We looked at 185,791 visits and 7,127 enquiries across 100 websites to answer one question before Black Friday: which channels actually produce enquiries, and which ones only produce traffic.
The two lists are close to opposite. The biggest source of visits was the weakest source of enquiries, and the source almost nobody measures converted better than everything else.
Here is what the data says, with the counts, so you can check the arithmetic yourself.
The headline numbers
- Paid search and social produced 44% of all enquiries from 16.7% of all visits. That is 3,137 of 7,127 enquiries, from 30,936 of 185,791 visits.
- Direct traffic was the biggest source and the weakest performer: 49.7% of visits, 27.3% of enquiries, converting at 2.1%.
- AI assistants converted 2.8 times better than organic search: 7.2% against 2.6%.
- Claude.ai converted at 11.1% and ChatGPT at 6.7%. Google organic search converted at 2.6% across 48,685 visits.
- Google Business Profile converted at 28.4%, the highest rate of any source in the study.
- A paid visitor was 4.8 times more likely to enquire than a direct one.
AI assistants send almost nothing, and it is the best traffic on the site
AI assistants sent 3,062 visits, which is 1.6% of the total. Those visits produced 220 enquiries, a 7.2% conversion rate.
Google organic search sent 48,685 visits over the same period and converted at 2.6%. Bing was much the same at 2.8%.
So a visit arriving from ChatGPT, Claude or Perplexity was 2.8 times more likely to end in an enquiry than a visit from a Google search result. Broken down, Claude.ai converted at 11.1% from 461 visits and ChatGPT at 6.7% from 2,488.
The likely reason is where in the decision the referral happens. Someone who asks an assistant which supplier to use and then clicks through has already had the comparison done for them. Someone scrolling a page of search results has not.
The volume is small enough to ignore, and the quality is high enough that ignoring it is expensive. On these numbers, 3,062 visits from AI assistants were worth about as many enquiries as 8,400 visits from Google.
Paid sends one visit in six and produces four enquiries in ten
Paid traffic was 30,936 of 185,791 visits, or 16.7%. It produced 3,137 of 7,127 enquiries, or 44.0%, converting at 10.1%.
Everything else, by share of visits and share of enquiries:
- Direct: 49.7% of visits, 27.3% of enquiries, 2.1% conversion
- Organic: 29.2% of visits, 21.7% of enquiries, 2.9% conversion
- Paid: 16.7% of visits, 44.0% of enquiries, 10.1% conversion
- Referral: 2.9% of visits, 4.7% of enquiries, 6.2% conversion
- Social: 1.4% of visits, 1.7% of enquiries, 4.7% conversion
- Email: 0.2% of visits, 0.5% of enquiries, 9.7% conversion
Google Ads converted at 10.0% and Facebook Ads at 10.3%, which is closer than most people expect. Instagram Ads converted at 15.9%, though from only 239 visits.
The practical point for Black Friday is that a traffic report and an enquiry report rank your channels in almost opposite orders. Budget decisions made on the traffic report will be wrong.
Half of all traffic is direct, and direct converts worst
Direct was the largest single source in the study at 92,295 visits, 49.7% of everything. It converted at 2.1%, the lowest rate of any category.
Direct is not really a source. It is the bucket a visit lands in when nothing identifies where it came from. A link in a messaging app, an email client that strips the referrer, a QR code on a flyer, someone typing your address after seeing an ad. All of it reads as direct.
So the biggest line in most analytics reports is the one that explains least, and its low conversion rate is partly an artefact. Real paid and referral visits are hiding inside it, diluted by people who were coming anyway.
That matters most in the week when ad spend triples. The more you spend on channels that lose their referrer, the more of your own performance disappears into a bucket labelled direct.
The channels too small to show up in a report
Two sources converted far above everything else, and both are small enough that most reports round them to nothing.
- Google Business Profile: 430 visits, 122 enquiries, 28.4% conversion
- Instagram Ads: 239 visits, 38 enquiries, 15.9% conversion
- LinkedIn: 319 visits, 36 enquiries, 11.3% conversion
- Email: 339 visits, 38 enquiries, 11.2% conversion
Google Business Profile converted better than any other source in the study, at more than ten times the rate of organic search, from 0.2% of the traffic.
Sorted by visits, these sit at the bottom of the page. Sorted by enquiries per visit, they are the entire top of it.
What the purchase data shows so far
Purchase tracking launched ten days before this study and recorded 80 purchases, so treat this as a signal rather than a finding.
Of those 80 purchases, 51 came from paid, 12 from direct, 10 from organic, 3 from referral, 3 from email and 1 from social. Paid produced 64% of purchases from 16.7% of visits, which runs in the same direction as the enquiry data, only harder. Google Ads and Facebook Ads produced 25 purchases each.
AI assistants produced none, from 3,062 visits. Hold that loosely. Ten days is not long enough to say AI referrals do not buy, only that none of them bought yet in this sample.
What to do before Black Friday
- Tag every paid link with campaign parameters, not just the platform's own click ID. An untagged click reports as direct or social, and the spend behind it becomes invisible.
- Check what share of your traffic reads as direct. Above 40% means a large part of your reporting is guesswork.
- Rank your channels by enquiries per visit, not by visits. The two orders are close to opposite.
- Claim and fill out your Google Business Profile. It converted at 28.4% here and takes an afternoon.
- Tag your email links. Most email clients strip the referrer, so untagged email traffic disappears into direct.
- Look at which pages AI assistants send people to, and make sure those pages answer the question that was asked.
- If your checkout sits on a different domain to your storefront, confirm the attribution survives the hop. It usually does not by default.
How we collected this
The data covers 100 websites tracked by Lead Recorder between 9 April and 12 November 2026: 185,791 visits and 7,127 enquiries. The sites span ecommerce and services, mostly Australian and New Zealand businesses.
A visit is a tracked session. An enquiry is a form submission, click to call or click to email recorded against that session. Source is classified from the referrer and any campaign parameters on the landing URL.
We excluded 148 enquiries that arrived through the API with no session behind them, since there is no visit to convert from. We also excluded our own infrastructure, site builder preview URLs, localhost and bare IP addresses from the referral counts.
Limitations
The sample is self-selected. These are businesses that chose to install attribution software, so they are likely better measured than average.
Direct is a residual category and will contain real paid and referral traffic whose origin was lost, which means paid is more likely understated than overstated.
AI assistant referrals total 3,062 visits and 220 enquiries. That is enough to see a pattern and not enough to break down by assistant with confidence, with the exception of ChatGPT.
Purchase data covers ten days and 80 purchases, and is reported as an early signal only.
We sell attribution software, so we have a commercial interest in these conclusions. The counts and denominators are published above so the arithmetic can be checked.