How accurate is AI search attribution, and how do you prove lift?
Attribution tells you who got credit. Incrementality tells you what changed because of your work. Test one change on a group of pages and keep another group unchanged as a control. Report accuracy as a range, with counted numbers marked separately from estimates.
organic click rate for brands cited vs not cited in an AI Overview
Seer Interactive, Sep 2025paid click rate, cited vs not cited
Seer Interactive, Sep 2025of visits with a click on a result link, with vs without an AI summary
Pew Research Center, Mar 2025Pages you improved pulled away from pages you left alone, which is the lift that attribution cannot show
Sample 12 weeks of AI referred visits, indexed to week 1 = 100. Content was updated in week 6.
Sample data for illustration. Your own numbers will differ.
How to read it. Both groups move together before the change. The gap after it is the evidence that the work caused the lift, not just the season.
In ThriveStack citedby Citation tracking and AI referred visits are charted per page, so you can compare changed pages with unchanged ones. See it for your site →
Attribution says who got credit, and incrementality says what changed.
Credit is a rule. Lift is a result. You need both before you move budget.
- Assigns credit by a rule
- First touch, last touch, survey
- Measures what changed
- Test group vs control group
Test one change on a group of pages and keep another group as a control.
Update the group in week six. Leave the control alone. The gap after the change is the lift.
- 1
Pick a test group and a control group of pages
Pages that are alike in traffic and topic.
- 2
Change only the test group
Update content or structure in a single week.
- 3
Compare the two groups for 6 to 8 weeks
Watch citations and AI referred visits for both.
Correlation is not proof, and the studies say so.
Seer found organic click rate was 0.70% for brands cited in an AI Overview and 0.52% for brands not cited. The study notes this is correlation, not proof of cause.
organic click rate, cited vs not cited in an AI Overview
Seer Interactive, 3,119 terms, 42 organisationspaid click rate, cited vs not cited
Seer Interactive. Correlation, not proof.Report attribution accuracy as a range, not one number.
A counted number and an estimate are different kinds of evidence. Label them.
Three kinds of evidence, labelled by strength
| Evidence | Strength |
|---|---|
| Counted referrals | High |
| Survey answers | Medium |
| Modelled dark traffic | Low, always a range |
ThriveStack citedby runs your prompt panel across 10+ AI engines, reads server-log crawler hits, matches human visits to signups and reports AI-sourced and AI-influenced pipeline.
AI attribution incrementality, answered.
How accurate is AI search attribution?
Counted referrals are accurate for what they count. Dark and influenced traffic are estimates and should be reported as ranges.
What is incrementality?
The lift caused by a change, measured against a control group.
Can I prove AI caused revenue?
Not with attribution alone. A test with a control group gets closer.
Sources for this lesson
- Seer Interactive: AI Overview impact on Google CTR, September 2025 update 3,119 terms, 42 organisations
- Pew Research Center: Google users are less likely to click when an AI summary appears (Jul 22, 2025)
Figures were checked against each page in Nov 2026. Vendor studies are marked by their sample. Charts marked sample data are illustrations.
Questions this lesson answers
See all 10 questions in this cluster
- How accurate is AI search attribution?P1
- AI search attribution confidence scoringP1
- Causal incrementality of AI searchP1
- AI search holdout experiment designP1
- AI search marketing mix modelingP1
- Correlation vs causation in AI visibilityP1
- AI attribution false positivesP1
- How to validate self-reported AI attributionP1
- AI attribution data quality auditP1
- AI search attribution statistical methodsP1