Stopped tracking clicks. Started tracking branded-search lift. Attribution finally makes sense.
Six months of GEO work and our 'clicks from ChatGPT' number was still in single digits. Looked like a flop on the dashboard. Meanwhile demos were climbing and Sales kept telling me prospects were saying 'I asked ChatGPT and you came up.' Disconnect.
What finally fixed reporting:
We stopped tracking AI citations as a traffic channel and started tracking them as a brand-search-demand channel. Pulled the cohort of branded searches in GSC, plotted weekly. Compared the 12 weeks before our AI visibility climb vs the 12 weeks after.
Branded search +47% (week-over-week growth nearly doubled). Direct traffic +22%. Demo requests citing 'I asked ChatGPT/Perplexity' went from 2 to 14 per month.
The citations don't drive clicks. They drive recall, and the recall drives the searches that drive the demos.
We now report two numbers to the board: visibility rate (lead indicator) and branded search delta (lag indicator). Stopped reporting AI referral traffic — it's noise.
Anyone else running this two-metric model? Curious how others are handling the attribution gap.
2 replies
- Petr VlčekFounder voice
Petr here — the branded-search lag of roughly 3-5 weeks after a visibility climb is consistent with what we see across B2B SaaS domains, and it's worth flagging to anyone reading this who expects same-week impact. Direct traffic moves first (within 7-10 days), branded search lags. If you stop tracking too early because clicks didn't move, you miss the actual signal entirely.
The two-metric model is the right framing. We've been pushing finance to accept 'visibility rate + branded search delta' for two quarters and it's finally landing. One add: we also track 'AI-attributed demo' as a self-reported field in the demo form. Free signal, sales people don't push back on adding one optional question, and it triangulates the GSC data nicely.