All posts

AI for Google Ads: What It Actually Does and What It Does Not

AI for Google Ads: What It Actually Does and What It Does Not

META TITLE: AI for Google Ads: What It Does and What It Does Not

META DESCRIPTION: What Google's AI genuinely decides in a Google Ads account, what it never touches, and the conversion volume Smart Bidding needs, sourced from Google docs.

SLUG: ai-for-google-ads

TAGS: google ads, smart bidding, performance max, ai advertising, shopify ads, ppc

AI for Google Ads means a specific set of features, not one button. Smart Bidding, Performance Max, broad match, recommendations and AI Max each hand a different decision to Google's models.

Good use looks like clean conversion data, enough volume to train on, and a human holding the guardrails. Bad use looks like switching on Target ROAS with six conversions a month and calling it automation.

This guide separates what Google's AI decides for you from what stays your job. Every claim below cites Google's own documentation, and the closing section covers where a Shopify-side tool fits.

What AI for Google Ads Actually Controls

Nothing inside a Google Ads account is labeled "the AI" and switched on once. Several separate features route different decisions to Google's models. Bidding carries the most weight, so start there.

Smart Bidding Reads Auction Signals You Cannot Reach

Smart Bidding, in Google's own words, means strategies that use Google AI to optimize for conversions or conversion value in every auction. Help documentation calls that auction-time bidding. Treat the description as a vendor claim. Four strategies qualify:

Signal access is where these strategies pull away from manual bidding. Per Google Ads Help, these strategies read identifiable attributes about a person or their context at the time of a particular auction. Device, location, time of day, query text, browser, operating system and remarketing list membership all sit in that set.

Seasonal performance trends feed in on Shopping. No manual bid adjustment reaches that far.

Target ROAS and Target CPA Need Different Starting Volume

Target ROAS predicts the value of a potential conversion every time someone searches. Bids then adjust to maximize return on those searches. Volume gates the whole thing.

Google documents a specific conversion threshold for each campaign type. Search and Shopping campaigns need at least 15 conversions in the past 30 days. Demand Gen asks for 50 in 35 days, with 10 of those in the past 7 days. Video Action wants 30 conversions in 30 days.

That first number decides most small Shopify accounts. Under 15 conversions a month, Target ROAS has almost nothing to learn from. Work out your profit floor with a break-even ROAS calculator before you set a target at all.

Target CPA is gentler on new accounts. It finds a bid from historical campaign data plus contextual signals present at auction time. Advertisers can start with no conversion history, according to the same help page.

Judging the result still takes volume. Google recommends evaluating over the last 30 days, including at least 30 conversions. Give the strategy that full window before you judge it.

How Performance Max Ranks Against Search Campaigns

Performance Max reaches Google's full inventory from one campaign, so placement choice moves to the model. We break the campaign type down in our Google Ads for Shopify guide.

Query priority, though, stays exactly where you put it. Exact match keywords in Search campaigns receive priority over Performance Max when a query matches exactly. Merchants often assume PMax will quietly eat their Search build, and that rule is the documented reason it does not.

Budget splits across channels are a separate question. Our comparison of Google Ads and Facebook Ads covers where each one earns its share.

Broad Match Is an AI Layer on Keyword Matching

Broad match no longer works on literal strings. Ads may show on searches related to your keyword, including searches that do not contain their direct meaning. Recent search activity, landing page content and other keywords in the ad group all inform the match.

Broad match is built to run with Smart Bidding. It also spends fastest when conversion tracking is wrong, which brings us to the limits.

What AI Does Not Do for Your Google Ads Account

Automation covers the decisions. Inputs, plumbing and judgment stay with you.

AI Does Not Fix Broken Conversion Tracking

Bad conversion data poisons every bid strategy downstream. Help pages name the usual causes directly: tagging issues, website outages and data import issues. A data exclusions tool exists for exactly those windows.

Even that tool carries a warning. You should not use data exclusions frequently or for prolonged periods, because doing so can hurt Smart Bidding performance. Fixing the tag beats excluding the week.

AI Does Not Set Up Merchant Center or Your Feed

Running Performance Max on a product feed requires a linked Merchant Center account. Google is blunt about getting that wrong. You cannot change the Merchant Center account after the campaign is created, and a different account means a new campaign.

None of that setup is automated. Feed approval, account linking and product data quality stay manual, before any AI bidding starts.

How Google Ads Recommendations and Optimization Score Work

Recommendations are generated automatically from account performance history, campaign settings and trends across Google. Optimization score sits beside them on a 0% to 100% scale. Google calls that score an estimate of how well your account is set to perform.

Each attached uplift number is an estimated impact Google may or may not deliver. Auto-apply exists for merchants who want the queue handled for them. One documented limit matters most, and it is a budget one.

Auto-applying recommendations will not increase your budget. Budget-limited accounts therefore still need a person watching. Review stays available by design, with a visible queue of what will run on a given day.

AI Cannot Show You Every Search Term

Search terms reporting has a privacy floor. Thresholds ensure user anonymity by only reporting terms with sufficient search volume across all Google searches. No AI layer, Google's own or a third party's, can surface what Google does not release.

Spend conclusions therefore rest on partial query data. Reading performance at account level helps, which is why blended metrics like MER matter alongside campaign reports.

What You Still Have to Do Manually

Three jobs stay manual, and they set the ceiling on everything the models can do. Margin sits underneath all three of those jobs. No bid strategy knows your contribution margin unless you tell it, so a CAC calculator beats a target pulled from a forum.

Enhanced Conversions and Data-Driven Attribution

Google's documentation says enhanced conversions improves the accuracy of conversion measurement by sending hashed first-party data. The same page positions it as unlocking more powerful bidding. Note the order of operations. Measurement accuracy is a prerequisite for AI bidding rather than an AI feature in itself.

Attribution belongs in the same bucket. Data-driven attribution distributes credit using your own account data. Google describes it as calculating the actual contribution of each interaction across the conversion path.

When to Use Seasonality Adjustments

Seasonality adjustments are a manual override for big, short conversion rate swings. Documentation calls them ideal for events of one to seven days. Beyond 14 days at a time, they may not work as well.

Ordinary seasonal movement is already handled by Smart Bidding. Reaching for the override every month works against the thing you are paying for.

How Negative Keywords Work With Performance Max

Account-level negative keywords apply to search and shopping inventory in Search, Performance Max, App, Shopping, Smart and Local campaigns. A limit of 1,000 negative keywords can be excluded for each account. Inside Performance Max, negatives apply to Search and Shopping inventory only.

That last constraint is easy to miss. Your negatives will not touch PMax placements on YouTube, Display, Discover or Gmail.

Where AI Max Fits in Search Campaigns

AI Max is an optimization layer inside Search campaigns. Google describes a suite of targeting and creative enhancements. Improved search term matching, better text customization of headlines and descriptions, and final URL expansion all sit inside it.

Google attaches an uplift number to the feature. Advertisers that activate AI Max in Search campaigns will typically see 14% more conversions or conversion value at a similar CPA or ROAS. Read that as vendor marketing rather than independent measurement.

Your own baseline is the only honest comparison. Incrementality testing is the method for getting one.

How Agency AI Handles the Google Ads Side

Agency AI: Meta, Google Ads AI reads your Meta and Google Ads accounts alongside Shopify. Recommendations then arrive measured against your break-even ROAS rather than against Google's optimization score. Shopify-side setup is covered in our Google Ads for Shopify guide.

Approval is the detail that matters on Google. Recommendation cards carry an Apply button, and the change reaches Google Ads when the merchant clicks it. That click is the human guardrail on an otherwise automated bid layer.

Agency AI publishes one customer result. Lolahemp's weighted average cost per purchase fell from $143.75 to $63.56, per the company's case study page. First window ran August 1 to September 30, 2025 across 33 purchases.

Second ran October 10 to November 6, 2025 across 95 purchases. Read it as one store's Meta result, and treat any Google Ads equivalent as unproven.

The MCP Plan for Claude and ChatGPT

A separate MCP plan connects the same data to Claude and ChatGPT. Its own description sets the boundary: it reads and explains, offering analysis, plans and clear next steps. It cannot spend money or change your ads.

Pricing there is $9 per month or $99 per year. Setup takes a couple of clicks, and our walkthrough on how to connect Claude to Google Ads and Meta Ads covers both assistants.

Chat answers are the point of that plan, so there is nothing to build or export. A free analytics dashboard now comes bundled with it as well, for the moments you would rather look than ask.

Frequently Asked Questions

How many conversions does Smart Bidding need before it works?
Fifteen conversions in the trailing 30 days is the documented floor for Target ROAS on Search and Shopping. Other campaign types sit higher. Demand Gen asks for 50 conversions in 35 days, with 10 of those in the past 7 days. Video Action needs 30 in 30 days. Target CPA will run from a standing start with no history, so judge it over a full 30-day window with roughly 30 conversions.
Can Google's AI change my campaigns without me?
Auto-apply recommendations can act on your account, though Google keeps the day's queue visible for review. One hard limit applies. Nothing auto-applied raises your daily budget, so a budget-limited account still needs a person to make that call.
Does Performance Max steal traffic from my Search campaigns?
Not on exact matches. Google gives an exactly matching keyword in a Search campaign the first claim on that query. Everything looser goes to the model, so your negative keyword list and campaign structure still decide a lot.
Why can I not see every search term that spent money?
Search terms reporting applies volume thresholds for user anonymity. A term too few people searched stays hidden, however much you spent on it. No outside tool recovers it, because no third party sees data Google withholds.
Is AI Max worth turning on?
Google publishes a typical 14% uplift for advertisers who switch it on, with CPA or ROAS held steady. That figure is the vendor's own marketing claim rather than an independent measurement. Test it against your own baseline before treating it as your expected result.

Sources

Free to install. Launch campaigns and pay nothing for your first 30 days.

Download Agency AI free on the Shopify App Store