec, 2026
  • Rédiger une politique de confidentialité, guide explicatif pour les entreprises, Commission d'accès à l'information du Québec
  • Guidelines on privacy and online behavioral advertising, Office of the Privacy Commissioner of Canada, modified August 11, 2025
  • Text of Regulations (CCPA Updates, Cyber, Risk, ADMT, and Insurance Regulations), California Privacy Protection Agency, 2025
  • Announcement on approved regulations, California Privacy Protection Agency, September 23, 2025

New to artificial intelligence?

Start with the basics, then follow the guides to tools, rules, and effects on society.

Start with AI Basics
Ultimate Guide to B2B Email Marketing :