AI-generated ad creative: how to keep control of your brand in Google Ads

Updated: 29 July 2026

Short answer: you are liable for copy you did not write

This is the sentence to start with, because it changes the whole risk calculation. Google now generates headlines, descriptions and images from your landing page and existing ads, but Google’s terms are explicit: advertisers bear full responsibility for all ad content, including content generated by Google’s AI tools. The editorial policy cuts the same way: the policy does not care that a machine made the mistake.

Control is available but scattered across the interface, and partly advisory rather than binding. You have four layers: brand guidelines, term exclusions, messaging restrictions, and a recurring review of what the system is actually serving. None of them works alone, because brand guidelines are not hard, fast rules. Google will take them into account, but things won’t be perfect.

The practical consequence: brand control in automated campaigns stopped being a one-off setup task and became a recurring editorial process. Below is what you can set, what you cannot, and the review checklist.

What automatically generated assets are, and where they appear

Two things are worth separating, because the interface blurs them. Automated assets are elements the system adds to your ad. Automatically created assets are new copy and images the model produces from your material. Google scans your landing page as well as your existing ads, then dynamically creates new assets it predicts will perform well.

Where this is live today:

• Performance Max. The campaign assembles assets into ads across all Google surfaces.
• Demand Gen. Automatically Created Assets has been enabled by default in Demand Gen campaigns since Q3 2024, and in 2026 Google expanded it to include image background generation and product image enhancement for e-commerce advertisers.
• AI Max in Search campaigns. A search campaign setting that uses broad match keywords, final URL expansion and automatically generated assets, aiming to capture additional searches and match ad copy and landing page content to the initial query.
• Asset Studio. The central creative workspace. In the May 2026 updates Google positioned it as a central workspace that can use marketing briefs, brand guidelines, website content and campaign goals to generate assets across formats, with refinement through natural-language input.

Four real risks, not one

The conversation about AI creative usually stops at “it might not sound like us”. That is the cheapest of the risks. Four more serious ones:

Risk one: unsupported claim elevation. ACA may elevate a minor landing page mention into a prominent ad claim that constitutes an impermissible health or financial claim. This is regulatory exposure, not an aesthetic problem, and it hits regulated verticals hardest.

Risk two: omitted disclosures. ACA-generated descriptions may omit required disclosures that your manually created ads include.

Risk three: competitor trademarks. ACA may incorporate competitor brand terms from your landing page content into ad headlines. If you run comparison pages naming competitors, the exposure is real.

Risk four: image quality as a policy breach. Image quality rules still ban assets that are blurry, unclear, unrecognizable, or contain illegible text, and that clause matters more in 2026 than ever, because AI generation tools produce exactly those failure modes at scale.

Then there is the fifth risk, the one most discussed and the easiest to manage: stylistic drift. If there are no clear brand rules, the system can generate a lot of material that is technically acceptable but stylistically unstable.

What you can actually configure

Brand guidelines

This is the base layer. Brand guidelines let you control how your brand is represented in your Performance Max campaign automated assets or formats. Two mechanics matter before you configure anything.

First, the system guesses on your behalf if you let it. Performance Max campaigns use Google AI to infer key brand elements from your final URL, and the first time you use a final URL, Google will infer your brand guidelines from it. If Google successfully detects a business name and logo, these assets are pre-filled by default.

Second, some settings require deliberate confirmation. Google AI will infer font and colors from your final URL and recommend them in the UI, but you have to manually confirm or modify these values. Skipping that step means the campaign runs on the model’s guess about your visual identity.

There is also a formal requirement worth checking: the provided business name must match exactly with your domain or verified legal name.

Term exclusions and messaging restrictions

Two further layers sit on top of brand guidelines. Per a 2026 creative standards summary, term exclusions cap at 25 per campaign, messaging restrictions govern required and disallowed content patterns, and brand guidelines pin colors, logos and voice. The settings are stackable, and they apply to every AI-generated asset that ships from your account.

The 25-term cap is a meaningful design constraint. It means you cannot exclude a long list of phrasings, so the list has to be built around genuine legal or reputational risk rather than around wording you simply dislike.

What you cannot configure

Three limits worth stating to a client or board before they ask:

• Guidelines are advisory, not absolute. Google will take them into account, but things won’t be perfect.
• Control is distributed across campaigns, not centralised. Not all brand guidelines live at the account level. Instead, they’re at the campaign level, so make sure you have all controls in place as needed throughout your account. At several dozen campaigns this is a real operational load.
• Your logo will not always serve. Business information assets, including logos, are not guaranteed to serve in every impression, because the system optimizes for the best performing format.

And the governing principle worth repeating inside the team: guidelines are only as strong as your input, so the better information you put in, the better outputs you’ll get.

The editorial process: what to do and how often

Quarterly review, plus after any significant site change

1. Pull the asset report. The practitioner recommendation following the July 2026 terms update is direct: pull a report on your Performance Max and Demand Gen campaigns, review the asset groups in detail, check which headlines, descriptions and images Google’s system is using most frequently, and cross-reference those with your current brand guidelines.
2. Check three things in every generated headline: whether it makes a claim your page cannot substantiate; whether it omits a required disclosure; whether it contains a competitor name.
3. Check message match against the landing page. This risk grows with automation: the more Google expands dynamic and visual campaign environments, the more dangerous it becomes to create assets that look good in isolation but do not prepare the click properly.
4. Review image quality against policy, not just aesthetics: blur, illegible text, distorted detail.

Refresh cadence

Automation accelerates creative burnout, because the system tests combinations faster and distributes them across more placements, so fatigue hits faster. The published recommendation is refreshing static assets every 4 to 6 weeks and video every 8 to 12 weeks. Treat that as a starting point to calibrate against your own data rather than a rule, since burn rate depends on budget and audience size.

The compliance and disclosure layer

This is where care is needed, because industry sources sometimes run ahead of official platform documentation. There is a circulating account of an AI content disclosure requirement in Google Ads, described as a rule mandating that advertisers disclose when ad creative, including headlines, descriptions and visual assets, has been generated or substantially modified by artificial intelligence, with Google handling the labeling on its end once the advertiser correctly flags the creative. It is linked to the EU AI Act, which mandates transparency for AI-generated content that users might believe was human-created.

Methodological caveat: this specific requirement is described by a single agency source and could not be confirmed in Google’s own documentation. Verify the current position in the Google Ads Help Centre and in AI Act guidance before adopting a labelling procedure. The regulatory direction is nonetheless clear, and maintaining an asset library that records the provenance of each creative is a low cost against the exposure.

Implementation checklist

• [ ] Brand guidelines filled in manually, not left on values inferred from the final URL
• [ ] Business name matching the domain or verified legal name
• [ ] Font and colours confirmed or modified, not skipped
• [ ] The 25 excluded terms built around legal risk, not stylistic preference
• [ ] Messaging restrictions set for both required and disallowed content
• [ ] Guidelines configured in every campaign, not just one
• [ ] Recurring asset review in the calendar, with a named owner
• [ ] Refresh schedule for static and video assets
• [ ] Provenance record for creative, against future disclosure requirements
• [ ] Message match between ad and landing page re-checked after every offer change

Worth noting: controlling what the system says about your brand in an ad is the same problem as controlling what language models say about it in an answer. The mechanism is identical in both cases. A model reads your site and constructs a description from it. Diagnosing that second layer is what a Brand Search Presence audit does, and consistent brand definitions improve the outcome on both sides at once.

If you need support at the account level, this work belongs inside ongoing Google Ads management as part of a recurring creative review rather than as a separate project.

Frequently asked questions

Can I switch off automatically generated assets entirely?
In some campaign types yes, in others no. In campaigns built around automation, such as Performance Max, disabling generation means abandoning the format. The better strategy is guidelines plus exclusions plus recurring review, rather than trying to block the mechanism outright.

Who is liable for an error in an ad Google generated?
The advertiser. Platform terms assign full responsibility for ad content to the advertiser regardless of whether the content was written manually or produced by Google’s tools.

Do brand guidelines guarantee my ads match my visual identity?
No. They are taken into account but do not operate as a hard rule. Your logo is not even guaranteed to appear on every impression, because the system optimises for format performance.

How often should I review automatically generated assets?
At minimum quarterly, and always after a material change to the landing page, since the page is the source the model draws from.

Is AI Max brand-safe?
It depends on the vertical and the configuration. It combines broad match, final URL expansion and asset generation, so it stacks three sources of unpredictability at once. In regulated verticals it needs tighter exclusions and more frequent review than a standard Search campaign.

Does AI-generated content require a label?
The regulatory direction points toward mandatory transparency, but specific platform requirements move quickly and individual industry sources run ahead of the documentation. Verify the current position in the Google Ads Help Centre before adopting a procedure.

Sources

• Google Ads Help, About brand guidelines → https://support.google.com/google-ads/answer/14934472?hl=en
• WordStream, How to Use Google Ads Brand Guidelines for Cleaner Automated Assets, June 2026 → https://www.wordstream.com/blog/google-ads-brand-guidelines
• Taikun Digital, Google Ads Creative Standards 2026 → https://taikundigital.com/insights/google-ads-creative-standards-2026
• Audit Socials, Google Demand Gen Campaign Compliance Guide 2026, April 2026 → https://www.auditsocials.com/blog/google-demand-gen-campaign-compliance-guide-2026-creative-audience-policy
• RichClicks, AI for Google Ads: how to improve assets in 2026, June 2026 → https://www.richclicks.co.uk/digital-magazine/ai-for-google-ads-how-to-improve-assets-in-2026
• Novoads, Google Ads Creative Standards in 2026: What Changed for Video Ads, June 2026 → https://novoads.ai/en/blog/google-ads-creative-standards-2026
• Impression Digital, Protecting your brand when implementing AI Max in Google Ads, November 2025 → https://www.impressiondigital.com/blog/protecting-your-brand-when-implementing-ai-max-in-google-ads/
• Common Thread Collective, Google Ads New Terms of Service: July 2026 AI Automation Changes, June 2026 → https://commonthreadco.com/blogs/coachs-corner/google-ads-tos-july-2026-ai-automation-ecommerce
• groas, Google Ads AI-Generated Label Requirement In 2026, May 2026 → https://www.groas.com/post/google-ads-ai-generated-label-requirement-2026-compliance-guide

Data currency and limitations note. Google Ads creative features changed every few weeks through 2026, and several controls rolled out in stages. The 25-term exclusion cap and the recommended refresh cadences come from industry write-ups rather than platform documentation and should be verified in your own account. The AI content disclosure requirement described here comes from a single agency source and was not confirmed in Google’s documentation. This article reflects the position in July 2026 and requires quarterly review. It is not legal advice.