Google Analytics is finishing a cleanup it started years ago, and the timing could not be worse. Starting in mid-July 2026, four attribution models, first click, linear, time decay, and position based, stopped being selectable for any new conversion action in GA4. By September 2026 they disappear completely. Any conversion action still running one of them gets migrated automatically to data-driven attribution. Google is not asking permission or waiting for anyone to click through a warning. The switch to the remaining models runs on its own timetable, regardless of whether an advertiser ever logs in to review it.
This is not a new idea, just a delayed one. Google Ads removed the same four models back in 2023, when then-product lead Ginny Marvin said the change was justified because fewer than 3 percent of Google Ads conversions were still being attributed using first click, linear, time decay, or position based models. Analytics is now catching up to Ads. Once September passes, marketers are left with exactly two choices: data-driven attribution, a probabilistic model that compares converting and non-converting paths, or last click, which ignores everything upstream of the final touch.
Marketers are not asking for fewer models. They are asking for visibility Google cannot sell them. The AgencyAnalytics 2026 Marketing Agency Benchmarks Report, based on 494 agency professionals surveyed between February and April 2026, found that 48 percent of agencies name tracking AI-driven discovery, prospects who find a brand through ChatGPT or AI Overviews, as their hardest attribution problem. A close 44 percent say traditional attribution models are losing reliability outright. Those two numbers sit within four points of each other, which is its own answer: the complaint is not about which model to pick, it is that whole stretches of the customer journey never generate a trackable path to begin with.
Multi-touch attribution is growing anyway, into a blind spot. Separate 2026 survey data covering more than 1,200 B2B teams, cited by Forrester, Bizible, Dreamdata, and HockeyStack, puts multi-touch attribution adoption at 47 percent, up from 31 percent in 2023. More teams are building models that need a sequence of touchpoints to compare. Data-driven attribution works the same way: it needs converting and non-converting paths to weigh against each other. Neither approach has anything to measure when the first real touch happens inside a chat answer that never fires a pixel or logs a referrer.
What September actually changes is smaller than it looks. Removing four models nobody used clears out dead weight, it does not close the AI discovery gap. Agencies dealing with the 48 percent problem are patching it with brand lift surveys, server side tagging, and direct questions on lead forms rather than waiting for Google’s remaining two models to catch up. For attribution and consulting teams, the practical read is to treat the GA4 migration as housekeeping and put the real budget toward measuring the part of the funnel Google’s dashboard still cannot see.
Sources referenced in reporting: