AI in Google Ads: automation and oversight

Laptop with advertising results and a gold control dial

By Patricia Jarmek Digital Advertising 7 October 2026

7 min read

Automation is no longer an optional layer added to mature Google Ads accounts. Smart Bidding can determine bids at auction time, AI Max can extend query matching, customize ad text and select a more relevant landing page, while Performance Max can optimize across multiple Google channels. [1], [2], [8]

The useful question is therefore no longer whether AI should be used.

It is:

Which decisions does the system have an informational advantage in making — and which decisions should a business never outsource to it?

Automate where machines have the informational advantage

A person cannot manually evaluate every available signal and calculate the optimal bid for every auction in real time.

Smart Bidding is designed for precisely that task. Target CPA, Target ROAS, Maximize conversions and Maximize conversion value use Google AI to optimize for conversions or conversion value at auction time. [1]

But an automated system optimizes the objective it is given.

If a business treats every contact-page visit as an important conversion, the system may become extremely effective at finding people who visit contact pages.

That does not mean it has found good customers.

Defining the conversion is therefore a strategic decision, not an algorithmic one.

Better automation requires better measurement

For ecommerce, purchases and their values need to be recorded correctly. For B2B businesses, a form submission may be only the beginning of the sales process. Where practical, feeding qualified-lead or closed-sale information back into the advertising system can give optimization a much more meaningful signal.

Google distinguishes between bidding for conversion volume and value-based bidding, which optimizes toward the business value assigned to conversions. For offline conversions, Google recommends a regular flow of data and says daily uploads are optimal. [6]

Automation does not make measurement less important.

It makes bad measurement more expensive.

AI Max changes Search campaigns without eliminating controls

AI Max combines search term matching, text customization and final URL expansion. As of September 2026, Google has begun automatically upgrading certain legacy Search settings, including automatically created assets and campaign-level broad match, while new Search campaigns have AI Max selected by default during creation. [3], [4]

Advertisers still have controls including brand inclusions and exclusions, URL controls and ad-group-level settings. [2]

The role of campaign management is therefore shifting from manually making every micro-decision to defining the guardrails within which automation can operate.

Landing-page automation is only as good as the website

Final URL expansion can be valuable when a website contains several strong pages serving different intents. Google may be able to route a user to a page that better answers a particular query.

But if the site contains obsolete pages, poor translations, irrelevant content or URLs that should never be used for paid acquisition, that same freedom creates risk.

The useful question is not simply whether Final URL Expansion is good.

It is whether the website is sufficiently well governed to give the system that freedom.

AI cannot repair poor information architecture. It can only work with it.

Automated creative still needs brand boundaries

Text customization can adapt headlines and descriptions to the user's intent. [2]

The system, however, does not inherently know every legal restriction, brand promise, commercial priority or phrase a company deliberately avoids.

Those inputs remain human responsibilities.

Google's own product direction reflects this need for guidance. AI Brief, powered by Gemini, lets advertisers provide richer context about their business, audience and messaging. On 23 September 2026, Google announced that the closed beta was expanding to additional languages, including German. [2], [5]

The better automation becomes, the more valuable precise strategic input becomes.

Do not judge automated campaigns only by their headline number

Automation has often been criticized for becoming a black box. Reporting is improving.

Performance Max now provides more channel-level reporting, fuller search-term reporting and richer asset metrics. [7]

On 23 September, Google also announced a new AI Max reporting view designed to connect the search term, creative assets shown and landing page used in one journey. Google said further availability details would follow later in 2026. [5]

Use those reports to audit what automation is actually doing rather than assuming that a good aggregate CPA means every underlying decision is sensible.

Human oversight does not mean daily interference

Constant intervention can make automated campaigns harder to evaluate.

For value-based bidding, Google recommends waiting one to two complete conversion cycles before comparing average target ROAS with actual ROAS and advises against repeatedly changing the target within a conversion cycle unless there is a genuine business reason. [6]

Oversight means distinguishing noise from a meaningful change.

There are exceptions. If conversion tracking breaks, Google provides data exclusions so Smart Bidding does not learn from faulty periods. If a short promotion is expected to cause a major conversion-rate shift, seasonality adjustments can inform Smart Bidding. Google says these are best suited to short events, typically one to seven days. [9], [10]

So what should be automated?

Good automation candidates include auction-time bidding, processing large numbers of contextual signals, discovering additional relevant queries, testing asset combinations and — on a well-maintained website — matching intent to a relevant landing page.

Human control should remain over business objectives, conversion definitions, conversion value, budgets, profitability, market priorities, positioning, brand promises, legal and reputational constraints, landing-page quality and the definition of a commercially good result.

The biggest risk is not using too much AI.

It is automating a decision that was never properly defined in the first place.

Google Ads AI can be an extremely capable executor.

It still needs someone to decide what success means.

Frequently asked questions

Should advertisers use Smart Bidding rather than manual CPC?

For conversion-focused goals, Smart Bidding can optimize each auction using Google AI. The right setup still depends on the business goal and reliable conversion measurement. [1]

Is AI Max mandatory for new Search campaigns?

Google currently selects AI Max by default during new Search campaign creation, but advertisers can choose its individual features or opt out. [3]

Are Dynamic Search Ads disappearing?

Yes, but the timeline changed. Google updated its plan on 11 June 2026: the DSA sunset and automatic AI Max transition will begin in February 2027, rather than September 2026. [4]

Does Performance Max still require management?

Yes. Automated bidding and distribution do not remove the need to inspect measurement, channels, search terms, assets, budgets and the commercial quality of conversions. [7]

Are more conversions always better?

No. When conversions have materially different business values, value-based bidding can optimize toward conversion value rather than simply volume. [6]