AI-Generated Ads Are Getting Labels: What Brands Need to Document Now

AI-generated creative has moved from experiment to everyday production.

A small team can now produce image variations, resize a campaign, draft headlines and test new concepts faster than it could a year ago. That speed is useful. It also creates a basic operating question that many businesses have not answered:

Who is responsible for knowing what the AI changed, proving that the claims are true and deciding when a disclosure is required?

Google made that question more immediate on July 9, 2026. The company announced new ways for people to see when generative AI helped create or alter an ad, along with new controls for advertisers using AI tools outside Google.

This is not a reason to stop using AI. It is a reason to build a simple record of how creative moves from idea to approved ad.

Google changed how AI-created ads are disclosed.

Google's July 9 announcement introduced a “How this ad was made” section in My Ad Center for ads across Search, YouTube and Discover.

Google says that assets created with its own generative AI advertising tools can receive disclosures automatically. Advertisers creating or editing assets elsewhere can use a control to indicate that generative AI was involved. Depending on local requirements, a label may appear directly on the ad.

Google also updated its advertising policy guidance. Starting in July 2026, advertisers can place text or visual AI labels inside image and video creative without violating Google's usual restrictions on watermarks and text overlays. Google is also rolling out an AI label setting across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Ads Editor.

The most important sentence in the policy update is easy to miss. Google says that using its label setting does not guarantee compliance with every applicable regulation.

The platform can provide a disclosure mechanism. The advertiser still owns the accuracy of the ad and the decision about which legal requirements apply.

A label does not make an inaccurate ad acceptable.

AI disclosure and truth in advertising are related, but they are not the same responsibility.

The Federal Trade Commission's advertising guidance states that advertising claims must be truthful, must not be deceptive or unfair and must be supported by evidence. Those standards apply whether a person, an agency or a model produced the creative.

An AI label does not repair:

• A generated product image that shows a feature the real product does not include.
• A home service ad that implies emergency availability when the company does not offer it.
• A franchise ad that uses a testimonial from someone who was never a customer.
• A nonprofit campaign that presents a fabricated beneficiary or outcome as real.
• A technology ad that promises performance the product has not demonstrated.

The practical rule is simple. Disclosure explains how an ad was made. Substantiation proves that what the ad says is true.

Brands need an AI creative record, not another committee.

Most small and mid-sized organizations do not need a complicated governance program. They need a repeatable record that follows each materially AI-assisted asset.

Create a short AI creative record with these fields:

1. The source material is identified. Record the approved product photos, customer quotes, performance data, offer details and brand guidance used to create the asset.
2. The tool and type of assistance are recorded. Note whether AI generated the scene, changed a background, extended an image, adapted copy, created a voice or produced a full video.
3. The material changes are described. State what the final asset depicts that did not exist in the original source.
4. Every factual claim has an owner. Assign a person to verify prices, service areas, availability, product details, results, testimonials and regulated claims.
5. Disclosure requirements are checked. Confirm the platform control, the markets where the ad will run and any legal review needed for the category or location.
6. The approved final asset is retained. Save the version that actually ran, along with its approval date and campaign destination.

This record can live inside the existing project-management system. The point is not the software. The point is making the decision traceable.

Different businesses face different failure modes.

The same workflow should adapt to the risk behind the offer.

Home service businesses should protect local accuracy.

Generated ads often make a local service company look larger, faster or more available than it is. Review every service, city, response-time statement, license reference, financing claim and warranty before publishing.

If a generated image shows a technician, truck, uniform or jobsite, make sure the scene cannot reasonably be mistaken for a real employee, branded vehicle or completed project when it is not one.

The campaign controls in our small-budget AI Max readiness guide address a related issue. Automation works better when geography, conversion signals, landing pages and exclusions are already accurate.

Outdoor and direct-to-consumer brands should protect product truth.

Generated lifestyle scenes can show products in conditions that were never tested. They can also change proportions, materials, included accessories or safety details.

Keep the real product as the visual source of truth. If AI changes the item itself, the team should compare the final asset with the current product specification before it reaches the media plan.

Franchise brands should separate shared approval from local claims.

A national team may approve the concept while a location supplies the offer, inventory, service area or customer proof. The creative record should identify which facts were approved centrally and which were verified locally.

This prevents a strong national template from carrying the wrong local phone number, promotion, availability or testimonial into dozens of markets.

Technology startups should slow down performance promises.

AI can make an unfinished feature look production-ready. Product marketing should confirm that every workflow, integration, speed claim and customer outcome reflects the current product.

Screens that are conceptual should be treated as conceptual. A generated interface should not be presented as a live product experience when customers cannot use it.

Nonprofits should protect dignity and donor trust.

A fabricated person or scene can create an emotional response without representing a real program participant. Even when the campaign goal is worthy, an invented story can weaken trust.

Use real, permission-based stories when possible. When illustrative imagery is necessary, avoid presenting it as documentary evidence of a specific person, event or impact.

Human review should happen before the asset enters the platform.

Do not wait for Google to add a label or flag a policy concern.

Add a short approval gate before upload:

• The brand reviewer confirms that the asset looks and sounds like the organization.
• The subject-matter reviewer confirms that products, services and outcomes are accurate.
• The campaign owner confirms that the offer, geography, landing page and tracking match the ad.
• The responsible lead confirms that the required AI and endorsement disclosures are present.
• The final approver confirms that source files and evidence are stored with the asset.

This is the same discipline that should already exist in a sound Google Search campaign. The ad, destination and measurement plan should describe the same offer.

Platform tools do not replace clear disclosure judgment.

Businesses should use the controls Google provides, but they should not assume that one platform setting resolves every disclosure question.

The FTC makes a similar point in its endorsement guidance. When a material connection requires disclosure, the responsibility stays with the brand and endorser, even if a platform offers a disclosure tool.

AI disclosure and endorsement disclosure address different facts. An ad may need one, both or neither depending on how it was created, what it claims, who appears in it and where it runs.

When a campaign involves regulated products, political content, health claims, financial claims, children or several legal jurisdictions, the marketing team should involve qualified counsel. A checklist can improve operations. It cannot make a legal determination.

A practical seven-day cleanup can reduce the risk.

Use one week to make the current creative process more defensible:

1. Inventory the AI tools used by employees, contractors and agencies.
2. Review active image and video ads for material AI generation or editing.
3. Confirm whether Google's AI label control is available in each advertising account.
4. Create the short AI creative record inside the team's existing workflow.
5. Assign claim approval to named people instead of a general marketing inbox.
6. Recheck product, service, offer and testimonial evidence for active campaigns.
7. Update agency and creator briefs so disclosure decisions happen before final delivery.

Do not turn the audit into a hunt for every grammar suggestion or automated crop. Focus first on changes that could affect what a reasonable customer believes about the business, offer, product, person or outcome.

The practical takeaway is accountability.

Google's new AI disclosures are a visible sign of a larger shift. Generated creative is becoming normal, while the expectation for traceable, accurate advertising is becoming more explicit.

The advantage will not belong to the brand that generates the most assets. It will belong to the brand that can move quickly without losing the source of truth.

AI can accelerate production. Human owners still need to approve the promise.

Build a faster creative process without losing control.

Strats & Roadmaps helps lean teams connect paid media, content production, brand standards and performance reporting into a practical operating system. Review the paid-media approach or start a conversation to build an AI-assisted campaign workflow that protects accuracy and customer trust.