OwnlyticsDo AI-referred visitors convert better? How to check
Maybe, but nobody can tell you for your site. Visits from AI answers are often small in number and sometimes high in intent, yet whether they buy more than search or social visitors is something you have to measure with revenue, not guess from visit counts.
Fig.AI sources next to everything else. Sample data, not a real site.
We will not quote a percentage here, because we cannot verify one for your kind of site, and a small shop's own data beats any average.
Why might AI visitors convert differently?
There is a plausible reason. Someone asks an assistant for a tool or a recommendation, reads an answer, and clicks the one that was named. That reader may be further along than a person who typed a broad search query. The same logic suggests the opposite risk too: an assistant may send people who wanted a quick fact, not a purchase.
Both stories are guesses until you look at your own numbers. Sample sizes also bite. AI channels are usually small, so one extra sale swings the rate a lot.
What should you compare?
Conversion rate is the obvious metric, but it hides things. A channel with few visitors and two sales looks great on percentage and tells you almost nothing. Use these in order:
- Revenue per visitor (revenue divided by visitors) for each channel. It folds in both conversion and order size. We cover it in revenue per visitor, the metric that beats traffic.
- Number of sales behind each figure, so you know whether a rate rests on 2 sales or 200.
- Time to buy. AI visitors may decide faster or slower. See how long buyers take to purchase.
- Refunds, if you have them, so you compare net revenue.
Example (made-up numbers for illustration)
Imagine a one-person SaaS over 90 days. These numbers are invented.
| Channel | Visitors | Sales | Revenue | Revenue per visitor |
|---|---|---|---|---|
| Organic Search | 9,000 | 72 | $2,088 | $0.23 |
| Organic Social | 4,000 | 12 | $348 | $0.09 |
| AI (all assistants) | 240 | 5 | $145 | $0.60 |
| Direct | 2,500 | 25 | $725 | $0.29 |
At first glance, AI looks 2.6 times better than search. But 5 sales is a thin base. If 3 of those buyers had skipped the purchase, revenue per visitor would drop to about $0.24, level with search. The honest reading: "AI visitors look promising and deserve a closer look as volume grows."
A cautious rule of thumb (ours, not a statistical standard): do not draw conclusions from a channel until it has at least a few dozen sales, and keep re-checking.
| Source | Visitors | Revenue |
|---|---|---|
| chatgpt.com | 1,411 | $196 |
| perplexity.ai | 620 | $98 |
| claude.ai | 388 | $49 |
Fig.AI assistants as their own channel in the source table. Sample data, not a real site.
How do you get the data?
You need two things joined together: where each visitor came from, and what they paid.
- Source data comes from referrers and UTM tags. See how to track ChatGPT referrals.
- Payment data comes from your provider, connected as described in the revenue attribution docs. Without a link between visitor and payment, you only have visit counts.
Ownlytics does this join for Stripe, Polar, Dodo Payments and Creem. It puts ChatGPT, Claude, Perplexity, Gemini, Copilot and others into an AI channel and shows revenue and lifetime value per assistant. How a sale is matched to its visitor, and what is marked as estimated, is spelled out in the attribution docs. The overview is on the AI traffic analytics page.
What can distort the comparison?
- Missing source data. Clicks from assistant apps or copied links can arrive with no referrer. Those buyers look like Direct buyers, which makes your AI channel look smaller than it is, and Direct look better.
- Attribution model. If you credit the first touch, AI may win more sales than if you credit the last non-direct visit. Pick one model and keep it. See first-touch vs last-touch attribution for solo founders.
- Google AI features. AI Overviews and AI Mode send visitors from google.com, so they usually sit in Organic Search and cannot be separated out.
- Small samples. Covered above, and the most common mistake.
- Page mix. If assistants link to your pricing page and search sends people to blog posts, you are comparing different intents.
When does this question not matter?
If AI visits are under 1% of your traffic and you have fewer than a handful of sales from them, the answer will not change what you do this month. Keep the channel visible, check quarterly, and spend your time on whichever channel already pays.
A high conversion rate on 30 visitors is a lead to follow up, not a result to bank.
At a glance
| Metric | What it tells you | Watch out for |
|---|---|---|
| Visitors | Reach | Says nothing about buying |
| Conversion rate | Share who buy | Tiny samples swing wildly |
| Revenue per visitor | Value of a visit | Needs sales linked to visitors |
| Sales count | Confidence | Aim for dozens, not units |
| Refunds | Real revenue | Compare net, not gross |
If you want to look at a channel table with revenue per visitor on a sample site, the live demo has one.
FAQ
Do ChatGPT visitors buy more than Google visitors?
It depends on your site, and no one can answer it in general. Compare revenue per visitor for each channel over a long enough window, and check how many sales sit behind each number.
How many visits do I need before the numbers mean anything?
There is no magic threshold. Look at sales, not visits. A few dozen sales in a channel gives a rough signal. Under ten, treat any difference as noise.
Why does my AI channel show so few sales?
Partly because AI sends few clicks, and partly because some clicks arrive without a referrer and land in Direct. Both are normal. Some buyers also read a recommendation and visit later by typing your name.
Can I improve conversion from AI visitors?
You can make the pages assistants link to clear: a direct answer, a visible price, and an obvious next step. Do not expect a specific lift, and change one thing at a time so you can tell what worked.