How to Tell If AI Is Sending You Traffic

At some point you will want to know if any of this AI search effort is doing anything. Reasonable. The problem is that AI referral tracking is genuinely messy right now. You can see some of it, you cannot see a lot of it, and the parts you cannot see are easy to either ignore or overclaim. This guide is about measuring honestly: what shows up, what hides, and how to check without fooling yourself.

The short version#

When someone clicks a link inside an AI answer, your analytics can often record where they came from, and AI tools tend to leave recognizable referrer traces. So you can usually spot at least some traffic arriving from assistants and AI search, and it often lands in your reports as a small but distinct source.

But two big gaps exist. First, zero-click answers leave no trace at all: if the AI answered the question and the person never clicked, you got exposure you cannot see. Second, people who read an AI answer, get curious about your brand, and later type your name into a browser show up as "direct" traffic, disconnected from the AI that actually sent them. So your analytics undercount AI's influence, sometimes badly, and no tool can fully close that gap today.

The practical stance: track what you can, treat it as a floor not a full picture, and never make a citation count into the thing you optimize. Conversion is still the number that matters.

The mechanism: how a referrer works, and where it fails#

When a browser follows a link, it usually tells the destination which page it came from. That is the referrer. Analytics tools group visits by referrer, which is how you see traffic labeled as coming from a search engine, a social platform, or a specific site. AI tools that link out are, in this respect, just another referrer, and they tend to show up under recognizable names.

The failure points are simple once you see them.

Path 1: AI answer with a clicked link
  person clicks -> browser sends a referrer -> shows in analytics   (visible)

Path 2: AI answer, question fully satisfied
  person reads, never clicks -> no visit at all                     (invisible)

Path 3: AI answer, brand remembered
  person types your name later -> lands as "direct" traffic         (misattributed)

Only path one is clean. Path two is pure loss to your measurement, even though the exposure was real. Path three quietly inflates your "direct" bucket, so some of what looks like people who already knew you is actually AI working, uncredited. Any honest read of your analytics has to account for all three.

What you can actually check#

Look for AI sources in your referrers. In whatever analytics you use, check the referral or source report for the names of AI assistants and AI search tools. If they appear, that is confirmed clicked-through traffic. Treat a new or growing entry there as a real signal, however small.

Watch for unexplained direct-traffic bumps. A rise in direct traffic with no obvious cause, especially alongside signs you are being cited, may partly be AI-influenced brand searches. You cannot prove it cleanly, so hold it as a soft indicator, not a fact.

Ask the assistants directly. Search your brand and your key topics inside the assistants themselves and see whether they name or cite you. This is manual and imperfect, and results vary by person and by day, but it tells you something your analytics never will: whether you appear in zero-click answers.

Tie it to conversions, not just visits. If AI-sourced visits are converting, that is the number worth watching. A source that sends visits which never act is not worth optimizing regardless of how it is labeled.

A clearly hypothetical example#

Made up to show the shape, not a promise.

Suppose you publish a clear answer to a specific question. Over a couple of months you notice three things at once. A small new entry appears in your referral report from an AI tool, say a modest number of visits. Your direct traffic ticks up a bit with no campaign to explain it. And when you ask an assistant your key question yourself, it names your page.

A naive read says "the referral report shows only a small number, so AI is barely doing anything." A more honest read is that the referral number is the visible floor, the direct bump is probably some AI-driven brand searches you cannot cleanly attribute, and the zero-click exposure, real but invisible, is larger than any of it. You will never get a tidy total. What you can do is confirm the direction, then judge it by whether those visitors convert.

Where the money comes from#

Measurement does not earn anything. It just tells you where to point effort. And the thing worth pointing effort at is the same as always: conversion.

AI exposure (partly visible, partly not)
        |
Visits you can and cannot fully attribute
        |
Some reach a page you own
        |
Offer, opt-in, or affiliate link earns
        |
Revenue (this is the number that settles the argument)

Because AI attribution is incomplete, revenue and conversions are your honest scoreboard. If you obsess over a citation count you can only partly measure, you will optimize a vanity number. If you watch whether people who arrive take action, you are measuring the thing that actually pays, exactly as in does ranking in AI search actually make money and how SEO makes money.

What to do#

• Set up basic analytics if you have not, so referrers and conversions are captured at all.
• Check your referral report for AI tool names, and note them as a floor.
• Watch direct traffic for unexplained movement and hold it as a soft signal.
• Spot-check the assistants by asking your key questions and seeing if you are named.
• Judge by conversions, not by citation counts you cannot fully measure.

What people get wrong#

• Assuming analytics capture all AI influence. They do not. Zero-click exposure and misattributed brand searches are invisible or mislabeled.
• Overclaiming the other way. Because some influence is invisible, it is tempting to credit AI for everything. Resist that too. Use the visible floor and the soft signals, and stay honest about the uncertainty.
• Optimizing citation counts. A number you can only partly see and that does not directly pay is a bad target. Optimize conversion.
• Trusting a tool that promises perfect AI attribution. Nobody can fully close the zero-click and direct-traffic gaps today. A tool claiming otherwise is overselling.
• Ignoring it entirely because it is messy. Messy is not the same as useless. The visible floor plus manual checks still tell you the direction.

Related links#

• The pillar: SEO for AI: how to get cited by ChatGPT, Claude, and AI search.
• Why conversion is the real scoreboard: does ranking in AI search actually make money.
• The money mechanics: how SEO makes money.
• More in the AI Search guides.

If you want the conversion side working so your measurements have something to measure, the free First $100 Blueprint is a plain starting point, and the newsletter follows the tooling as attribution slowly improves.