Google AI Can Now Rewrite Your Ads — And You’re Still Liable

Google AI Can Now Rewrite Your Ads

The Change in Plain Language

Google has updated its Google Ads Terms of Service to expand the language around how advertiser-provided inputs can be used across its “AI-powered” campaign features. The new terms clarify that information entered into conversational experiences and similar Google Ads tools — including URLs, product descriptions, and campaign copy — may be utilized by Google AI systems to generate ad assets automatically.

The previous terms generally allowed advertisers to opt in or opt out of many automation features. That opt-in language has been replaced with broader authorization language. The new terms state that Google and its affiliates are authorized to serve ads using assets derived from advertiser inputs — across automated campaign types — as part of standard account operation.

“Google Ads features may generate your headlines, suggest your sitelinks, build your asset groups, and recommend targeting changes. The legal and operational responsibility for what runs still belongs to the advertiser.”

The new terms also reinforce the advertiser’s obligation to review, approve, or remove all campaigns and ad assets generated automatically by Google AI features built into their ad platform — and confirm that advertisers are responsible for ensuring they have the rights to their own inputs.

In other words: Google writes the copy. You are responsible for it.

Why This Should Concern Advertisers

For most advertisers, the concern is not the Google AI automation itself. The concern is accountability without visibility. Google AI can now generate ad headlines and descriptions drawn from your website and campaign inputs, and those assets can run in live auctions before you have reviewed them — unless you have explicitly set up review workflows to catch them.

For regulated industries — financial services, healthcare, legal, and pharmaceuticals — this is particularly bad. Automatically generated assets can include outdated pricing, unsupported product claims, broad guarantees, or copy that does not match approved brand language. When that copy runs and a regulator or consumer takes issue with it, the advertiser answers for it. Not Google.

The new terms also narrow the 30-day liability cap to the specific advertiser account associated with a dispute, rather than the overall account spend. That is a meaningful shift for agencies managing multiple accounts under the same relationship.

For brand and compliance teams that have spent years managing message control, sequence, and approval workflows, the July 2026 update is not a technical footnote — it is a structural change to who controls what your company says to customers in paid search.

This Did Not Start Last Week

Anyone surprised by the July 2026 Terms of Service (ToS) update may not been following this closely. It is the formal, contractual codification of a direction Google has been moving for four years.

The first major signal came in July 2022, when Google deprecated Extended Text Ads (ETAs) and made Responsive Search Ads (RSAs) the only available format in search campaigns. Up to that point, advertisers had full control over which headlines and descriptions appeared in each ad, in what order, and in what combination. That control was handed to Google AI — down to the individual auction level.

At the time, this was described as an improvement: better testing, better performance, better use of machine learning. The advertising community’s skepticism was not technophobia. It was a reasonable request for evidence. The most important metrics — operating margin and conversion dollar volume — were never used to validate RSA superiority over ETAs. ROAS ratios were cited instead, which can mask serious deterioration in volume. A campaign can show an improving ROAS while total conversion revenue contracts sharply, and no one in a weekly status meeting raises a flag.

The lack of transparency and observability in RSA performance — which combination ran, in which auction, against which query, with what result — was never resolved. The July 2026 ToS update does not resolve it. It expands the automation and formalizes Google’s authority to use it.

The Data That Was Never Cited by Google

This was not speculation. In February 2022 — just months before Google made RSAs mandatory — Brad Geddes, a PPC technical pioneer and author respected industry-wide (co-founder of ad testing platform Adalysis), published a large-scale analysis comparing RSA and ETA performance across a substantial sample of ad groups containing both formats. The findings were difficult to reconcile with Google’s case for mandatory RSAs.

Across the ad groups studied, ETAs outperformed RSAs in conversion rate in the majority of cases. On CPA, ETAs outperformed RSAs in 67% of ad groups analyzed. On conversion per impression — a metric that combines click frequency and conversion rate, and arguably the most meaningful single measure of ad effectiveness — ETAs outperformed RSAs in 55% of ad groups. On ROAS, ETAs won by an overwhelming margin, which Geddes attributed directly to their higher conversion rates. Critically, when one format outperformed the other by a large margin, it was almost always the ETA.

RSAs did show higher CTR in many cases — but as Geddes noted, that advantage largely reflects RSAs appearing in a broader range of auctions, including lower-quality ones. More impressions across less relevant queries produces higher raw click volume. It does not produce better campaign performance.

The lack of transparency and observability in RSA performance — which combination ran, in which auction, against which query, with what result — was never resolved. Google deprecated ETAs anyway. And now, four years later, it is asking advertisers to trust that its AI will generate better copy than the format its own independent data suggested was outperforming.

What Advertisers Should Do Before July 1

The new terms take effect automatically. There is no opt-out. The only protection available to advertisers is operational: review what is running, understand what Google AI is generating on your behalf, and put controls in place before the deadline.

• Audit whether automatically generated assets are enabled in your active campaigns — and review what is currently running if they are
• Ensure all campaign inputs (headlines, descriptions, URLs, feed data) reflect current, accurate, and compliant information
• Identify any brand, legal, or regulatory requirements that must be enforced through pinning, exclusions, or asset restrictions
• Brief your compliance and legal teams on the new ToS structure — specifically that Google can generate copy from your site without per-asset sign-off
• Verify that conversion tracking is accurate enough to evaluate what the automation is actually producing, including operating margin — not just ROAS ratios
• Consider using Google’s AI Brief tool (introduced at Google Marketing Live 2026) to provide brand guidelines, goals, and audience context — treating it as a guardrail, not a handoff
• Establish a regular review cadence for automatically generated assets in all active ad groups

The Google AI Brief tool is worth using. Providing explicit campaign context — business goals, audience personas, brand language — gives Google AI systems better inputs and reduces the chance of off-brand or non-compliant copy appearing. But it does not replace human review, and the new terms do not reduce the advertiser’s review obligation. It is a guardrail, not a handoff.

The Bigger Question

Google has built one of the most profitable advertising platforms in history. It has also, over the past four years, systematically reduced advertiser control over how messages are assembled and delivered — first through RSAs, then through Performance Max, and now through ToS language that formally authorizes AI to generate ad assets from advertiser inputs.

Each step has been framed as improvement. Each step has made it harder to observe, test, and validate whether the automation actually performs better. The advertising community has not asked for proof because it is afraid of technology. It has asked for proof because that is how responsible campaign management works.

The question worth sitting with is this: if a single viable direct competitor to Google existed, would advertisers accept an arrangement where Google AI writes their ad copy, they bear full legal liability for it, and the algorithm that made the decisions is a black box? Or would they go elsewhere?

That question has no good answer right now. Which is precisely why the July 2026 Terms of Service update deserves more scrutiny than a one-paragraph email disclosure.