Which AI Search Engines Bring You Leads and Sales?

Which AI search engines bring you leads and sales? To find out, look at what people do after they arrive.

I checked 283 AI visits to a content website. Of those, 38 included a click on a retailer link.

Perplexity sent 108 visits, compared with 65 from ChatGPT. But the number of visits with a retailer click was much closer: 14 from Perplexity and 12 from ChatGPT.

That makes both sources worth a closer look. Traffic alone doesn't tell you how much an AI source helps your business.

Below, I show the full example, what the public studies say, and how to run the same check on your website. I built Clickport to connect each AI source to the enquiries and sales it brings.

What did the website show?

The example covers June through August 2026. The website publishes content and links to retailers. Its name, page addresses, and products stay private.

A click to a retailer takes a reader from the website to a shop. That's the next step I can measure here.

Here is how the six AI sources compare:

Perplexity led on both measures. But ChatGPT and Claude sent less traffic while still bringing 12 and 10 visits with a retailer click.

I'd look at the pages people reached from those three sources first. Are they finding what they need? Is it easy to take the next step?

I wouldn't pick a winner from a sample this small. Your website can produce different results.

How I counted: I used Clickport's AI Search channel for visits that started from 1 June through 31 August 2026, using UTC. I then checked which visits included a retailer-link click during that period. A visit counts once, even if it includes several clicks.

A visit means one session, so the same person can visit more than once. The figures come from the website's analytics. They don't include orders placed with retailers.

Why check each AI engine separately?

The Minuttia and Convex Works study examines 24,067 prompts across five search and AI systems.

For questions about alternatives, the researchers checked which engines recommended a brand in response to the same question. Of those brand-and-question pairs, 53.7% appeared in only one engine. Just 8.05% appeared across all five.

In other words, the engine you check can change which brands you see. The calculation includes only questions answered by all five systems. The study methodology explains how the researchers counted the results.

The study looks at recommendations. The website example looks at visits and clicks. Neither tells you which engine brings the most sales to your business.

That's why I check each source separately and follow it through to the result that matters.

What do the public studies say?

My example is one website. Larger studies point the same way.

One result points the other way. In the Visibility Labs data, the average order from ChatGPT was $204, compared with $238 from organic search. Revenue per session was still higher for ChatGPT: $3.65 compared with $3.30.

The volume is small, and it grows fast. Previsible analyzed 19 GA4 properties and found that AI sessions rose 527% from January to May 2025. Shopify said on its earnings call of November 4, 2025 that orders attributed to AI searches were up 11x since January.

Few companies measure it. McKinsey found that only 16% of brands systematically track their performance in AI search.

These studies use different definitions, and two of them come from a single company's data. None of them tells you what AI traffic brings your business. For that, you need your own result.

Which result matters for your business?

Start with something you want people to do on your website. Be specific about what counts.

For a service business, that could be a completed enquiry. For a booking website, it could be a confirmed reservation. For an online shop, it could be a paid order.

In this example, I can measure clicks to retailers. I can't see their sales, so I don't count those clicks as purchases or earned commission.

Clickport calls these actions goals. You choose what to count, such as someone sending an enquiry.

Start with one goal and try it yourself. Send a test enquiry, for example, then check that it appears in your report. The goals guide shows you how to set this up.

For your own checkout, Clickport connects payments to AI sources and landing pages. Connect Stripe, Paddle, or the Revenue API. Follow the setup guide to link each payment to its recorded visit.

Here is what that looks like for a website that sells directly:

That €49 payment gives you a result to compare with other sources and pages. You can see where recorded sales start and decide which page to improve or promote.

Start your 30-day free Clickport trial and check which AI sources bring results on your website. All features are included, with no credit card required.

Use these reports once your goal and payment integration are ready:

1. Select AI Search under Sources → Channels. Sources → Sources lists the recognized AI engines.
2. Select a goal under Goals → All. Compare conversions and conversion rate in the Sources report.
3. Clear the goal filter and activate Revenue in the header. Compare payment revenue by AI source.
4. Select an AI source. Apply an Entry page filter to check Revenue for visits that started on that page.

Check the sales against your order records. Some payments won't have a known traffic source. And if you assign a value to an enquiry, remember that it's an estimate of what it might be worth.

Which page should you check first?

Look at the first page people reach from each AI source. Does it answer their question? Can they easily see what to do next?

In this example, 68 AI visits started on one page, and 12 of those visits included a retailer click. Another page brought 27 visits, including eight with a retailer click.

I'd look at both pages. The second received fewer visits, but people still clicked through to retailers during eight of them.

These totals include all six AI sources. Select one engine to see which pages its visits started on.

In Clickport, select the AI Search channel, then open Pages and choose Entry. This shows where those visits start. Use the Entry page filter to check results for visits that started on a particular page.

With that filter active, select your goal to check results for the landing page. Clear the goal filter and activate Revenue to check its attributed payments.

Read the page before you change it. You might find an unclear link, an unanswered question, or an offer that doesn't fit. The traffic count can't tell you that.

What if the report shows no leads or sales?

First, check that your tracking works. Try the action yourself and look for it in the report.

Make sure you're counting the right thing, too. Opening a contact page isn't the same as sending an enquiry. Clicking a payment button isn't the same as paying.

Then look at how much traffic you have. Kagi and Mistral sent just 11 visits between them in this example. None included a retailer click.

I wouldn't dismiss either source after 11 visits. Give the report more time before making a large change.

Some visits arrive without telling your website where they came from. Browser rules called referrer policies can limit that information. You can't reliably assign those visits to an AI engine. A "How did you hear about us?" question can show part of what they miss.

Bot filtering isn't perfect, either. A source name alone doesn't prove that a visit came from a person. These are the visits the website's analytics recorded.

I explain this in more detail in how to track AI search traffic.

What should your monthly report say?

Keep it short: where the visits came from, what happened, and what you'll do next. Include the dates so everyone knows which period you're discussing.

For this example, I'd write:

For your business, use the goal you've chosen, such as enquiries or paid orders. Only report sales when you can confirm them in your payment records.

Use the same dates for every source and count results the same way. If you change what a goal counts, mention that in the report.

A useful report helps you decide what to do next.

For practical ways to turn those figures into a client update, see our 10 AI report examples for agencies and website owners. If orders also complete inside an AI channel, our agentic commerce analytics example shows how to reconcile them with website-attributed revenue without counting the same sale twice.

Count the results, not the visits

In the example, Perplexity sent the most visits. ChatGPT and Claude came close on retailer clicks with far less traffic. A visit count alone hides that.

So start with one useful action, such as an enquiry or a paid order. Then compare your AI sources by that result. I would use it to choose the next page to improve.

I built Clickport to make that comparison short: the AI source, the landing page, the goal, and the payment in one report.

Start your 30-day free Clickport trial, add your website, and choose your first goal. All features are included, with no credit card required. The guide to switching from Google Analytics takes you through the setup.