Since AirOps deployed in March, direct sessions have grown to become the #1 source on joblogic.com. Where AI assistants do carry their referrer, ChatGPT-referred traffic now converts at a better rate than google/cpc. This is the honest version of what we can — and can't — claim.
Cumulative Feb 1 – May 18 2026, filtered to the demo_form_submission key event. All major sources, ranked by sessions:
Direct is the largest — by a clear margin.
Direct is high-volume but low-conversion. Paid converts best (for now).
Sessions, demo submissions and conversion rate by source/medium. ChatGPT-referred highlighted.
| Source / Medium | Sessions | % of total | Demos | Conv rate |
|---|---|---|---|---|
| (direct) / (none) | 52,911 | 37.47% | 112 | 0.21% |
| google / cpc | 41,049 | 29.07% | 572 | 1.39% |
| google / organic | 16,701 | 11.83% | 48 | 0.29% |
| bing / cpc | 12,966 | 9.18% | 67 | 0.52% |
| bing / organic | 5,324 | 3.77% | 24 | 0.45% |
| chatgpt.com / (not set) | 76 | 0.05% | 2 | 2.63% |
| chatgpt.com / referral | 153 | 0.11% | 1 | 0.65% |
Weekly direct/(none) sessions. The acceleration sits inside the AirOps tracking window — that doesn't prove causation, but the trajectory is consistent.
When ChatGPT, Perplexity, Claude or Gemini pass through their source tag, GA4 captures it. The volume is small. The quality signal is unambiguous.
ChatGPT dominates. Perplexity is the named gap.
When AI traffic converts, it converts better than we can buy.
ChatGPT-referred traffic on the (not set) row converts at 2.63% — higher than google/cpc (1.39%) — with a bounce rate of 7.89% (vs site average 12.21%) and engagement of 1m 12s. Users arrive on Joblogic from ChatGPT with a clear intent to evaluate. The volume is currently small (76 sessions in 3.5 months) but on a unit-economics basis, AI-referred traffic is the highest-quality traffic on the site.
This is correlation, not causation. Three things weaken the headline if not acknowledged up front. Don't bury these — name them.
The 6-page paid investigation we ran this week attributed ~1,768 additional direct/(none) sessions per month to a Performance Max campaign that strips UTM parameters. The AEO claim must be net of that. Excluding the four affected feature pages gives a cleaner — but smaller — signal.
It captures bookmarked visits, iOS Mail referrer-strips, Slack/Teams shares, password-protected redirects AND AI assistant traffic that loses its referrer. We can't separate these cleanly without UTMs on every AI citation — which we don't control.
A truly clean test compares the AirOps period vs the same period in 2025. We have GA4 data starting later than that. 3.5 months is enough to see a pattern, not enough to rule out seasonal effects. YoY comparison is the next step.
Move the sliders to model ChatGPT referral growth and conversion rate. The pipeline impact updates in real time. Conservative assumptions: 10% demo→close rate, £7,500 blended AVC.
This is illustrative only. The conversion rate assumes ChatGPT-referred traffic continues to convert at its observed quality. Scale effects, sample size and seasonality not modelled.
The current evidence is consistent with AEO impact but not conclusive. Each of these moves the needle from "consistent with" to "demonstrated."
Pull direct/(none) sessions for Feb–May 2025 vs Feb–May 2026 once we have the historical data. A YoY trajectory removes seasonality from the picture and either confirms or kills the AEO inference cleanly.
Exclude direct sessions landing on /products/premium, /features/gas-certificate-app, /features/boiler-servicing-software, /features/route-scheduling. What's left is the AEO-attributable share. That's the number to track monthly.
AEO often shows up as increased branded queries on Google (people search "joblogic" after hearing about us in ChatGPT). The google/organic rate is 0.29% on 16,701 sessions — worth pulling the brand vs non-brand split separately.
GA4 has both demo_form_submit AND demo_form_submission firing. Separate from this analysis but worth flagging — once resolved, historical comparisons get cleaner.