Search is changing.
Measure before you change direction.

There are more ways to encounter a business.

A person may see a conventional search result, read an AI answer with supporting links, or find the information they need in a business listing. Some of that discovery happens before they visit your website.

Google explains that its generative search features retrieve relevant information through its established search systems. SEO foundations remain relevant to these experiences. An AI mention is therefore another outcome to understand, rather than proof that conventional website work has become obsolete. Google’s AI optimisation guidance.

A search can also be useful without producing a website click. If someone checks a venue’s address before travelling, the immediate goal is finding the location. By contrast, a complex service enquiry may require detailed reading and a conversation. Judge the journey against the task.

For the broader background, our article on how search has evolved covers the shift from desktop results to mobile decisions and AI answers.

Keep three signals separate.

VisibilityWas your information shown or referenced?

Search impressions and AI citations can indicate exposure. They do not establish that someone read, trusted or acted on the information.

VisitsDid someone reach your website?

Search clicks and referral sessions show different parts of arrival. They are measured differently, so totals from separate tools need not match.

OutcomesDid the visit lead to something useful?

Relevant enquiries, completed bookings or sales connect the journey to the business. Include quality: a high volume of unsuitable enquiries can still create work without value.

Investigate the change before reacting.

First, check that the measurement still works. A changed analytics setup, consent configuration or enquiry process can change reported figures without the same change in demand.

Then identify which pages, queries, devices and countries account for the movement. Compare suitable periods, including the same season where that is meaningful. Check recent website changes, indexing problems, outages and search demand. Google’s traffic-drop guidance sets out these different possibilities. Google’s diagnostic guide.

Review a small sample of relevant search results to add context. Record the query, date, location and device. A result containing an AI answer is useful evidence about that sample; it does not prove that AI caused your overall decline.

Keep the explanation provisional until the evidence supports it. That helps avoid rewriting successful pages to solve the wrong problem.

Use the reports available—and respect their limits.

Google Search Console

Alongside its search performance reporting, Google provides a Generative AI performance report for impressions in AI Overviews and AI Mode. It can help identify which pages appear and how that exposure changes. It is an impressions view, not a complete account of visits or sales; limited impressions can also affect report availability. Google’s report documentation.

Bing Webmaster Tools

Bing’s AI Performance report provides citation information for supported Microsoft Copilot and partner experiences. Use it to understand exposure within that scope, not as a measure of every AI service. Bing’s report documentation.

Your own enquiry and sales records

Connect website reporting to what your team sees. Separate enquiries that fit your offer from spam, duplicates and requests outside your scope. Ask how customers found you when appropriate, while recognising that recall may be incomplete.

A small set of repeatable AI questions can supplement these records. Save the exact question, service, date and cited pages. Answers can vary; a few tests are observations, not a universal ranking.

Give each page a decision: keep, improve or combine.

Durable guidance can remain useful. Review it when the information or customer need changes. Age alone does not make a page obsolete, and changing its date alone does not improve its value.

Start with the purpose of each important page. Does it help someone understand a service, compare options, check evidence or complete an action? If so, measure its contribution with that purpose in mind.

• Keep material that remains accurate and serves a clear audience need.
• Improve a useful page when an explanation, example or next step is missing.
• Combine overlapping pages when one clearer resource would better serve the reader. Plan URL changes and redirects before removing anything.

Original evidence can give people a reason to choose your resource: explain a real decision, document a method or publish an authorised case study. Avoid claiming that any format is impossible for AI to summarise or guarantees a citation.

For ChatGPT-specific access and discovery, see our guide to website visibility in ChatGPT. For practical page improvements, start with our small-business SEO health check.

Run a focused review, then decide.

Choose one business priority and the pages supporting it. Record a baseline for relevant exposure, visits and outcomes. Note gaps in the data instead of filling them with assumptions.

Make a limited set of changes you can explain: a clearer comparison, current service information or a simpler enquiry step. Keep a record of when they went live. Compare subsequent performance over a period appropriate to your traffic and sales cycle.

Also consider where customers already know you: professional relationships, relevant communities, repeat business or referrals. Strengthen channels that suit your audience rather than joining a platform solely because it is fashionable.

OPEXPER’s approach is to connect a website’s structure, content and customer journey. Better measurement helps direct that work: towards the places where clearer information and a more usable experience can make a meaningful difference.