Measure AI search visibility without confusing activity with demand.

For this guide, AI search visibility means the observed presence of a business or its pages in a defined set of AI-assisted answers. It is a sampled observation, not a complete market share figure. Traditional search impressions and clicks belong in the same review, but they should remain separately labelled.

Google says its existing SEO practices apply to AI Overviews and AI Mode; there is no special additional optimisation requirement. Eligible supporting pages must be indexed and eligible for a search snippet, and inclusion is not guaranteed. See Google's guidance on AI features.

Begin with the commercial question. For a business selling CRM implementation, it might be: “Are people finding our explanation of lead follow-up, and does it help suitable buyers request a conversation?” That is more actionable than trying to increase an undefined AI score.

Google includes AI feature traffic within the Search Console Web performance reporting described in its measurement guidance. Do not present an ordinary Web click total as a separately isolated AI Overview total.

Use CDN and server logs for delivery, crawler and security investigation. Raw request totals include monitoring, assets and automation. They are not a substitute for human visits or sales opportunities.

Select questions that represent the decisions your customers make. For the illustrative CRM business, include a definition question, a workflow question, a comparison question and an implementation question. Keep the exact wording and report the sample size.

• “What is a CRM lead follow-up workflow?”
• “How should an AI agent get approval before sending a follow-up?”
• “When should a small team choose n8n instead of Zapier?”
• “What information do we need before connecting website enquiries to a CRM?”

Record the answer system, available model or mode, date, region, login state and prompt. Separate a brand mention from a clickable citation. Capture the cited destination, not just the domain. Repeat under comparable conditions and retain earlier observations; answers can vary even when the wording stays the same.

Use manual review to check whether a citation supports the statement beside it and whether the page is actually useful to a buyer. A mention that misdescribes the product is a content accuracy task, not an automatic win.

Observation ID: OBS-DEMO-006
Question group: CRM follow-up design
Exact prompt: How should a CRM stop obsolete follow-ups?
Observed system / mode: record the actual system used
Observation date / region: record when and where
Brand mentioned: yes / no
Marsen URL cited: exact URL or none
Answer accuracy: supported / mixed / unsupported
Evidence: saved observation reference
Landing page action: proposed edit, owner and review date

This is a template, not an actual observation. Do not populate it with assumed mentions. Keep raw observations separate from the interpretation so a teammate can review the same evidence.

For a sample-level citation rate, divide observations with at least one qualifying citation by all completed observations in that defined sample. Report the numerator and denominator. Do not call that percentage your share of the entire AI search market, and do not compare it with a differently composed sample without explaining the change.

Connect each recurring question to one useful destination. A product page explains what the product does, who it serves and what implementation requires. A guide shows the process. A comparison helps someone choose. Linking these pages gives the reader a clear route from question to decision.

Suppose the follow-up guide earns impressions but sits far from the first results. Review its query intent, example quality, internal links and competing explanations before only rewriting its title. If people reach the page but do not continue, check whether the next action matches the question they came to answer.

Use descriptive headings, a direct answer near the beginning, concrete examples and visible limitations. Add sources where factual claims depend on an external system. Keep product facts consistent across the product page, guide and contact conversation.

Structured data should describe the visible content. It can help Google understand a page, but eligibility does not guarantee a rich result. Validate any markup using the feature-specific rules in Google's structured data introduction. The Marsen schema generator can help prepare markup; it does not replace content review or indexing checks.

Track intent and completed outcomes as different events. Opening the contact page is an intent signal. Preparing an email is another intent signal; it does not prove that the email was sent or received. A booking page view does not prove that an appointment was confirmed.

Use received enquiry records and confirmed booking records as the authoritative business events. Apply a documented qualification rule after review, such as relevant service need, fit with your scope and an agreed next step. Keep spam, duplicates, internal testing and monitoring outside the qualified funnel.

Where consent and the implementation allow it, carry a non-sensitive source reference into the enquiry record. Preserve unknown attribution as unknown. A missing referrer should not be rewritten as an AI conversion merely because the business also received crawler visits.

Report enquiries, reviewed qualified enquiries, confirmed bookings and cancellations alongside traffic. At low volumes, show counts and individual patterns rather than drawing strong conclusions from a small percentage change.

1. Baseline: record the current indexed pages, search queries, visits, enquiry outcomes and the initial answer sample.
2. Choose: select one or two pages with relevant demand and a clear content gap. Assign an owner and state the expected reader benefit.
3. Improve: update the explanation, evidence, examples and internal links. Record the publication date and what changed.
4. Verify: check the rendered page, canonical URL, indexing eligibility, structured data and enquiry path.
5. Review: compare equivalent reporting windows and repeat the defined question sample. Account for seasonality, ranking volatility and other changes.
6. Decide: continue the topic, correct the page or choose the next gap based on the evidence.

Marsen Search connects search context, AI visibility and research into priorities a team can act on. Bring your existing Search Console observations and one customer question to a walkthrough. The starting point is a measurable decision, not a promised ranking or citation.