Google Ads’ July 2026 Terms Changes Reclassify Your Conversion Labels as Smart Bidding Training Data: Here’s the Account Split That Protects Them

Google’s July 2026 terms changes are not a legal footnote you forward to your lawyer. They are a Smart Bidding training-data event. Google notified advertisers on June 1, and the new terms take effect July 1, 2026 (PPC Land).

The press coverage is mostly right and mostly useless for an operator. Yes, the change is automatic. Yes, it “requires no action.” But the inputs Google is clarifying rights to are the same conversion and audience signals your lead-gen and pay-per-call accounts feed the algorithm every day. It lands in the same window as the Google Ads API v20 sunset, so operators tracking one platform deadline are already living through the other.

So the question isn’t whether you have to do something. It’s what you should choose to do before the labels you upload start pulling double duty as training data.

What the Google Ads July 2026 Terms Changes Actually Do, and Why ‘No Action Required’ Doesn’t Mean Do Nothing

The update clarifies that advertiser inputs can be used across Google’s automated tools to shape and improve your campaigns (Search Engine Roundtable).

The old terms treated automation as an optional feature you switched on. The new terms describe how Google Ads actually runs in 2026, where automation is the system, not a setting (ZATO Marketing). The language now covers what you type into Google’s tools, the URLs you hand over for setup, and the site content you let Google crawl.

Here’s the part the recaps skip. “No required action” is not the same as “no smart action.” Nothing in your account breaks at midnight on July 1. What changes is the role every signal you feed plays downstream.

A “qualified call” you mark for billing was always an optimization signal too. So the right response isn’t to sit still. It’s a deliberate audit of what you feed the algorithm, because the algorithm is built to learn from it.

Where Your Qualified-Lead Signal Actually Leaks: Enhanced Conversions for Leads and Customer Match

Your qualified-lead signal leaks into Smart Bidding through two specific doors: Enhanced Conversions for Leads and Customer Match lists. The press coverage never names them.

Enhanced Conversions for Leads lets you upload offline outcomes, like a closed sale or a qualified call, and match them back to the click using hashed customer data, a pathway that is itself migrating to the Google Ads Data Manager API (Google Ads Help). Customer Match does the same matching for audience lists you build from your own data (Google Ads Help). Both run on first-party data. Both feed the optimizer.

Now connect that to the update. A billable call label you upload to get paid is also a training target. Your qualification definition quietly becomes an optimization objective. Smart Bidding learns to find more of whatever you marked as the win.

That sounds fine until you look at what you actually labeled. If your billable event fires on a call that just crossed a duration threshold, you trained the model to chase duration, not sales. The optimizer does its job and fills your pipeline with calls that talk long and close nothing.

Pay-per-call and form-lead accounts leak differently here. A form account uploads a single lead event, so the risk is one over-broad signal. A pay-per-call account uploads call dispositions, and every tier you mark is a candidate training target. That’s why the disposition decision matters more for call accounts. (If your match rate is already soft, our Enhanced Conversions for Leads troubleshooting walkthrough finds the real leak.)

The Account Split to Make Before the Rollout: Lean Conversions In-Account, Rich CRM Outcomes in BigQuery

The fix is an architecture decision, not a toggle. Keep a lean billable event in-account, and hold your richer CRM outcomes in BigQuery where Smart Bidding can’t train on them. You push only the signal you actually want the model to learn.

Three moves make this real this week.

Keep one lean event in-account. Pick the single conversion action that defines a usable lead, like a billable-duration call or a verified form lead. That’s what you upload through Enhanced Conversions for Leads. Keep it clean and keep it singular. The more granular you get in-account, the more shapes the optimizer has to chase.

Hold rich CRM outcomes in the warehouse. Sale value, policy bind, booked-job revenue, true close data: route all of it to BigQuery and stop there. Don’t push it back through Enhanced Conversions or Customer Match. Your most sensitive qualification data is exactly the stuff you don’t want reclassified as a modeling input. You still report on it. You still build your real economics from it. You just don’t hand it to the auction.

Decide which disposition tier bidding should optimize toward. This is the pay-per-call call the recaps miss. A billable-duration call and a sold call are two different training targets. Optimize toward billable duration and you get volume. Optimize toward sold and you get fewer calls and a model that’s harder to feed but pointed at revenue. Make that choice on purpose before July 1, not after you notice the volume shift.

The two numbers this split protects:

• Cost per qualified call = total campaign cost ÷ qualified calls
• Call qualification rate = qualified calls ÷ total calls

When your in-account label drifts toward volume, qualification rate falls and cost per qualified call climbs even while the dashboard conversion count looks great. We’ve written before about how conversions jumping while qualified calls don’t is the classic sign the model is pacing toward the wrong lead. The update doesn’t cause that. It just removes any excuse for letting a careless label create it.

If you already route offline conversions through a call platform, the same logic applies to your call-tracking Enhanced Conversions setup: the match rate gets the attention, but the event you match is what trains the model.

Decide What You Feed the Algorithm Before July, Not After You See the Volume Shift

The cheapest version of this correction is the one you make in advance. Once your labels are working as inputs and the optimizer has spent a few weeks learning on them, you’re not editing a setting anymore. You’re unwinding a trained model and waiting out a relearn period while spend keeps moving. Decide what you feed it now, while the only cost is an afternoon of audit.

We’re media buyers and lead-gen operators sharing what we see in the field. This isn’t legal advice. How a platform’s terms apply to your data and contracts varies, so talk to an actual attorney before you change how you handle first-party data or vendor agreements.

If you want a second set of eyes on what your account is actually feeding Smart Bidding, that’s the kind of audit we run. Book a free consultation and we’ll walk your conversion-input setup and first-party data handling before the July rollout, not after the volume shift shows up in your qualified rate.