Google Ads Just Removed Your Attribution Model: What Changed and the 15-Minute Check

What actually changed

In mid-July 2026, Google removed first click, linear, time decay, and position-based attribution from Google Ads. They are no longer selectable models for conversion actions. Two options remain: data-driven and last click. Any conversion action still set to one of the retired models is being migrated automatically, to data-driven where the account has enough conversion volume to qualify, and to last click where it does not.

Here is the part that matters for operators: this happens at the conversion action level, silently, with no change to your campaigns, your budgets, or your dashboards. If nobody on your team opened the conversion settings this month, the account looks untouched. But the math that decides which campaign gets credit for each conversion, and the signal that Smart Bidding trains on, changed underneath you.

Who gets hit

Most accounts moved to data-driven attribution years ago, when it became the default for new conversion actions. The migration hits the accounts nobody has looked at closely in a while:

• Inherited accounts. You took over the account last year. The conversion actions were created by two agencies ago, and nobody has audited their settings since. This is exactly where a forgotten position-based model has been quietly steering credit.
• Legacy conversion actions. Actions created before 2022, when marketers picked models deliberately. Someone chose time decay for lead gen in 2019 for a good reason. That reason left the company. The setting stayed.
• Agencies managing portfolios. If you run thirty accounts, you almost certainly have a handful configured pre-2022. You will not know which ones without checking every conversion action, and "a handful of accounts" is exactly how a portfolio-wide reporting surprise starts.

Why your numbers move when a model changes

An attribution model does one job: it splits credit for a conversion across the clicks that led to it. Change the model and the same conversions get re-divided. Position-based gave 40% of the credit to the first click; data-driven might give it 12%. Your totals barely move, but credit shifts between campaigns, and that is where two things break.

First, reporting. Your prospecting campaigns suddenly "lose" conversions and your brand campaigns "gain" them, or the reverse. Nothing changed in the market. The scorekeeping changed.

Second, and more expensive: Smart Bidding trains on attributed credit. Target CPA and target ROAS make bid decisions based on which clicks earn conversion credit. Redistribute the credit and the bidding algorithm starts re-learning where to spend, which means a re-learning period, which means volatility in exactly the weeks your client is comparing against last month. If your GA4 and Google Ads numbers already disagree, this adds a third version of the truth. We covered why those two never match in Why GA4 and Google Ads Never Match, and attribution settings are a big piece of that story.

The 15-minute check

Run this on every account you touch, this week:

1. Open the conversion actions list. Goals > Conversions > Summary. Make sure the attribution model column is visible in the table.
2. Inventory anything legacy. Sort by model. Anything that still shows a retired model is queued for migration; anything that recently flipped to data-driven or last click shows up in the account change history. Note every affected action and which campaigns feed on it.
3. Screenshot the before-state. The settings page and a campaign-level conversion report for the trailing 30 days. Change history tells you a setting changed. It does not show you what your reports looked like before the credit moved. Screenshots do.
4. Annotate the migration date. Everywhere reporting lives: Looker Studio, the client dashboard, the monthly deck template. One line: "Google migrated attribution models on this date; conversion credit redistributed."
5. Flag the Smart Bidding campaigns. List every tCPA and tROAS campaign that optimizes toward an affected conversion action. Expect movement for a couple of weeks and resist the urge to chase it with target changes while the algorithm re-learns.

The reporting trap: August comparisons will lie

Every week-over-week and month-over-month comparison that straddles the migration date is comparing two different credit systems. Campaign A did not actually lose 30% of its conversions in the first week of August. It lost credit it was never really earning under the old model.

How to handle it in client reporting:

• Break trend charts at the migration date, or at minimum drop a labeled reference line on it.
• Compare pre-migration weeks against pre-migration weeks, and post against post. Do not narrate a trend across the seam.
• Anchor on totals, not per-campaign splits, for the transition month. Total conversions are roughly stable; the split is what moved.
• Put the caveat in writing once, in the report itself, so it is on the record before anyone asks.

What we do when platforms force a migration

When a platform pushes a change like this, we run the same play across every client account at once: inventory the affected configurations, capture the before-state, annotate every reporting surface, and watch the bidding impact through the re-learning window. Our monitoring stack, TagPipes, watches tracking configuration continuously, so a silent settings migration gets flagged the week it happens instead of discovered in a quarterly review. That is the whole reason portfolio-wide monitoring exists: platforms change things without asking.