AI Mentions, Website Citations and Visits: Three GEO Results to Measure Separately

“Our company already appears in AI answers, so why have website visits not noticeably increased?” This is a common question when owners start paying attention to GEO. A brand appearing in an answer, an official website link being included and a user actually visiting the website are three different actions. One figure cannot represent all three.

When assessing GEO, separate brand mentions, official website citations and identifiable website visits, then connect them with actual enquiries. This shows early changes in visibility without counting clicks or business that have not happened.

When exploring NEWPAGES GEO services, first identify the problem you want to address: customers cannot find the company, AI describes its business inaccurately, website information is not used as a reference, or visitors do not submit suitable requirements. Different goals call for different measures.

What does each of the three results tell you?

For example, an AI answer to “How should a factory in Johor choose an air-conditioning repair service?” may cite a company’s maintenance guide without recommending that company as a provider. The website supplied useful information, which is worth recording. Describing the citation as “AI recommends our company” would go beyond the evidence.

Alternatively, AI may mention the company but link to a business directory. The brand appeared, but the source was not the official website. Marketing should check whether the directory’s company name, URL and service area are accurate, and whether the website has more complete relevant information, rather than automatically counting a third-party link as an official website citation.

Use a fixed sample so the figures have a clear meaning

A small team can define its observation method around a fixed set of buyer questions. Consider this hypothetical example: twenty query observations are scheduled for the month, eighteen return complete answers and two encounter errors. Of the eighteen complete answers, six mention the brand and three cite the official website.

Using complete answers as the denominator, the brand mention rate is six divided by eighteen, or approximately 33%; the official website citation rate is three divided by eighteen, or approximately 17%. Also report twenty attempts, eighteen completed answers and two errors so readers understand sample completeness. If all attempts are used as the denominator instead, explain that choice and keep the calculation consistent thereafter.

An answer that writes the company name three times would normally count as one brand mention, not three successes. An answer citing two official website pages can count once as “contains an official website citation”, with the number of cited pages recorded separately. Agree on the rules first so different months or employees do not produce different counts for the same event.

These rates describe only the selected questions and test conditions. They are not the company’s share of the entire AI search market. Ten industrial equipment questions and ten catering equipment questions address different buyers, so directly comparing rates across industries is usually unhelpful.

If the website is cited but visits do not follow, what should you check?

First consider whether the answer already resolved a simple question. Someone who learns what to prepare before requesting a quotation may not click immediately; that does not mean the information served no purpose. Next check whether the cited page connects with a commercial need: does the article naturally guide readers towards suitable products, service areas and enquiry requirements?

For an office renovation company in Kuala Lumpur and Selangor, an article about considerations for a shop renovation reaches people in an early planning stage. Detailed scope and project explanations are closer to provider comparison. If everything stops at design inspiration, readers cannot judge whether the company handles their project. Connect articles to real service information rather than simply repeating brand slogans at the end.

Then examine the source, landing-page and conversion reports actually available in the website analytics account. Observe identifiable AI-source visits separately, but retain unidentified visits as unknown. Users may later search for the company name on Google or contact it directly through WhatsApp. Source labels alone generally cannot reconstruct every customer’s complete decision process.

AI SEO and website optimisation can be planned alongside this work, helping readers understand what the company does, the evidence supporting it and how to submit a specific requirement after finding the content.

Monthly reports should cover quantity, relevance and business progress

Place visibility records and sales information on one page, but list them separately. The first part should show the test scope, completed observations, brand mentions and official website citations. The second should cover relevant page visits and data limitations. The third should show actual enquiries, requirements meeting your service conditions and opportunities reaching quotation.

Look at the context of appearances as well as the count. If the company prioritises factory maintenance but additional mentions concern residential cleaning, it may not be reaching its target buyers. Conversely, even a small number of appearances in highly relevant questions such as choosing a provider for factory shutdown maintenance may deserve further observation.

Sales staff can naturally ask how a customer first heard about the company and record the response in the customer’s own words. “I found you through ChatGPT, then Googled your website” can be recorded as a customer-reported journey involving multiple channels. Do not force it into the final contact channel just to make the report tidy. Keeping the uncertainty helps budget decisions.

End each monthly review with one or two priority actions: correct an inaccurate business description, add service information to a cited page or improve the enquiry route. A GEO report becomes a business tool when the data leads to specific work.

Frequently asked questions

Does an official website citation guarantee a click?

No. A citation proves that the link appeared in a particular answer. A click requires supporting visit records. Distinguish clicks made by the testing team from visits by real potential customers, so internal checks are not counted as new customer acquisition.

Can the test mention rate be called market share?

No. A fixed question sample does not represent all market demand, and answers change with time and conditions. A more accurate description is “brand mentions observed in this round of twenty queries”, accompanied by the question scope, completion figures and calculation method.

If enquiries increase, should all the growth be attributed to GEO?

Check whether other channels changed at the same time. Promotions, advertising, exhibitions, existing-customer referrals and sales follow-up can all affect enquiries. Report concurrent changes and use identifiable sources and customer accounts as clues. Where evidence is insufficient, do not attribute all growth to a single service.

To identify the figures that matter to your business, bring your website and existing customer acquisition records and discuss GEO goals and measurement with NEWPAGES.