Attribution is broken for AI search. Here’s what to use instead
Attribution was built for a world of cookies and clicks. That world is leaving. If you’re still using click-based models to justify your non-paid marketing budget, you’re measuring a shrinking fraction of what actually drives demand.
Kevin Indig at Growth Memo wrote a sharp piece on this with input from George Bonaci, VP Growth at Ramp. Ramp processes over $100B in annualized payments for more than 70,000 businesses and hit a $44B valuation on a $750M round. When their head of growth says attribution is a crutch, that’s worth paying attention to.
Why Attribution Is Failing Right Now
Three platform shifts are happening at the same time, and each one removes a data point you used to rely on:
- Google AI Overviews and ChatGPT answer questions without requiring a site visit. Users get the answer. You lose the referral, the click, and the conversion signal.
- Google Analytics increasingly models conversions rather than observing them directly. Consent restrictions and fragmented journeys make direct measurement impossible in a growing share of cases.
- Apple’s 2021 opt-out for cross-app tracking blinded advertisers to which ads drove purchases. Meta responded with modeled conversions, replacing observed behavior with estimates.
The deeper problem: attribution can only assign credit for the parts of the customer journey it can see. As more discovery shifts to AI-generated answers, attribution explains less and less of what drives demand.
According to Graphite, AI can be underattributed by as much as 10x. That doesn’t mean you throw attribution out. It means you stop treating it as the primary instrument.
The Classic Journey Is Gone
The old model was: search, click, convert. It was always messier than it looked on a dashboard, but clicks gave you something to anchor on. The new model is: prompt, synthesize, direct visit. No referral. No click trail. Just someone showing up already sold.
Bonaci frames it directly:
Attribution and direct measurement in general has become a crutch replacing critical thinking. Every attribution model is wrong, there is not a perfect one and this is not a solved problem, which is why it should be used as a tool to align incentives and the pros and cons of the model must be top of mind.
His team at Ramp still uses a multi-touch attribution model and still sets org-level goals against it. But according to Bonaci, there isn’t an hour that goes by where the nuance of that model isn’t discussed and channel-level impact isn’t measured separately. Attribution as alignment tool, not as truth.
Two Better Frameworks
1. Triangulation
Instead of one metric, measure three signals with different blind spots simultaneously. For AI visibility, that could mean AI Share of Voice, self-reported discovery or CRM tags, and conversions. When all three move together, your confidence rises. When they diverge, you know where to investigate.
One concrete example from the piece: a user behavior study found a medium-strong correlation between AI Share of Voice and conversions. Most of those conversions show up as direct traffic with no clear source. Measuring Share of Voice gives you a proxy for the upstream demand you can’t directly observe.
Triangulation is already common in mature measurement teams. A BCG survey of 3,000 senior measurement professionals found that 46% use marketing mix modeling, incrementality testing, and multi-touch attribution together. Leaders using that integrated approach achieve up to 70% stronger revenue growth.
2. Incrementality
Attribution tells you who gets credit. Incrementality tells you what actually works. The core question is: what happened because of this activity, compared to what would have happened without it?
The main methods the piece covers:
- Randomized holdouts: Withhold the activity from a randomly selected group. Compare outcomes against the exposed group.
- Geo experiments: Run the activity in selected markets. Compare lift against similar markets where nothing changed.
- Phased rollouts: Roll out across markets or pages at different times. Use the waiting groups as temporary controls.
- Switchback tests: Alternate between active and inactive periods and compare outcomes while controlling for recurring patterns.
- Quasi-experiments: When randomization isn’t possible, construct a counterfactual using matched controls, difference-in-differences, or synthetic controls.
- Marketing mix modeling (MMM): Estimate contribution of broad, overlapping channels from historical variation. Calibrate with experiments where possible.
Incrementality is already widely used. The IAB found that 76% of U.S. buy-side decision-makers use incrementality tests, slightly more than attribution analysis at 73% and MMM at 67%. But only 39% use all three together, leaving most teams with individual tools rather than a coherent system.
The Bigger Shift: Unmeasurable Work Gets More Valuable
This is the part of the piece that has the longest shelf life. Bonaci’s observation at Ramp cuts to the core of where marketing value is moving:
Almost everything directly measurable is becoming commoditized thanks to AI. Most measurable work is no longer differentiated and the least measurable work is not valued. CMOs and CFOs will see how the alpha is in things that are not directly measurable by things like attribution models.
The practical consequence, in Bonaci’s words:
Just looking at attribution alone would have led us to cancel all the brand marketing we do, all our stunts, all our direct mail, all our events. Instead we would be plowing more and more money and time into generic outbound campaigns and simplistic paid ads. This would have been fine for a while but the results would have been very average and we are trying to be great, not average.
If AI automates everything that’s directly measurable, the advantage moves to creative, original, hard-to-quantify work. Attribution models, by design, cannot capture it. That doesn’t make it less valuable. It makes the case for better measurement infrastructure so you can defend it anyway.
When This Works
Triangulation and incrementality are most useful when your highest-impact channels (brand, content, events, AI visibility) don’t produce clean click trails. If your growth is mostly paid acquisition with clear conversion tracking, your existing attribution model is probably fine for now.
When It Does Not
Neither framework establishes causality on its own. Triangulation strengthens evidence but can’t prove a channel drove an outcome. Incrementality requires controlled conditions that are often difficult to set up cleanly in fast-moving campaigns. Both demand more analytical lift than plugging into a dashboard.
The point isn’t to replace attribution with something perfect. It’s to stop using it as a substitute for judgment about channels that were never designed to produce clicks in the first place.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
