November 20, 2026Study: Visitors from ChatGPT and Other AI Assistants Convert More Often (Data from 6 Indonesian Business Sites)
Across six Indonesian business websites we work on, visitors who arrived from ChatGPT and other AI assistants contacted the business far more often than the average visitor. Over the 90 days to 18 November 2026, the conversion rate of AI-referred sessions ranged from 15.7% to 72.5%, while the sites' overall averages ranged from 5.1% to 28.9%. The direction was the same on all six sites: 2.5 to 5.7 times higher. The session counts are small, though (16 to 324 AI sessions per site), so treat this as an early signal, not a general rule.
Here, a conversion means a WhatsApp click or a form submission recorded as a key event in Google Analytics 4 (GA4). That is the first step towards a lead, not necessarily a sale. Below you will find the full data, how we calculated it, its limits, and how to check the same pattern on your own site.
Key findings
- 6 out of 6 sites show a higher conversion rate from AI than the site average.
- The gap is 2.5x to 5.7x. Smallest at Pet Travel Asia (72.5% vs 28.9%), largest at Angelic Agent (29.6% vs 5.2%).
- Volume is still small. Across all six sites: 515 AI sessions out of 23,554 total sessions, about 2.2%.
- The site with the most AI sessions, Pramika Uniform (324 sessions), recorded 15.7% vs 5.1%, roughly three times higher.
- In two manual ChatGPT checks on 19 November 2026, Pramika Uniform and Angelic Agent were named first for questions close to their business.
Full data by site
All figures come from each site's own GA4 property, covering the 90 days to 18 November 2026. All six site owners agreed to be named.
| Site | Business | Total sessions | Site conversion | AI sessions | AI conversion |
|---|---|---|---|---|---|
| Pramika Uniform | B2B workwear manufacturer, Jakarta | 18,852 | 5.1% | 324 | 15.7% |
| Pet Travel Asia | International pet relocation | 2,038 | 28.9% | 80 | 72.5% |
| Angelic Agent | Purebred cat import service | 993 | 5.2% | 27 | 29.6% |
| Peptivia | Product catalogue with WhatsApp ordering | 555 | 7.4% | 30 | 40.0% |
| Hoamedia | Personal branding production | 447 | 10.7% | 38 | 28.9% |
| Pet Travel Indo (AU landing page) | Pet relocation to Australia | 669 | 9.3% | 16 | 43.8% |
At a glance
Each block is roughly 4 percentage points.
| Site | Site average | From AI |
|---|---|---|
| Pramika Uniform | █ 5.1% | ████ 15.7% |
| Pet Travel Asia | ███████ 28.9% | ██████████████████ 72.5% |
| Angelic Agent | █ 5.2% | ███████ 29.6% |
| Peptivia | ██ 7.4% | ██████████ 40.0% |
| Hoamedia | ███ 10.7% | ███████ 28.9% |
| Pet Travel Indo (AU landing page) | ██ 9.3% | ███████████ 43.8% |
How much of the traffic comes from AI?
Very little. Keep that in mind before reading too much into the conversion rates.
| Site | Share of sessions from AI | Estimated converting AI sessions | AI conversion vs average |
|---|---|---|---|
| Pramika Uniform | 1.7% | about 51 of 324 | 3.1x |
| Pet Travel Asia | 3.9% | about 58 of 80 | 2.5x |
| Angelic Agent | 2.7% | about 8 of 27 | 5.7x |
| Peptivia | 5.4% | about 12 of 30 | 5.4x |
| Hoamedia | 8.5% | about 11 of 38 | 2.7x |
| Pet Travel Indo (AU landing page) | 2.4% | about 7 of 16 | 4.7x |
We worked out the converting session counts from the GA4 percentages and rounded them, so read them as estimates. At Pet Travel Indo, a single session moves the rate by about 6 percentage points.
Methodology
We want this section to be clear enough for others to judge, repeat, or challenge the findings.
- Data source: each site's own Google Analytics 4. Not a survey and not third-party data.
- Period: 90 days to 18 November 2026.
- AI sessions: sessions whose session source is chatgpt, gemini, perplexity, claude, or copilot. The five sources are combined, not split by tool.
- Conversion: a session that triggered at least one key event, meaning a WhatsApp click or a form submission. See what a conversion is and our SEO and ads glossary.
- Baseline: the conversion rate of all sessions on the same site, across all channels.
What to keep in mind before quoting these numbers:
- Small samples. Only 16 to 324 AI sessions per site. On small sites, one or two sessions can shift the percentage noticeably.
- Not a random sample. All six are client sites we built or manage, not a random pick from the web. Results may differ for other sites.
- Key events differ by site. Some count WhatsApp clicks, some count quote forms, some count both. That is why we do not merge all six into one average conversion rate. We always compare AI sessions with the average on the same site.
- The site average includes AI sessions, and on some sites paid ad traffic as well. Since AI makes up only 1.7% to 8.5% of sessions, the effect on the average is small.
- Google AI Overviews and AI Mode are not included. According to GA4's default channel definitions, visits from AI Overviews and AI Mode count as Organic Search, not as AI assistants.
- Some visits from AI apps arrive without a source and may land in Direct. So the AI session counts here are probably lower than reality.
If you treat every session as independent, the rough 95% uncertainty range for AI conversion on each site still sits above that site's average. Angelic Agent, for example: about 16% to 48%, against an average of 5.2%. The ranges are wide, so the exact figures are not reliable yet, but the direction is consistent.
Two direct checks in ChatGPT
On 19 November 2026 we typed two questions into ChatGPT (temporary chat, Instant mode). Both queries were in Indonesian:
- "rekomendasi konveksi seragam kerja Jakarta 100 pcs" (workwear manufacturer recommendation in Jakarta, 100 pieces): Pramika Uniform was named first.
- "beli kucing British Shorthair import terpercaya" (buying an imported British Shorthair from a reliable seller): Angelic Agent was named first.
That is two questions, from one account, at one point in time. AI answers change with wording, conversation history, location, and model updates. This is not proof of a fixed ranking, only an example that these sites do appear as answers to questions that sit close to a buying decision.
Why might AI visitors get in touch more often?
GA4 data cannot tell us why. What we can do is offer a reasonable explanation and list the other explanations that are also possible.
Main hypothesis: their intent is further along. Someone asking ChatGPT for a workwear maker in Jakarta for 100 pieces has already named the product, the location, and the quantity. The early research happened inside the conversation. By the time they click through, they often just want to confirm and then reach out. This fits the businesses in our data: all of them sell services or products that people usually ask about over WhatsApp before buying.
Google describes something similar for AI features in its own search. In AI features and your website, Google says clicks from results pages with AI Overviews tend to be higher quality, with visitors more likely to spend more time on the site. That is Google's observation about AI Overviews, not ChatGPT, so we mention it only as a reference point.
Other explanations we cannot rule out yet:
- People who already know the brand. Some visitors may ask AI about a brand they already know, then click to find contact details. They would have reached out with or without AI.
- The type of user. People who use AI assistants to research purchases may simply decide faster.
- Small-sample chance. On a site with 16 or 27 AI sessions, a handful of sessions can create a pattern.
What makes us fairly confident something is going on is the consistency: six sites, six different lines of business, all pointing the same way.
What this means for your business
Do not chase volume yet. At around 2% of sessions, AI traffic will not replace Google search or ads any time soon. Its value is in quality: few visitors, but more of them get in touch.
Make sure the pages AI might cite are ready for people who are ready to buy. AI visitors often land directly on a product or service page, not the home page. That page needs to answer what they check last: what exactly is on offer, who it is for, proof that the business is real, and a WhatsApp button or form that is easy to find.
The foundation is still SEO. Google states there are no additional requirements to appear in AI Overviews or AI Mode: the page must be indexed and meet Search's technical requirements. No special files or special schema markup are needed. Google also notes that meeting every requirement does not guarantee a page will be shown. For AI assistants outside Google, we do not yet have enough data to say what decides inclusion. We cover the difference in SEO vs GEO and our GEO guide for businesses.
Measure before you decide. If your site does not record WhatsApp clicks as a key event, you will not see this pattern at all.
What we do not know yet
- Whether AI leads turn into sales more often. This data stops at WhatsApp clicks and forms. We have not linked it to transactions.
- Whether the pattern holds with larger samples. We need more sites and a longer period.
- Which AI tool matters most. ChatGPT, Gemini, Perplexity, Claude, and Copilot are combined here.
- The effect of Google AI Overviews. Those visits are mixed into Organic Search, so this data cannot separate them.
- How many AI visits end up in Direct because their source was not passed on.
If we repeat this measurement in a later period, we will add the results to this page.
How to check in your own GA4
- Step 1: Make sure key events are set up. WhatsApp clicks, phone clicks, and form submissions should be marked as key events (in Admin, under Key events). Without them, conversion figures stay empty.
- Step 2: Open Reports > Acquisition > Traffic acquisition. Switch the primary dimension to Session source.
- Step 3: Filter for AI sources. Add a Session source filter that matches the regex chatgpt|gemini|perplexity|claude|copilot. GA4 also has a default channel called AI Assistant (see GA4 channel definitions), which you can use as a cross-check.
- Step 4: Show Sessions and Session key event rate. Compare the AI rows with the total.
- Step 5: Use at least 90 days and note the session count. If you have fewer than 30 AI sessions, treat the result as a note, not a decision.
If you find a different pattern on your site, we would like to hear about it. Contradicting data matters as much as supporting data.
Frequently asked questions
How do I see ChatGPT traffic in Google Analytics?
Open the Traffic acquisition report, switch the dimension to Session source, and look for chatgpt.com or any source containing chatgpt. The full steps are in the section above.
Does AI traffic always convert better?
Not necessarily. The pattern was consistent across our six sites, but the samples are small and all of them are sites we work on. Check your own data before making decisions.
Do I need llms.txt or special schema to appear in AI answers?
For AI Overviews and AI Mode, Google says no special files or markup are required. What you need is the SEO basics: pages that can be crawled, are indexed, and say clearly what they offer in text.
Are Google AI Overviews counted in this data?
No. GA4 counts visits from AI Overviews and AI Mode as Organic Search, so they are not part of the AI session figures in this study.
Citing this study
You are welcome to quote the figures in this article in your writing, presentations, or research. We ask for two things: link to this page as the source, and mention the limits (six sites, 90 days to 18 November 2026, small samples, not a random sample). Example: "According to GA4 data from six Indonesian business websites compiled by Bienara (90 days to 18 November 2026), sessions from ChatGPT and other AI assistants converted at 2.5 to 5.7 times the site average."
Per-project results, including Google Search Console impressions and clicks, are on our case studies page. If you want to know whether your site is ready to be referenced by search engines and AI assistants, see our SEO & GEO service and GEO service, or send us your site address on WhatsApp. We will first check whether your tracking can already pick up a pattern like this.
All posts