How to Measure SEO and AI Search: From Visibility to Qualified Leads
Start with the short answer ↓
Measure SEO and AI search in four separate layers: visibility, visits, enquiries and qualified opportunities. Connect Search Console, GA4 and your CRM without treating rankings, mentions or clicks as sales.
Prepared with AI assistance. These are practical scoping recommendations; examples are illustrative, not client results.
Start with the business decision, not the dashboard
A search report should help you decide what to improve next: discovery, the landing page, the offer or the follow-up. Before choosing SEO KPIs, define a qualified enquiry with the people who handle sales. For example: a real business, a relevant need and an agreed next conversation. A newsletter signup does not automatically meet that definition.
Give every metric a definition, an owner and a time window. Record the service, landing page and language where available. Keep unknown values visible. Start with the few fields your team can maintain reliably; an elaborate dashboard built on incomplete records is not a better measurement system.
Keep four layers of evidence separate
Use this as a reporting contract, not a universal benchmark. Each layer answers a different question. Review it by service and page before combining the whole site: a research article and a service page can attract very different audiences.
| Layer | Evidence to record | Decision it supports |
|---|---|---|
| Visibility | Search impressions; separately, mentions in a documented AI prompt sample | Are relevant topics and pages being discovered? |
| Visits | Search clicks and measured landing-page sessions, kept as distinct counts | Which pages bring people into the site? |
| Enquiries | Accepted submissions or recorded conversations; remove tests, spam and duplicates | Can interested visitors reach the team? |
| Qualified opportunities | CRM records meeting your agreed qualification criteria | Does the offer attract businesses you can help? |
Use Search Console and GA4 for different questions
Search Console reports search impressions, clicks, click-through rate and average position. Compare a consistent page, country, device and date range. Average position is not one permanent place your company holds for every searcher.
GA4 Traffic acquisition uses session-level acquisition dimensions. Use it to examine measured visits and key events after people arrive. Do not silently substitute first-user acquisition for session acquisition or divide CRM leads by Search Console impressions and call the result a website conversion rate.
For the reporting process, maintain separate counts and document collection gaps, time zones and exclusions. Investigate a change in measurement before interpreting it as a change in demand. If a form event fires on a button click, test whether it also fires when validation fails; a click alone is not evidence that the enquiry arrived.
Measure AI visibility without inventing AI traffic
Google includes traffic from AI Overviews and AI Mode in the Web search type in Search Console. That does not make every Google organic visit an AI visit. Google also says these features do not require special AI schema or a separate AI text file.
For a repeatable AI visibility check, define a small set of buyer questions and record the service, prompt, language, date and citation returned. Keep the same sample when comparing periods. This is an observation of selected answers, not a census of all AI users, searches or clicks. Label any mention rate with its exact sample size.
Keep three source fields distinct in your CRM: a recorded referral when available, what the prospect says brought them to you, and your attribution confidence. A prospect may discover you in an AI answer and return directly later. If the evidence is missing, leave the source unknown instead of filling the gap with a claim.
Turn the report into a next action
Illustrative example, not an AgenticShip result or a benchmark: a service page receives 200 measured sessions in a month. There are eight confirmed enquiries, two duplicates and three qualified opportunities among the six unique enquiries. The qualification rate is 3 ÷ 6 = 50%. It is not 3 ÷ 200, and it does not prove which visit caused each enquiry.
If relevant impressions rise without clicks, review the query, title and page promise. If visits rise without confirmed enquiries, check the offer, mobile contact flow and collection before rewriting everything. If enquiries rise but qualification falls, tighten service scope and examples. If qualified opportunities stall, inspect response ownership and follow-up. These are investigation paths, not diagnoses from one metric.
Review operational issues weekly and compare business outcomes over a period long enough for your sales cycle. Record one change, its owner and the next review date. Avoid changing targeting, copy and qualification rules together if you want to understand what happened.
Protect the customer data behind the numbers
Google Analytics prohibits sending personally identifiable information such as email addresses and personal phone numbers. URLs and user-entered form content can expose that information too. Keep contact details and conversation notes in the access-controlled CRM, not Analytics event parameters.
Test the success event, duplicate handling, privacy preferences and CRM handoff before relying on the report. The goal is a small, trustworthy loop: an agreed definition, a captured result, a person responsible for the next step and a documented decision. More tracking is not automatically better tracking.
Primary references
The guidance above is AgenticShip's proposed approach. These references provide technical background for the relevant recommendations.