How a Port Klang logistics firm shows up in AI search: the DNE case study

A real, AEO-built Malaysian SME site is showing up in AI search and the record now runs long enough to show a trend, not a snapshot. DNE Logistics, a Port Klang freight forwarder and Acclaira's own live reference, drew 5,078 visitors in the 30 days to 27 August 2026 with zero paid ad spend. Its WhatsApp enquiry button fired 176 times from 154 people. Asked realistic buyer questions with no mention of the company, four of five AI assistants named DNE in their answers on the day we captured them.

Editorial correction: 14 September 2026. Measurement snapshot: 28 August 2026. Figures re-pulled from DNE's Google Analytics 4 on that date.

Key takeaways

• In the 30 days to 27 August 2026 the site drew 5,078 visitors and 6,763 sessions, and its WhatsApp enquiry action fired 176 times from 154 people — roughly five enquiry taps a day. There was no paid channel in the property at all.
• Against the previous 30 days (3,429 visitors, 114 people tapping WhatsApp) that is +48% visitors and +35% people tapping WhatsApp, in one month.
• Monthly visitors have gone 273 in May → 1,890 in June → 3,814 in July → 4,499 in August with four days still to run. May is the first complete month of measurement, so that is a 14-fold rise across the first three full months, not a single good week.
• Asked buyer questions about Port Klang customs, haulage and sea freight without naming the company, the assistants named DNE on the day of capture: Perplexity 8 answers out of 8 (citing dnelogistics.com.my), Copilot 5 of 5, ChatGPT 6 of 8, Claude 3 of 4 (27 June 2026), and Gemini 9 of 30 on a separate 30-question run (23 July 2026).
• We publish no figure for how much of this traffic came from AI, because no such figure exists. An earlier version of this case study reasoned that about 90% of visits were AI-driven; we withdrew it. We have since stopped quoting the analytics "AI Assistant" channel as well — see below for why a number that looks measured is not one.
• The point stands, and is now better evidenced. With Malaysian internet penetration at 98.0 percent (DataReportal, 2026), clear, accessible business content is useful. The dated answer captures and website metrics document different observations; they do not establish that AI caused the enquiries.

Whose site is this, and why use it as proof?

This is DNE Forwarding (M) Sdn Bhd, a family-run Port Klang freight forwarder operating since 1999. It runs JKDM-licensed customs clearance, KA-licensed container haulage, and sea and air freight forwarding for manufacturers across Selangor.

Disclosure, because it matters: this is the founder's own logistics company. Dan Duar is a director of DNE. It is not a paying Acclaira client and it is not an arm's-length reference.

We use it as the headline proof anyway, for one reason: it is the site whose analytics we can actually open. Most agencies show you polished case studies you cannot check. Here we can show the raw numbers, good and imperfect, because we built it and we own it. Treat it as first-party proof we can measure end to end, and weigh it accordingly. The other businesses we build for are listed on our work, each one a live site you can open and check for yourself.

What does the site actually do?

In the 30 days to 27 August 2026 the DNE site drew 5,078 visitors across 6,763 sessions and 7,858 pageviews, with an engagement rate of 38.8 percent. The job that matters most — a customer tapping the WhatsApp enquiry button — happened 176 times from 154 different people.

There was no paid ad spend. That is not a claim we are asking you to take on trust: the property's channel report contains no Paid Search, Paid Social or Cross-network channel at all for the period. Every visit came from search, direct, an AI assistant, or a referral.

These are modest numbers by consumer-brand standards, and that is the point. This is a niche business-to-business logistics firm in Port Klang, not a viral shop. For an SME, a site that quietly produces four or five tracked WhatsApp taps a day, every day, is a working sales channel.

Is it growing, or was it always like this?

It is growing, and this is the part the earlier version of this case study could not show.

Measurement on this property began on 6 April 2026, so April is a partial month and there is no earlier baseline to compare against. We are not going to claim a multiple against a month we only half-measured. If a before-and-after window is what you want to see, our Carriera case study has one, on a client we do not own. From the first complete month onward, the record is:

That is a 14-fold rise in monthly visitors between May and July, the first three complete months of measurement, with August already ahead of July on 27 days of data. The shape matters as much as the size: this is a steady climb across four months, not one spike from one viral moment.

Can you prove AI is sending the visitors?

We cannot establish AI’s total contribution from this record. Identifiable referrals can show some visits, but they do not account for every AI-influenced journey.

Google Analytics does report an "AI Assistant" channel, and we used to quote it on this page. We have stopped, because the number cannot mean what it appears to mean. Assistants strip or omit the referrer inconsistently, so an unknown and variable share of AI-prompted visits lands in Direct instead. Whatever the channel reports is the residue of whichever assistants happened to pass a referrer that month, not a measure of AI's contribution. The breakdown by engine is worse still: it mostly measures how each engine handles referrers, not how often it recommends you. Presented as a floor it is no better, because a caveat attached to a misleading number still leaves the number in your head. If you want to look at the channel in your own account anyway, our guide to seeing ChatGPT traffic in Google Analytics shows where it sits and what it leaves out.

So here is what we can show instead, none of which depends on attribution.

The assistants do name DNE, and we have the captures. On 27 June 2026 we put realistic buyer questions to five assistants — importing through Port Klang, customs clearance and haulage, sea freight, demurrage — never mentioning DNE. Perplexity named DNE in 8 answers out of 8 and cited dnelogistics.com.my as a source; Copilot in 5 of 5, with DNE as the first row of its comparison table; ChatGPT in 6 of 8; Claude in 3 of 4. Gemini named it in none of four that day. A separate 30-question run on 23 July 2026 found Gemini naming DNE in 9 of 30, usually in the first two positions.

And it is not a standing fact, whatever anyone tells you. On 28 August 2026 we ran the same Google AI Overview question eight times from clean, signed-out Malaysian browser profiles. Five were blocked by rate limiting; of the three that returned, DNE was named in two, and appeared in Google's own written sentence in only one. Minutes apart, the same question put DNE first in the prose, sixth in the attached business list, and absent from the page entirely. An AI answer is one draw from a distribution, not a ranking that holds still. Any agency showing you a single screenshot as proof of a permanent position is showing you one draw.

The commercial result needs none of this. The site went from 273 monthly visitors in May to 3,814 in July with no paid channel in the property at all, and its WhatsApp enquiry button fired 176 times from 154 people in the 30 days to 27 August. Those numbers are unambiguous, they are first-party, and they do not rest on knowing which engine sent whom.

How honest is the attribution, really?

A correction we owe you. An earlier version of this page reasoned that about 90 percent of DNE's visits were AI-driven. On the current data that is not defensible: in the last 30 days Organic Search alone accounted for 4,416 of 6,763 sessions, and Direct for a further 1,997. We withdrew the 90 percent figure rather than quietly leave it up. It was an estimate presented alongside measurements, and estimates like that drift into being quoted as fact — which is exactly what happened to the analytics figure we have now dropped as well.

So we will not tell you "AI drove X leads" with false precision, and we will not replace one guess with another. Reporting the uncertainty is part of the proof. Numbers you can trust are worth more than numbers that flatter — and if you are evaluating an agency, an unsourceable percentage is exactly the thing our own red-flags guide tells you to walk away from.

Through rigorous evaluation, we demonstrate that GEO can boost visibility by up to 40% in generative engine responses. Aggarwal et al., the Princeton GEO study (ACM KDD 2024).

What did we do to get here?

We built DNE the way an answer engine needs to read a business. The work is plain, not magic, and it follows the same AI-search visibility programme Acclaira runs for any Malaysian SME.

1. Answer-first content in real text. Every service, location and credential is stated in clear, crawlable words, not locked inside images. JKDM-licensed clearance, KA-licensed haulage, sea and air freight, Port Klang, Selangor. The facts an assistant needs are right there to read.
2. Structured data and consistent facts. Schema describes the organisation, its services and its area served, and the same name and details appear consistently across the web, so an assistant can resolve who DNE is with confidence.
3. Question-shaped pages and guides. The site and its insights answer the exact questions a manufacturer would ask an AI, which is what gets a page pulled into an answer. Adding clear statements, statistics and citations is a measured lever, worth up to a 40 percent visibility lift in Princeton's GEO study.
4. Measurement from day one. Google Analytics with a defined AI-Search channel and a WhatsApp key event, so we can see what is actually working rather than guess. That is how we can show you these numbers at all.

Which fixes mattered most?

The most useful work was the least glamorous: making DNE's basic facts machine-readable and consistent. We will not rank these by exact contribution, because the analytics cannot cleanly separate one lever from another. But this is the concrete checklist we worked through for DNE, and it is where any Malaysian SME should start.

• Schema.org structured data describing the organisation, its services and its area served, so an assistant can parse who DNE is without guessing.
• Services stated as real, crawlable text, not images. Clearance, haulage, sea and air freight, and the locations served are written in words a crawler can read, not trapped in a picture.
• A complete, consistent Google Business Profile, with the same name, contact details and services that appear on the site, so trusted sources agree about DNE.
• An FAQ that answers the real questions buyers ask, phrased the way a manufacturer would actually ask an assistant, which is what gets a page pulled into an answer.
• A fast, crawlable build where the content is in the page itself, not loaded later by script, so AI crawlers and Google can both read everything.

What can your business take from this?

That an ordinary Malaysian SME, in an unglamorous industry, can be the named answer and turn it into steady WhatsApp enquiries, and prove it with its own analytics. You do not need a huge audience. You need to be the clearest, most trustworthy, most machine-readable option for your customer's question, and a way to measure the result. The five-minute version of that measurement, which you can run yourself today, is in how to check if ChatGPT recommends your business in Malaysia.

The same starting point is open to you: a free 15-minute strategy call where we ask ChatGPT, Gemini and Google the exact questions your customers ask, show you whether the answer names you or a competitor, and hand you the single highest-impact fix to earn your first mention. It is how we began with DNE, and it costs nothing. Book a call, or browse the rest of our Insights library for the how-to behind each step.

When your customer asks AI who to trust, you want one name to come back. Make it yours. Be the answer.

Sources

• DNE Logistics, its own Google Analytics 4 account (first-party data), 30 days to 27 August 2026 and monthly totals from 6 April 2026, Asia/Kuala_Lumpur time zone. Re-pulled 28 August 2026. Held privately by DNE, not a public URL.
• Pew Research Center, 2025, Google users are less likely to click links when an AI summary appears
• Aggarwal et al., GEO: Generative Engine Optimization, Princeton (ACM KDD 2024)
• Google, May 2025, AI Overviews expansion to 200+ countries and 40+ languages
• DataReportal, Digital 2026 Malaysia