Start with a comparable sample.
Save the exact question, answer, platform, mode, date, region and visible sources. Distinguish an explicit recommendation from a mention. Keep each platform identifiable rather than blending everything into one score.
Report the number of comparable observations alongside any percentage. Explain changes to the question set or collection method before comparing periods.
Inspect the visits you can observe.
In your available analytics, examine recorded referral sources, landing pages and completed enquiry actions. Check that a test enquiry is recorded correctly before using conversion counts.
Do not assign all direct traffic to AI. Equally, no recorded AI referral does not establish that AI had no influence: a person may learn your name and visit later through another route.
Google includes traffic from its AI Search features in overall Web performance reporting in Search Console. This is not a dedicated Gemini referral report. Google’s measurement guidance explains the reporting scope.
Ask what led to the conversation.
Add an optional question to your enquiry process: “How did you first hear about us?” Offer familiar channels, an open answer and “Not sure.” If someone names an AI assistant, ask what they were looking for, without requiring a private conversation.
Define qualification before reviewing results: a relevant need, a market you serve and a plausible fit for your offer. Apply the same criteria throughout the review.
| Record | What it tells you |
|---|---|
| Recorded source | What your configured analytics can attribute. |
| Self-reported discovery | What the prospect recalls, including uncertainty. |
| Enquiry fit | Whether the need matches your agreed qualification criteria. |
| Other activity | Campaigns, seasonality or other changes that could affect the period. |
End the review with a decision.
Log content changes and publication dates next to comparable AI observations, recorded referrals and qualified enquiries. Keep self-reported discovery separate from technical attribution; neither should silently overwrite the other.
Decide whether to continue, adjust a specific change or gather more evidence. A useful report can inform the next action without claiming that every new enquiry is AI-generated revenue.
Keep a practical review log
Blank fields for changes, observations, enquiries and decisions.
Measure the progress that matters to you.
Start with the systems you already use, the customers you want and the enquiries you need to understand.
Talk about your visibility