Marketing Pulse | Google Ads
Google AI ad labels: which creative will be disclosed, and what does your team need to record?
Google now shows how generative AI was used to create or edit some ads. The label is only the visible end of the process. The business still needs to know where each asset came from, whether it shows the real product or service accurately, who approved it, and what buyers actually see.
Stan Consulting verdict
The new label is not an approval process. Build the approval process behind it.
What changed: Google added a How this ad was made section to My Ad Center for ads across Search, YouTube, and Discover. Google says assets created with its own generative advertising tools can be disclosed automatically. Advertisers can use a new control for assets created or meaningfully edited with AI elsewhere.
Does it affect the business? Yes, when a team, agency, freelancer, platform, or creative tool uses generative AI to create or substantially change images or video that run through Google. It can also matter when Google's automated features create and serve variations on the advertiser's behalf. The immediate task is not to stop using AI. It is to stop losing track of what was generated, approved, labeled, and launched.
What could happen if it is ignored? A buyer may see a disclosure that the marketing team did not expect. A generated scene may imply a product size, finish, location, result, employee, customer, or service experience that the business cannot support. An agency may say Google handled the label while nobody can identify the source file or approval. In some targeted regions, a visible overlay may appear on the ad itself. Google also warns that using its label control does not guarantee compliance with local requirements.
What can be checked in ten minutes? Open the Asset library, add or review the AI label status column, filter the assets, and choose one live campaign. For every image and video in that campaign, record the source, the tool used, whether the asset was generated or meaningfully edited, who approved it, the selected label status, and the date the live ad was checked.
Disclosure asks whether AI was used to create or edit the asset. Approval asks whether the final claim, product depiction, person, place, price, result, and brand presentation are accurate enough to run. A correctly labeled ad can still be inaccurate or weak.
Update log
What changed and what each change triggered.
This page changes when the platform changes the buyer-visible disclosure, the advertiser control, the asset report, or the legal boundary Google describes. A refreshed date without a changed decision is not an update.
Advertiser evidence edition published
Google's launch post, product instructions, policy update, generated-image guidance, and political-ad boundary were reviewed with current industry coverage.
Triggered: this permanent guide, the saved creative check, and a direct action from the current Pulse edition.How this ad was made reaches My Ad Center
Google announced a global section in the ad information panel on Search, YouTube, and Discover. Google-created assets can receive disclosures automatically; outside AI work can be identified through an advertiser control.
Triggered: verify both the source of the asset and the disclosure buyers can see.AI label controls begin a gradual product rollout
Google documents a July rollout across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor. Asset reporting gains an AI label status column and filter.
Triggered: create one account-level asset inventory instead of relying on memory or filenames.Election advertising keeps a separate disclosure rule
Google requires verified election advertisers to disclose synthetic or digitally altered content that inauthentically depicts real people or events.
Triggered: keep political requirements separate from the broader commercial-asset workflow and use the stricter applicable process.Who needs to act
Start with the asset source, not the campaign type.
Review now
AI has touched the final image or video in a way that creates a new realistic scene, person, event, product setting, voice, or substantial visual result.
- Images generated inside Google Ads.
- Creative made with Midjourney, Firefly, Canva, ChatGPT, Runway, or another outside tool.
- Product photos placed into generated rooms, landscapes, or lifestyle scenes.
- Still images animated into video.
- People, voices, testimonials, demonstrations, or results generated or materially altered.
- Google automation creates or serves asset variations for the campaign.
Document, then decide
Ordinary production edits may not carry the same meaning, but the business should still know what changed and why.
- Cropping or resizing for an ad format.
- Color, brightness, defect, or background cleanup that does not invent a realistic event.
- Brand-safe text layout added by a designer.
- Compression or export changes.
- An untouched licensed photograph.
- A real customer, employee, product, property, or project shown without synthetic alteration.
Google's political policy explicitly distinguishes inconsequential edits such as cropping, color correction, and defect correction from synthetic depictions of events. The broader commercial AI-label control is rolling out around assets created or edited with AI. When the boundary is unclear, preserve the source and edit history, review the current product prompt, and choose a decision the business can defend.
Use this now
The ten-minute AI ad creative check.
This saves progress in this browser. It does not upload creative or account information.
No account check recorded in this browser yet.
Decision table
What to record for each kind of asset.
This table is an operating guide, not a substitute for the current account prompt or applicable legal advice. Product availability is gradual, and Google says some labels can be applied automatically or inherited from other platforms.
| Asset situation | What Google documents | Business record to keep |
|---|---|---|
| Created with Google generative tools | Google says a disclosure is added automatically. Machine-readable SynthID and C2PA signals are also applied to Google-generated images and videos. | Prompt or brief, source input, selected output, accuracy review, approver, date, campaign, and live-ad check. |
| Created or materially edited outside Google | The advertiser can use the AI label control to identify the asset as created or edited with AI. | Tool, source file, edit description, label decision, approver, and exported final file. |
| Created by a fully automated Google feature | Google says some assets may be labeled on the advertiser's behalf. Certain automatic labels cannot be overwritten. | Automation setting, asset status, where it served, final depiction, removal decision, and owner. |
| Already carries outside AI metadata or a label | Google may show the source page or inherit signals. Google also says some labels cannot be overwritten. | Original source, metadata status, current label, ownership, and any platform conflict. |
| Advertiser adds a visible label to creative | Google allows these labels. It advises keeping the label outside the 5.5 percent perimeter safe area and disabling enhancements that could crop it. | Label language, placement proof for every ratio, enhancement settings, and rendered-ad screenshots. |
| Campaign targets the EU, India, or New York | Google says designated AI assets can receive visible overlays in these geographies and that its control does not guarantee legal compliance. | Targeted locations, applicable guidance, legal owner, chosen disclosure, and live regional proof. |
| Election advertising | A separate Google policy requires disclosure for synthetic or altered content that inauthentically depicts realistic people or events. | Verification status, campaign setting, prominent disclosure where required, source evidence, and policy review. |
Evidence register
One row per asset is enough to stop creative history from disappearing.
Minimum fields
Use the asset name or ID, campaign, source owner, source URL or file, creation tool, date created, what AI changed, final dimensions, label status, approver, approval date, target regions, live check date, and a link to the final screenshot. If the team cannot complete those fields, it does not yet have an accountable creative process.
Review the claim, not only the pixels
An AI asset can be visually polished and still create the wrong promise. A generated contractor can imply that a real employee will arrive. A staged product image can make an item look larger, better finished, or differently equipped. A synthetic customer can resemble a testimonial. A generated office can imply a physical location. Approval must compare the visual with the ad copy, offer, landing page, inventory, service delivery, and actual proof.
Check every rendered ratio
One approved master file can appear differently in square, landscape, vertical, in-feed, and video placements. Google specifically warns that advertiser-added labels can be trimmed and gives a perimeter safe area. A live check should confirm that the product, required text, disclosure, and main claim survive the final rendering. The source file is not the buyer experience.
Give one person final ownership
Creative can pass through a media buyer, designer, agency, AI tool, brand manager, and business owner. Shared access does not create shared accountability. Name the person who can approve, reject, relabel, pause, or remove an asset. Record that decision beside the asset, not in a chat thread that will be impossible to reconstruct later.
Verify from the buyer side
The Asset library shows what the account believes about an asset. My Ad Center shows part of what a buyer can learn from the live ad. Use both. Open a real served ad where possible, select the three-dot menu or information icon, and look for How this ad was made. Save the date, surface, disclosure, crop, advertiser identity, and destination. Repeat the check on a materially different placement when the campaign spans Search, YouTube, and Discover.
A preview is useful for catching layout problems, but it is not complete production proof. Automated combinations, responsive formats, asset enhancements, and regional treatment can change the rendered result. The final check should answer a simple question: if a buyer opens the ad information panel, does the explanation match the creative history the business has on file?
Make agency handoffs explicit
An outside agency may generate the asset while the business owns the account. The agency should deliver the source, prompt or edit brief, selected output, rights information, label decision, and approval record with the final files. The business should retain that packet after the engagement ends. Account access without creative provenance leaves the next operator guessing about which assets are real, licensed, generated, or altered.
Pause when the evidence is incomplete
Pause or remove an asset when nobody can identify its source, when the final scene changes a material product or service fact, when a synthetic person resembles a testimonial or employee without a defensible basis, when the label status conflicts with the known process, or when the live rendering hides required information. A small delay is cheaper than paying to distribute a claim the business cannot explain.
If an AI disclosure becomes visible, compare performance carefully. A change in click or conversion behavior may reflect buyer trust, the quality of the creative, the offer, the audience, or the placement. Do not declare that the label helped or hurt from one ad without a controlled comparison and enough volume.
Original sources
Platform facts used in this check.
Product and policy statements were checked against Google documentation. News and agency pages were reviewed to identify unanswered advertiser questions, not to establish the rule.
- Google Ads and Commerce: Expanding AI transparency in ads for the July 9 announcement, My Ad Center placement, automatic Google disclosures, and advertiser control.
- Google Ads Help: Use AI content label settings and disclosures for the AI label status column, setup flow, regional overlays, safe area, and non-overwrite conditions.
- Advertising Policies Help: Updates to AI labeling requirements for the July rollout and allowance for AI labels within image and video creative.
- Google Ads Help: About generated images in Google Ads for advertiser review responsibility, policy treatment, SynthID, C2PA, and generated-image restrictions.
- Advertising Policies Help: Image asset format requirements for image safe areas, text treatment, format constraints, and AI-label policy notes.
- Advertising Policies Help: Political content for the separate synthetic or digitally altered election-ad disclosure requirement.
Connected decisions
This update should improve the ads that actually run.
Need accountable paid creative?
Build the campaign, the creative record, and the revenue path together.
Stan Consulting manages Google Ads across campaign structure, creative, landing pages, tracking, lead handling, and sales feedback.