Rules
Part of How to assess digital marketing trend claims before you plan
How to prepare for five future digital marketing trends
Prepare for five future digital marketing trends: AI assistant discovery, automated campaigns, modeled measurement, retail media and privacy rules. Each gets three preparation steps and a funding gate.
Five trends are changing how brands are found, bought from and measured: AI assistant discovery, automated campaigns, modeled measurement, retail media, and privacy and AI rules. Each section below gives the dated development, three preparation steps and a funding gate. The gate is the evidence a team needs before it moves budget.
Five trends and the preparation steps for each
Each trend below has a dated development behind it, three preparation steps and a funding gate. No trend is funded on a forecast alone.
1. AI assistant discovery
Google launched AI Overviews in US search in May 2024 and AI Mode in May 2025, and ChatGPT and Perplexity also answer product questions directly. An assistant can only repeat facts it can read.
- Correct price, size and stock data in the Google Merchant Center feed.
- Keep product pages crawlable and add Product structured data.
- Ask each assistant the same ten product questions each quarter and log what it says.
Funding gate: fund more content only if the log shows assistant answers that match the feed.
2. Automated campaigns
Google introduced Performance Max in 2021 and Meta introduced Advantage+ shopping campaigns in 2022. Both let the platform choose placements and creative.
- Set a spend cap and name one owner.
- Read a daily report for the first two weeks.
- Write down how to switch the automation off.
Funding gate: raise the cap only after the daily reports show no unexplained spend.
3. Modeled measurement
Universal Analytics stopped processing data on July 1, 2023, so Google Analytics 4 and its modeled conversions became the default. Meta's open-source Robyn and Google's Meridian estimate channel effects from historical spend.
- Label each number observed or modeled.
- Reconcile modeled revenue with finance records monthly.
- Run one holdout test before trusting a model.
Funding gate: shift budget only when the holdout result and the finance records agree.
4. Retail media
Amazon Ads and Walmart Connect combine browsing, advertising and checkout inside one marketplace.
- Ask the network for incremental sales, not attributed sales alone.
- Check margin after ad fees, returns and fulfillment.
- Keep a route to customers that does not depend on one marketplace, such as email consent or a direct checkout.
Funding gate: increase spend only when the incremental test beats the next-best channel after margin.
5. Privacy and AI rules
California's CPRA amendments took effect on January 1, 2023, and the EU AI Act entered into force on August 1, 2024, with duties phasing in from 2025. The FTC's rule on fake reviews and testimonials took effect in October 2024.
- Build a compliance register that lists which rule applies to which campaign.
- Name a reviewer for each entry.
- Review the register on a fixed schedule and after any platform change.
Funding gate: launch a new automated or AI-generated campaign only when its register entry is complete and approved.
Applying the gates: an illustrative retailer
Suppose a retailer receives three requests at once: repair product feeds, test automation, and reconsider funding. It can treat them as one decision, using the gates above. A named owner records the decision and the conditions for continuing, pausing or correcting each test. The retailer here is hypothetical, so no results are claimed.
For context, the team reads case-study evidence, modeled journey evidence, and trend signals. These sources help frame its questions, but they do not show what this retailer's customers will do.
Start with the product record
Before spending, the team checks that product facts are current, feeds are accurate and pages are accessible. It compares a verified feed holding price, size and stock with the answer an assistant gives to a product-comparison question. Stale feeds or blocked pages can leave its products out of the answer.
Any figure from Google's reporting on generative features is logged as an observation, not a forecast. The team notes which details need correction and checks platform dependencies, rather than treating one interface as the whole journey.
Permissions stay inside the test
The retailer matches each task's authority to the cost of a wrong action and whether it can be reversed. Before testing, it sets risk tiers and defines the review records, limits, escalation, monitoring and manual fallback for each task.
- Tier 1an agent drafts ad copy or alt text, while a person publishes and keeps a review log.
- Tier 2an agent adjusts bids within a fixed budget, with a spend cap, daily report and named owner.
- Tier 3a person approves each refund or price change, with a named approver and full audit trail.
Agent Authority Controls
- Grant authority by risk tier
- Require logs and spending limits
- Approve claims and protect data
- Human escalation and monitoring
- Stop controls
The team logs actions so a person can reconstruct what happened. Before expanding permissions, it checks that someone can stop the system, correct an error and return to manual work. A successful low-risk task does not justify authority over a different task.
For synthetic creative, the team follows the IAB AI Transparency and Disclosure Framework, which is voluntary, risk-based guidance. It asks for disclosure only when AI materially affects authenticity, identity or representation in a way that could mislead consumers, such as a realistic synthetic person or voice. Routine editing and copy do not need a label.
The framework offers two routes: a visible label in the creative, or machine-readable disclosure using C2PA. C2PA is an open standard that attaches signed provenance data, called Content Credentials, to a file, recording who made it and how it was edited. That data can be stripped or incomplete, and it does not show the content is true.
For commerce, the retailer considers how Amazon and Walmart combine browsing, retail media and checkout. It tests incremental contribution rather than attributed sales alone, and protects inventory, support, returns and alternatives to dependence on one intermediary.
The funding decision follows the record
Funding moves only when the team can explain why the evidence and expected cost support a test. It ranks scenarios, records indicators that would strengthen or weaken each one, and treats forecasts as unsettled. Cross-channel repair lessons and a strategy decision cycle help keep ownership visible.
Before altering the plan, it checks the FTC and SEC advertising rules and California privacy requirements. The FTC rules cover deceptive advertising, endorsements and, since October 2024, fake reviews. The SEC Marketing Rule governs advertising by investment advisers, so it applies only to firms in that business.
Rules carried into the next review
Preserve portable evidence and assets, introduce capabilities through bounded tests, and change funding only when indicators warrant it. Account ownership, exports, documentation, source files and human service routes stay available so the retailer can correct course or leave a dependency.
At each review the team inventories owned assets, platform dependencies, data, AI uses, rights, vendors and skills, then funds limited tests with customer and risk guardrails. It connects the record to a strategy checklist and keeps the US Copyright Office AI materials for later checks on AI-generated creative.
The compliance register from the fifth trend lists each rule beside its source, such as the FTC's advertising substantiation policy, California's privacy statute, the EU AI Act text and the NIST AI Risk Management Framework. The list points staff to material to review. It does not decide which obligations apply or replace qualified advice.
