A ChatGPT search is not the same thing as a visit to your website. For a firm in Johor Bahru that serves people on both sides of the Causeway, that difference matters. ChatGPT search volume describes activity on a platform. Your analytics describes what people do on your own site. Neither number alone tells you how many people will ask about your services.

An AI answer can name a firm, link to a helpful page or answer a question without sending anyone to a site. A person might visit later by another route, but that path can be hard to trace. You do not need to give up on Google or ignore AI. Instead, track each stage on its own, improve useful content and act on evidence you can see. This applies to local firms and those selling to Singapore buyers who compare suppliers before they get in touch.

What ChatGPT search volume really tells you

In an interview with Big Technology, OpenAI consumer product lead Adam Fry made this claim:

ChatGPT’s volume of search-like queries is now roughly one-fifth the volume of queries of Google.

That claim is about search-like queries, not visits to websites or numbers of customers. Fry described these as questions people might once have taken to a search engine. They include requests for current information. His claim does not mean that every ChatGPT prompt is a search. Nor does it show that the two firms count queries in the same way.

The interview does not set out the exact Google baseline, the places covered or the time period. It does not give an independently verified sample or a full method for comparing the two. Treat one-fifth as OpenAI’s own claim, not an audited market share or a forecast of local demand.

An owner might read the headline and think one in five possible customers now arrives through ChatGPT. The claim does not show that. A query can lead to an answer without a link, a click or a sale. It may also be about a topic that has nothing to do with the owner’s firm.

ChatGPT search volume offers context on how people find information. It cannot tell you your share of attention, your referral rate or the value of a lead. To learn about those, you need to look at other data, ideally from your own site.

The type of question matters too. Someone asking for a quick definition has a different goal from a buyer who needs a supplier next week. Even a precise count for the whole platform would mix many topics and intents. Start with your service area and your customers’ needs when you plan content.

Why the website traffic comparison looks different

Another dataset looks at visits received by sites, not questions asked on a platform. In its AI versus search traffic benchmark, Ahrefs studied a fixed panel of 118,671 sites using its Web Analytics product.

For October 2026, Google made up 24.74% of measured traffic across that panel. ChatGPT made up 0.36%. Divide those shares and the result is about 69 to 1 in Google’s favour within this sample. These are shares of the panel’s measured website traffic. They are not shares of all searches or click-through rates.

The two sets of figures can both be true. A platform can receive many search-like questions while sending far fewer measurable visits to a given set of sites. The sources cover different groups and use different definitions and baselines. Neither source tells you what share of ChatGPT queries leads to a click.

Ahrefs used a selected analytics panel. It is not a representative count of Malaysia, Singapore or the whole web. Results can differ by industry, type of site and customer base. The benchmark updates, so keep the October 2026 date with these figures when you share them.

For a Johor firm, the lesson is to check your own traffic sources and enquiries before you shift your budget. An outside benchmark can prompt better questions. It does not promise that your own ratio will be 69 to 1.

For example, a specialist firm with detailed technical guides may attract visitors who act differently from those on a restaurant menu page. Both firms should review their own landing pages, countries and contact actions. They should not copy the panel average into a sales forecast. Changes to measurement or the mix of sites can also affect comparisons over time.

Separate searches, mentions, citations and enquiries

Think of discovery as a set of possible stages, not a funnel that every buyer follows:

• Search-like query: someone asks an AI a question. Your firm may not appear at all.
• Brand mention: the answer names your firm. It can do so without a link.
• Citation: the answer names a source and may include a link to one of your pages. That does not guarantee a click.
• Website visit: a person reaches your site. Analytics may show the source when that data is available.
• Enquiry: someone sends a form, calls or makes another relevant contact. A visit is not always a qualified lead.

Illustrative example: a Singapore buyer asks for a JB renovation contractor. An answer names a contractor and cites its guide. The buyer might read the answer, open the source, compare firms or do nothing. A mention alone does not tell you which, if any, of those things happened.

A citation may help people see a source even when no one clicks. A mention can also appear without a citation. Likewise, a referral session does not prove which sentence persuaded a customer. Xwork’s guide to AI mentions and citations explores this distinction further.

Put the numbers in separate columns. Do not add queries, citations, visits and enquiries as if they were the same unit. A few relevant enquiries can be worth more to a firm than many appearances in answers.

When you review a citation, note the page it links to and check whether its details are right. A link to your homepage serves a different need from a link to a service guide. If an answer gives the wrong service area or an old offer, correct your public details and watch future answers. Do not assume the correction will show up at once.

Use GA4 and Search Console without overclaiming

Start in GA4 at Reports → Acquisition → Traffic acquisition. Pick a consistent date range and look at Session default channel group. Google’s channel definitions describe AI Assistant as traffic from sources such as ChatGPT and Gemini when GA4 classifies it that way. Check session source/medium too. Channel rules and report availability can change.

Google’s AI Overviews and AI Mode are not in the AI Assistant channel. Their non-ad Google Search traffic falls under Organic Search. Do not label all Organic Search visits as AI-driven. Do not assume AI Assistant captures every AI-related visit either.

Filter or compare Country for Malaysia and Singapore. Then use an exploration or a suitable report to review landing pages. Record sessions and relevant key events, such as a form sent successfully. Check that the event fires as intended and is marked as a key event. If you track a WhatsApp click, call it a click, not a confirmed chat or sale.

In Search Console, the generative AI performance report shows impressions for links in Google AI Overviews and AI Mode. It includes page and country views. An impression is not a visit. This report does not give click counts for those features. The regular search performance report shows broader search clicks, not a clean count of clicks from these AI features.

Direct or unknown traffic is not proof of AI influence. A rise in branded Google searches is also a possible lead to investigate, not proof that AI caused it.

Before you compare months, check for changes to consent settings, the analytics tag, campaign labels or your site structure. Each can affect what your reports show. Search Console and GA4 also use different definitions and ways of counting, so their numbers need not match. Compare like with like rather than forcing the reports to agree.

A practical monthly plan for JB and Singapore businesses

1. Establish a baseline. Export the last full month’s sessions by channel and source, country and landing page. Include key events and, if you can, separately recorded qualified enquiries. Compare similar periods and take care with small counts. A rise from one to three leads may look dramatic, but it is still just two more recorded leads.

2. Repeat a small prompt set. Pick five to ten realistic questions in English and, if relevant to your customers, Bahasa Malaysia or Mandarin. One example is “Which JB accounting firms help Singapore-owned SMEs?” Run the same prompts each month. Record the date, platform, settings or location, brand mentions, cited URLs and answer accuracy. Answers can vary by run, account and location. This small set offers clues; it does not measure market share.

Improve pages and review results

3. Improve pages people need. Explain your services and locations. Add prices or quote conditions where suitable, and keep contact details current. Answer real questions clearly. Keep pages crawlable, use helpful internal links and ensure your business details are accurate and trustworthy. For more ways to measure results, see Xwork’s AI search analytics guide and its discussion of ChatGPT brand visibility.

4. Review outcomes, not hype. Note whether enquiries were relevant and whether tracking changed. An optional, non-leading form question can ask customers how they found you. Their answers are useful context, not perfect attribution. Generative engine optimisation (GEO) can fit within your existing SEO and content work. It need not be a mandatory standalone upsell. For another practical view, see CODE/RAVEN’s SEO and AI search protocol.

Give one person the job of keeping the monthly spreadsheet. Ask another to check that site changes are accurate. Record the date and a short note for each change. At the next review, compare what you see with your baseline. Allow for seasonal changes and other campaigns too. A change in a metric alone does not prove which action caused it.

Frequently asked questions

Does ChatGPT search volume show how many customers I can reach?

No. ChatGPT search volume is activity across a platform, not a count of people looking for your service. It does not directly show your local market, whether your brand appeared or whether an enquiry followed.

Does an AI citation always send a visitor?

No. A citation can make a source available, but a person still has to choose the link and reach the site. Even then, tracking may not retain every source. Keep citation observations separate from recorded sessions.

Where should I look for ChatGPT visits in GA4?

Check Traffic acquisition for the AI Assistant session channel, then inspect source/medium. Make sure tracking works and review landing pages and key events. Recorded traffic is an incomplete view, not a full history of AI influence. Some AI-related visits may lack identifiable referral data, and channel rules can change. Neither fact tells you how many AI visits went unrecorded.

Can I identify AI Overview clicks in Search Console?

The generative AI performance report shows impressions for supported Google Search AI features, not clicks specific to those features. General Search performance clicks cover broader results. Do not present them as an AI Overview click total. If the AI report is not available, Google says too few impressions may be one reason.

Should I stop investing in ordinary SEO?

No. Keep improving search visibility, helpful content and conversion tracking while you monitor AI referrals and appearances. If you want a clearer measurement plan, Xwork can help review your existing SEO and analytics without treating ChatGPT search volume as a traffic forecast.