Is AI search bringing you enquiries? What to measure

Keep four different measures

A business name appearing in an answer is a mention. A source link is a route the reader might follow. A recorded visit shows someone reached the site. A qualified enquiry tells you whether the person needs something you can supply.

Write down the page linked as well as the business name. An answer may name you while linking to another website, or describe an old price. Count accuracy separately so that an incorrect recommendation does not look like an unqualified success.

Look at the landing page and recorded source

In Google Analytics, session source and medium describe where a session came from and the type of traffic. Review those alongside landing pages and your agreed enquiry events. Keep the reporting period consistent when comparing results.

Treat attribution as incomplete. People can see a recommendation, search your name later and phone you. Someone may also refuse analytics consent. An optional enquiry-source question can help, provided you keep their answer separate from the analytics record.

Check how form submissions and calls are counted before using the figures. A button click can be useful evidence of intent, but it is not the same as a received enquiry. Remove your own test submissions from any manual lead count.

Source: Google Analytics: traffic-source dimensions

Know what Google reports together

Google says traffic from its AI features is included in the overall Web search reporting in Search Console. That means a rise in the total does not, by itself, tell you how much came from AI Overviews or AI Mode.

Its guidance also says there is no special AI schema or new machine-readable file required for those features. Keep important information readable, maintain crawl access and use structured data that matches the page. Those are useful checks; they do not promise inclusion.

I'd be wary of a report that presents all organic search growth as an AI result. Ask the person reporting it which figures are directly measured and which are an interpretation. Google's guidance describes Google's products, so avoid applying it to every other answer engine.

Source: Google Search Central: AI features and your website

Repeat a small set of ordinary customer questions

Choose questions that a customer could ask without already knowing your name. Include the service and any location or constraint that genuinely affects the choice. Save the wording so it does not become more flattering to your business each time you test.

Use fresh conversations and record the conditions you can see. Repeat on different dates. Save absences as carefully as appearances, and keep the full answer so another person can check how you counted it.

These observations are a sample. Answers vary, and a set of manual checks cannot establish a universal rank. They can still reveal a wrong price, missing service detail or outdated linked page that is worth fixing.

Use the evidence to choose the next change

If a relevant page receives visits but few suitable enquiries, check the offer, contact route and question the page answers. If an answer cites an old page, correct or redirect the content through the normal website process. If there is no movement, keep that in the report.

Start with a few commercially useful pages and note when they change. Review the observations alongside enquiries over an agreed period. Keep unrelated site changes in the same record so you do not credit everything to one piece of content.