How to measure visits and signups from AI search

A crawler can fetch your pricing page without anyone considering your product. Counting that request as a potential customer makes your marketing report look better while telling you less.

If you want to know whether AI sends you buyers, build two reports: one for automated requests and one for referred visits. Then follow those visits through signup and payment.

What is AI traffic?

AI traffic can mean visits from people clicking AI answers or requests made by AI systems. Those counts describe different things, so your report needs to say which one it measures.

Semrush defines AI referral traffic as website visits from links cited in AI-generated responses. Someone asks ChatGPT a question, clicks a cited link, and reaches your site. That's a referral visit.

An AI crawler fetching your page is different. The request tells you that software accessed your content. It doesn't tell you that someone read your offer, started a trial, or bought your product.

Even a real referral visit isn't automatically a buyer. A person might want a definition, a code example, or an answer to a support question. You still need to measure what happens after the click.

Use separate labels for the things you can observe:

• AI referral visits: visits attributed to links in AI answers.
• Crawler requests: automated fetches recorded by your server or content delivery network.
• Referred signups and customers: outcomes tied to measurable AI referral visits.

Don't put these in one total called “AI growth.” A request count and a visit count use different units. Adding them doesn't produce a useful measure of customer demand.

For your SaaS, the useful question is narrower than “How much AI traffic do we get?” Ask how many measurable visits arrive from AI answers, where they land, and whether they become customers.

Keep the crawler count too. It can help you understand how automated systems access your site. Just don't ask it to answer a sales question.

A crawler surge doesn't prove customer demand

Training crawlers accounted for 67.5% of AI-driven traffic in HUMAN's 2025 dataset. That figure describes the company's measured traffic, not the share of your SaaS visitors who came from AI answers.

The report also found that monthly AI-driven traffic grew 187% from January to December 2025. Those are substantial changes in requests generated by or on behalf of AI systems. They aren't evidence that your product received more qualified visits.

The dataset matters here. HUMAN says more than 95% of its measured AI-driven traffic came from retail and e-commerce, streaming and media, and travel and hospitality. Don't borrow that mix as a forecast for your software business.

For your own site, create a report from server or content delivery network logs. A content delivery network, often called a CDN, sits between visitors and your server and can record requests.

Start with the fields your tools expose:

• The bot or requester your tools identify.
• The page requested and the request time.
• The response status, such as a successful fetch or an error.
• Whether your server or network allowed the request.

Keep unknown requests labelled unknown. Don't treat a missing bot label as proof that a person made the request.

Use this report to investigate access and repeated fetches. If crawler requests jump, check which pages they requested and whether those requests succeeded.

Don't celebrate the jump as customer growth. Until your referral and signup reports show a change, you have evidence of more automated activity, not more buyers.

How do you track AI referral visits?

Group visits with known AI referrers in your website analytics, while keeping that group separate from crawler logs. The goal is to count measurable arrivals from AI answers, not every request associated with an AI company.

Semrush names ChatGPT, Perplexity, Claude, and Google's AI Overviews as places where people can click links in AI-generated responses. Use that definition to decide what belongs in your referral report. Don't assume your analytics tool can distinguish every one of those sources.

Begin with the referral information your analytics actually records. Inspect the source values before writing rules. Save the rules somewhere you can check when a new source appears or an existing label changes.

Build a report with a small set of fields:

• The recorded referring source.
• The page where the visit started.
• The number of visits during the reporting period.
• The signups associated with those visits.
• The paid customers associated with those visits, where you can measure them.

Keep the date range visible. Compare the same length of time when checking whether visits changed. Otherwise, a longer reporting window can look like growth.

Check your grouping against the underlying records. A tidy channel label isn't enough if its rules collect unrelated traffic or miss sources you intended to include.

If your analytics tool already has an AI channel, inspect what it counts before accepting the total. You need to know whether it groups referral visits, estimates something broader, or mixes unlike measures.

Name the finished report “Measurable AI referral visits.” That label is less exciting than “AI-powered growth,” but it tells you exactly what the number can support. It also leaves room for visits whose origin you can't identify.

How do you connect AI visits to signups and payment?

Track a signup event and connect it to the recorded visit source wherever your setup allows. An AI referral report becomes useful for sales decisions when it shows what happened after someone arrived.

The Semrush definition stops at a visit from a link in an AI answer. A visit alone doesn't establish that the visitor created an account or paid. Those are separate events you need to record.

For a small SaaS, start with a path you can explain without opening a diagram:

1. Record the AI referral visit and its landing page.
2. Record a completed signup, rather than just a click on the signup button.
3. Record payment if you can connect it to that customer.
4. Keep the source rule and reporting period consistent.

Be explicit about what gets credit. Are you counting signups completed during the referred visit? Or are you preserving an earlier source when someone returns later? Either approach needs a clear label. Don't silently switch between them when one produces a nicer number.

Test the path yourself. Check that a successful signup produces the event you expect. Check what happens if someone leaves your marketing site to register elsewhere. Treat missing links as measurement gaps, rather than assigning the customer to a source you can't prove.

Keep signups and paid customers separate. Someone trying your product has taken a useful step, but that step doesn't establish revenue.

A crawler request, a visit, a signup, and a payment are four different events. Your report should preserve those differences all the way through.

If payment tracking isn't ready, publish the visit and signup counts first. A limited report with honest labels beats a complete-looking report built on guesses.

Small samples can make signup rates look better than they are

A signup rate needs a visit count beside it before you use it to change your content plan. A strong percentage from a small sample tells you less than the same percentage repeated across a larger, steady flow of visits.

HUMAN's report illustrates why the underlying measure matters. Its 187% rise describes monthly AI-driven traffic across its measured customer base during 2025. It doesn't describe your referral visits, your signups, or your chance of winning a customer.

Apply the same care to your own numbers. Report the count behind every rate:

• AI referral visits during the period.
• Completed signups tied to those visits.
• Paid customers tied to those visits, if known.
• Signup rate, calculated as signups divided by visits.

Use the same counting rules each time. If one report counts visits and another counts individual people, their rates won't answer the same question. Write down which measure you're using.

Don't hide an empty result either. A page with referred visits and no measured signups deserves a place in the report. So does a signup you can't connect to a source.

For a solo founder, the decision is whether the evidence justifies spending scarce writing time. One promising result can earn a follow-up article or a closer look at the landing page. It doesn't need to decide your next month of publishing.

Choose your review period before looking for a winning result. Then watch whether the pattern repeats. Keep raw counts visible so a neat percentage can't hide how little evidence you have.

Can you measure every customer influenced by AI?

Your referral report can't identify an AI-origin visit when the recorded data doesn't reveal that origin. Label it as measurable AI referral traffic, not a complete account of AI's effect on your sales.

Semrush defines the channel around visits from links in AI-generated responses. That gives you a useful boundary. It doesn't make every earlier mention, recommendation, or interaction visible in your analytics.

Consider a possible customer journey: someone sees your product mentioned in an AI answer, then types your address into a browser later. Unless you have evidence connecting those actions, don't assign that visit to AI. The customer may have been influenced by AI, but your referral records don't establish it.

The reverse matters too. A recorded AI referral doesn't prove that the AI answer introduced the person to your product. They may already know you. The referral tells you how that measurable visit arrived, not everything that caused the purchase.

Keep separate places for different kinds of evidence:

• Recorded AI referrals belong in the analytics report.
• Customer comments about discovering you through AI belong in research notes.
• Visits with no known AI origin stay outside the measured AI referral bucket.

You can ask new customers how they found you. Treat their answers as another source of evidence, rather than quietly adding them to a count of recorded referral visits.

That distinction lets you say something useful without claiming too much: “We measured these visits from AI answers, and these visits led to these signups.”

You don't need to claim a complete view of every influence on a purchase. You need a report clear enough to support the next decision and honest enough to survive a closer look.

Use referred customers to choose your next content job

Choose your next content task by connecting measurable AI referrals to landing pages and customer outcomes. Crawler activity can guide a technical check, but it shouldn't choose your editorial calendar by itself.

The Semrush guide defines AI referral traffic around clicked links in AI answers. That makes the landing page a sensible place to start: it's the first page you can inspect after the person leaves the answer and reaches your site.

Review the page alongside the question it answers. Then compare its recorded visits with completed signups and payments, where available. Don't assume that a page earning attention also helps someone choose your product.

Use a short set of checks before commissioning another article:

1. Check the visit source. Confirm that the report contains the AI referrals you intended to measure.
2. Check the page. Does it answer the visitor's question clearly, with evidence you can support?
3. Check the next step. Can a suitable visitor understand what your product does and how to try it?
4. Check the outcome. Did measured visits produce signups or customers, and is there enough evidence to repeat the bet?

Treat each answer as a reason to investigate, not an automatic diagnosis. Visits without signups might justify checking the signup path before rewriting the article. Signups without payments might justify talking to those users before publishing more pages.

If only crawler requests increased, inspect the logs. Don't order another batch of content because a bot discovered the last batch.

For your next review, open two reports. Use the bot report to understand automated access. Use the referral report to inspect visits and customer outcomes. Pick one concrete task from what you can actually see, and keep the unknowns labelled.