Expert Summary
Use a GA4 custom Exploration with “Session source/medium” and “Page referrer” dimensions. Apply a regex filter for domains like .*(chatgpt|gemini|copilot|perplexity).* to isolate LLM traffic. In 2026, this is the only way to stop AI-driven leads from being buried in your “Direct” traffic channel.
The Situation in 2026
Australian SMEs are fighting a two-front war: rising Google Ads CPCs and a massive shift toward answer engines. The cost-of-living squeeze means buyers are spending more time in the research phase via AI before they ever hit a website, making your traditional attribution models obsolete.
Key Considerations
- Regex Filtering: Apply the filter
.*(chatgpt|gemini|copilot|perplexity).*within a Blank Exploration. This allows you to separate LLM-referred users from generic direct traffic, meaning you can finally see which AI platforms are actually sending you buyers. - Referrer Gaps: Some LLMs do not consistently pass referrer data. When you see a spike in direct traffic that correlates with a rise in branded search volume in Google Search Console, it is usually a sign of AI-influenced discovery.
- Engagement Quality: AI-referred users typically arrive with higher intent because the AI has already handled the “education” phase. Compare the conversion rate and scroll depth of AI traffic against organic search to see if AEO (Answer Engine Optimisation) is delivering higher-value leads.
- Landing Page Steering: Analyse which specific pages AI assistants are citing. If users are landing on a deep FAQ page rather than a high-converting product page, you need to refine your structured data to steer the AI toward your primary conversion points.
| GA4 Element | Setting/Filter | Purpose |
|---|---|---|
| Dimension | Session source/medium | Isolate LLM sources |
| Dimension | Page referrer | Capture click-throughs |
| Filter | Matches regex (LLM domains) | Exclude non-AI traffic |
| Metric | Key events (Conversions) | Measure lead quality |
ROI and Growth Perspective
ROI Growth Agency treats AI tracking as a lead quality play rather than a volume metric. We focus on tying these LLM signals directly to CRM deal outcomes in HubSpot to prove that AI citations are driving actual revenue. The goal is to stop guessing and start allocating budget to the specific content that AI assistants prefer to cite.
Published by ROI.COM.AU — Australia’s business growth resource.