Record what the source evidence actually shows
Identify whether the information comes from a recorded referral, a campaign parameter or the customer saying they used an assistant. These are different observations. In a fictional business, a prospect says an AI tool helped them research suppliers but then finds the company through another route. Record that account without inventing a direct referred visit. Equally, a visit associated with an AI service does not prove that the business was cited, recommended or described positively in an answer.
Connect the enquiry only where supported
Use the available enquiry reference and approved analytics arrangements to establish whether a visit and request can be linked. Do not assume two events belong to the same person merely because they happened close together. Where a reliable connection is unavailable, report the observations separately. Avoid collecting unnecessary personal information just to fill an attribution gap. Ask the person responsible for measurement to explain what the implementation records and what it cannot establish before using the figures in a results report.
Read the actual request
Assess whether the enquiry concerns a service the business offers, the relevant location and a next step staff can take. Keep those criteria explicit rather than calling every submitted form a qualified lead. In the fictional supplier example, an enquiry about an unavailable service may still be a real person but does not demonstrate useful demand for the current offer. Record duplicate, irrelevant and unresolved requests separately. The quality review should help the team understand enquiries, not improve a headline by excluding inconvenient cases without explanation.
Keep customer recollections in context
If staff ask how someone found the business, make the question neutral and optional. Record the answer as reported rather than translating every mention of AI into a named platform referral. A customer may not remember the exact tool, answer or sequence. Do not press them for screenshots or personal conversation history just to support a marketing claim. Where they voluntarily provide a relevant source, handle it through the business’s approved process and distinguish that evidence from assumptions made by the reviewer.
Compare outcomes with a clear denominator
Report the number of reviewed enquiries and the subset with each type of source evidence. Separate requests received, suitable enquiries and later business outcomes. Use comparable periods and explain gaps or a small sample. An increase from one enquiry to two is limited evidence even if the percentage looks large. Consider other changes in advertising, seasonality or the offer before attributing an outcome to AI visibility work. Do not turn the value of an enquiry into revenue before the business has established that result.
Use findings to improve the next step
Review whether the landing page explained the service and whether staff could respond to the requests received. Correct misleading or missing information through the normal content process. Keep a record of what changed and why, then continue observing rather than declaring a causal result immediately. The useful outcome may be a clearer contact path or better qualification information. State the limits of referral evidence in the report so decisions are based on actual enquiries rather than an assumed recommendation count.