The click was never the goal. It was the proxy
Marketing measurement has run on the same assumption for twenty-five years: if something influenced a buyer, it produced a click. An ad got an impression, the buyer clicked. A search result answered a question, the buyer clicked through to confirm it. A UTM parameter tagged the journey so a report could later say which channel deserves the credit. None of that was ever the actual goal. It was a convenient stand-in for a moment you couldn't otherwise observe: the buyer decided you were worth trusting.
Now the buyer asks ChatGPT or Perplexity the same question they used to type into a search box, gets a complete, confident answer with your name in it, and never opens a browser tab. If the answer was good enough to end the research, there is nothing left to click. Your marketing did exactly what it was supposed to do. It just didn't leave the artifact your reporting stack was built to find.
The stage that goes dark first is the one that mattered most
Attribution models can only assign credit to a channel that shows up in the path. Last-click, first-click, multi-touch, it doesn't matter which one you run: they all require an interaction to attach a source to. Zero clicks means zero data points, and a model can't distribute credit across events that never got logged. So budget keeps flowing to the channels that still generate a click, mostly paid search and retargeting, while the channel that actually swung the decision gets nothing, because it happened inside a conversation your tools can't see.
The cruelest part is which stage disappears first. It's not top-of-funnel awareness content, which was always hard to attribute anyway. It's the comparison and decision stage, the exact moment a buyer used to land on your pricing page or your head-to-head post before choosing. That's precisely where an AI assistant now resolves the question in one turn. The highest-intent, highest-value part of the journey is the part going dark, and it's growing every quarter, not shrinking.
Replace the click with presence, citation, and share of the answer
The fix isn't a better attribution model. It's a different object to attribute. A click was always a proxy for three underlying facts: were you present in the moment of decision, did the buyer trust the source describing you, and how much of that decision did you own versus a competitor. You can measure all three directly now, without needing a click to happen first. Presence is whether the assistant names you at all. Citation is whether it points to your source, not a rival's, as the reason it trusts the claim. Share of answer is how much of the recommendation you get relative to everyone else named in the same response.
This is what Crescive measures instead of waiting for a click that isn't coming. It runs your buyers' actual questions across ChatGPT, Perplexity, Gemini, Claude, and the other engines they use, records whether you're present and cited on each one, and tracks your share of the answer over time the same way a rank tracker used to track position on a results page. The difference is that this metric doesn't evaporate the moment the buyer stops clicking, because it was never built to depend on one.
Report the lift in a language leadership will actually accept
Start by pairing what you can already see with what you now measure directly. If direct and branded traffic climb while trackable referral sources stay flat, that gap is not noise to explain away, it's a rough proxy for zero-click influence, and presence and citation data on the same prompts over the same window fills in the mechanism. Better still, run the same discipline you'd apply to a paid campaign: capture a baseline on the prompts that matter before you ship a fix, then re-measure after. A jump in citation share on "best expense management software for a 50-person team" the week after you publish a comparison page is a result you can defend, click or no click.
The attribution model isn't broken beyond repair. It just needs a new unit to attribute, the answer instead of the click. Marketing teams that keep reporting on clickstream alone will keep explaining away their best quarters as unexplained direct traffic. The teams that start reporting presence, citation, and share of answer next to it will be the ones who can finally say what actually moved the number, and prove it.
Key takeaways
- Click-based attribution assigns zero credit to marketing that fully satisfies a buyer inside an AI answer, even when it worked exactly as intended.
- The comparison and decision stage, historically the highest-value part of the funnel to attribute, is the part going dark first as AI answers resolve it in one turn.
- Presence, citation, and share of answer measure the same moment a click used to stand in for, and none of them require a click to exist.
FAQ
What is the zero-click attribution problem in AEO?
The zero-click attribution problem is the breakdown of click-based marketing measurement when a buyer gets a complete answer inside an AI assistant like ChatGPT or Perplexity and never visits a website. Classic attribution models require a click or session to credit a channel, so marketing that successfully influences a buyer through an AI answer shows up as no data at all, not as a measured result.
How do you measure marketing impact when there's no click to track?
Measure the underlying facts a click used to stand in for directly: whether an AI assistant names your brand (presence), whether it cites your source as the reason it trusts the claim (citation), and how much of the recommendation you get relative to competitors named in the same answer (share of answer). Capturing a baseline on key buyer prompts before a change and re-measuring after lets you attribute a specific lift to a specific action without needing a click in between.