As of mid-July 2026, four attribution models are gone from Google Ads: first click, linear, time decay, and position-based. New conversions can no longer use them. By September 2026, any conversion action still configured with one of these four gets automatically migrated to data-driven attribution — whether or not your account is ready for it.
Two models remain: data-driven attribution (DDA) and last click.
What actually changed
Google framed this as cleanup, pointing to data showing fewer than 3% of conversions were using these models. That's probably accurate. But low adoption doesn't mean the people using them had no reason to.
The decision fits a larger pattern: remove advertiser-controlled signal inputs, route everything through automation, then point to aggregate performance numbers as proof it worked. The attribution model removal arrives one month before the August 17 Target CPA bidding changes and three months before the September AI Max migration. It's not isolated.
For most accounts, this is genuinely a non-event. If you were already on DDA or last click, nothing changes. The 3% stat holds.
The 3% who need to pay attention had reasons for their setup — and some of those reasons still apply.
The data-driven attribution problem low-volume accounts are about to run into
Data-driven attribution works by analyzing your actual conversion paths and distributing credit based on which touchpoints actually influenced the outcome. It's better than the rule-based alternatives — when you have enough data to train the model.
Google's own guidance recommends a minimum of 50 conversions per month per conversion action. Accounts below that see DDA behave erratically. Credit distribution shifts month to month without any obvious cause. The model is essentially guessing when data is thin, and it signals that uncertainty into Smart Bidding as if it were reliable.
A DTC brand generating 25 purchases a month in Google Ads doesn't have the conversion volume to run DDA with any stability. They'll be on it from September regardless. Their Target ROAS or Target CPA campaigns will inherit whatever signal DDA produces — noisy or otherwise.
For most mid-size e-commerce accounts, 50 conversions/month per action is achievable. But not every account is there, and some that are close run multiple conversion actions that individually fall below the threshold.
How this connects to Smart Bidding
Attribution and bidding aren't separate. Target CPA and Target ROAS consume your conversion signal to decide how much to bid in each auction. If DDA redistributes credit unpredictably because your volume doesn't anchor the model, Smart Bidding treats that noise as ground truth and builds bid decisions on top of it.
This is one of the ways accounts see unexplained performance shifts in Q4 with no obvious cause. Attribution model changed, bidding signal changed, and nobody connected the two because no single metric spiked.
That doesn't make DDA wrong. It makes DDA unstable in accounts where conversion volume is too thin for the model to be confident.
There's a second issue for accounts running PMax alongside Search: these campaigns have historically handled multi-campaign credit allocation inconsistently under rule-based models. DDA is actually better here — it distributes signal between campaign types more accurately when PMax and Search compete for the same funnel. So for those accounts, this change is net positive.
What to do before September
Check your active conversion actions now. In Google Ads: Tools > Measurement > Conversions. Any action still using first click, linear, time decay, or position-based will either show a flag or be unavailable for new conversions. Switch them manually — don't leave it for Google's auto-migration.
If your account generates 50+ conversions per month per action, switch to DDA. Mark the date, note your CPA and ROAS from the prior 90 days, and set a 45-day review window. Attribution changes are hard to separate from seasonal noise without a documented baseline.
If you're below the threshold, switch to last click. It's simpler, more predictable, and maintains consistent bidding signal. You lose some cross-channel visibility but you keep a stable anchor for Smart Bidding.
One specific thing to document now: your current bid strategies, their stated targets, and their recent actual performance. If something breaks in September or October, you need a clean pre-change record to diagnose against.
Google will not automatically alert you before the September enforcement. Notifications went out earlier in the year. The migration runs silently on September's schedule regardless of whether you acted.
The check takes five minutes. The consequence of missing it can take weeks to diagnose.
If you want to see how your current attribution setup is interacting with your Smart Bidding signal quality, the free audit shows where the gaps are. [Link to /audit]
Sources: Google Ads Help, Search Engine Journal, Boostify, ALM Corp, July 2026
