How to Track Traffic and Leads From AI Search

AI search measurement sits in an awkward middle ground. Some of it is straightforward: GA4 can show you a session that arrived from chatgpt.com or perplexity.ai as clearly as it shows a session from Google. Other parts are effectively invisible: there is no analytics report that tells you your business was named in a ChatGPT answer that generated no click, or that a phone enquiry started life as a Perplexity search three days earlier.

The practical answer is to stop looking for one report that explains everything and instead build four separate layers — referral traffic, on-site conversions, lead-source attribution and manual citation tracking — and read them together. Businesses working with us on GEO agency services typically run all four in parallel, because each one catches something the others miss.

It also helps to be precise about terminology from the outset. A citation is your business being named or linked inside an AI answer — you may never see this directly. A visit is a session in your analytics that can be traced to an AI platform. An enquiry is a form fill, call or chat that converts that visit into a lead. An assisted conversion is one where AI search played a role somewhere in the journey but was not the final touchpoint recorded by your analytics.

Key takeaways

1. GA4 can track direct referral clicks from chatgpt.com, perplexity.ai, gemini.google.com and Copilot, but not citations without a click.
2. Set up a dedicated AI-referral channel grouping or exploration segment so these sessions do not disappear into 'Unassigned' or 'Direct'.
3. Use unique UTM parameters on any links you control (bios, profiles, citations you can influence) to make attribution certain rather than inferred.
4. Add a 'how did you hear about us' field to forms and a source question to call scripts, because many AI-influenced enquiries arrive as direct traffic.
5. Track brand-search volume growth in Google Search Console as a proxy for AI-driven awareness that never produces a referral.
6. Build a manual prompt-tracking spreadsheet — no platform currently gives you a reliable citation report.
7. Monitor Google Search Console's AI Overview appearance data specifically, separate from standard organic reporting.
8. Report on AI referral traffic, conversion rate and qualified leads together, not traffic volume alone.
9. Expect and account for measurement gaps rather than treating them as tracking failures to be 'fixed'.
10. Review AI search performance monthly alongside the rest of your channel reporting, not as a separate annual exercise.

Which AI platforms can send website traffic?

Several AI platforms are capable of sending genuine referral traffic when they include clickable links in their answers. The extent varies by platform and by whether the user is on a free or paid tier.

ChatGPT

ChatGPT sends referral traffic when browsing or retrieval is used to ground an answer, appearing in analytics as sessions from chatgpt.com or, in some cases, oai.azure.com and related domains. See our detailed breakdown of how to get more traffic from ChatGPT.

Perplexity

Perplexity is built around cited, linked answers by default, which makes it one of the more reliably trackable AI referrers — sessions typically arrive from perplexity.ai.

Gemini

Gemini sends traffic from gemini.google.com and, within Google's own ecosystem, AI Mode results can appear as a distinct segment inside Search Console alongside standard organic listings.

Copilot

Microsoft Copilot, particularly inside Bing and Edge, can send referral sessions from bing.com and copilot.microsoft.com, often at lower volume but with reasonably high intent.

Google AI results

AI Overviews and AI Mode results sit inside Google's own search results, so any resulting clicks usually appear as standard "google / organic" traffic in GA4 — the only way to isolate them is through Search Console's search appearance filters.

How to find AI referral traffic in GA4

Start with the standard Traffic Acquisition report under Reports > Acquisition, and switch the primary dimension to "Session source" or "Session source / medium." Search for chatgpt.com, perplexity.ai, gemini.google.com, bing.com and copilot.microsoft.com.

By default, GA4 usually buckets these under the "Organic Search" or "Referral" channel group depending on how it classifies the domain, which makes them hard to spot at a glance. Build a custom channel group in Admin > Data display > Channel groups that explicitly defines an "AI Referral" channel matching these source domains, so it appears as its own row everywhere channel groupings are used.

For a more flexible view, use a free-form Exploration report with "Session source" as the row dimension and "Sessions," "Engaged sessions," "Key events" and "Total revenue" as metrics, filtered to the AI domains you care about. This gives you a single table you can revisit monthly without rebuilding it.

Which referral sources should you monitor?

chatgpt.com and chat.openai.com
perplexity.ai
gemini.google.com and bard.google.com (legacy)
copilot.microsoft.com and bing.com (Copilot-driven sessions)
claude.ai, where Claude's web search feature is enabled
you.com and other smaller AI search engines relevant to your market

How to build an AI traffic report

A useful AI traffic report goes beyond session counts. Structure it around the metrics that actually indicate commercial value:

Users and sessions from each AI referral domain, tracked monthly to show trend direction.
Landing pages receiving those sessions, so you can see which content is actually being cited and reinforce it.
Engagement rate and average engagement time, which indicate whether AI-referred visitors are genuinely interested or bouncing immediately.
Conversions (key events) attributed to AI referral sessions specifically, not blended into overall organic figures.
Revenue, where ecommerce or lead-value tracking is set up, attributed to the same segment.

Save this as a dashboard or Looker Studio report pulling directly from GA4 so it updates automatically rather than requiring a manual pull each month.

How to track leads from AI search

Traffic is only half the picture — the commercial question is whether AI search produces enquiries. Several methods work together here.

Form tracking

Tag enquiry form submissions as key events in GA4 and segment them by session source, so you can see how many enquiries in a given month arrived via an AI referral session.

Call tracking

Use dynamic number insertion so phone numbers displayed to AI-referred visitors are distinguishable in your call tracking platform, allowing calls to be attributed to the same source data as web sessions.

CRM source fields

Pass UTM and source data through to your CRM at the point of enquiry, and add a specific "AI search" option to your lead-source picklist so sales teams can log it when a prospect mentions ChatGPT, Perplexity or similar by name.

Customer surveys

A short "how did you first hear about us" question, either on the thank-you page or in a post-enquiry follow-up email, catches AI-influenced leads that never show up as an AI referral in analytics.

UTM links

Where you control a link that might be surfaced by an AI platform — a profile bio, a directory listing, a press mention you supplied — add UTM parameters (utm_source=chatgpt, utm_medium=ai-referral) so any resulting clicks are attributed with certainty rather than guesswork.

Why some AI-driven enquiries appear as direct traffic

A large share of AI-influenced enquiries never register as an AI referral at all. This happens for a few consistent reasons:

The user was named your business by an AI platform without a link, then opened a new tab and typed your URL or brand name directly.
The research happened on a mobile device and the enquiry was made later from a desktop, breaking the session chain.
AI apps sometimes strip referrer data entirely, particularly within native mobile apps, causing GA4 to classify the session as "(direct) / (none)."
The user asked a voice assistant or AI app that does not pass any referrer information at all.

This is why a rising trend in direct traffic and branded search, alongside AI referral growth, is a legitimate (if imperfect) signal of AI-driven demand — not something to dismiss as unattributable noise.

How to monitor brand mentions and citations

Because most AI citations produce no analytics trace, monitoring has to happen at the platform level rather than the analytics level. Run your own brand name and core service terms through ChatGPT, Perplexity, Gemini and Copilot periodically and record whether you are mentioned, linked, and how accurately.

Several third-party AI-visibility monitoring tools now track brand mentions across multiple models automatically, which is worth adopting once manual checking becomes too time-consuming to sustain consistently. Until then, a disciplined manual process — covered below — is a perfectly credible starting point.

How to track Google AI Overview visibility

Google Search Console is the only reliable first-party source here. Under Performance > Search results, use the "Search appearance" filter and select "AI Overview" (where available in your account) to isolate impressions, clicks and average position specifically for queries where an AI Overview appeared.

Compare this against your overall organic performance for the same queries. A high impression count with very low clicks for AI Overview appearances is normal and expected — it reflects visibility rather than clicks, and should be reported as such rather than judged against standard organic click-through benchmarks. For the underlying mechanics, see our guide to how to get in Google's AI Overviews.

How to track prompts manually

A simple spreadsheet, checked on a set schedule, is still the most reliable way to know whether you are being cited. Build a priority list of 15-30 prompts that reflect real customer questions, and log the same fields for each one every time you check.

Manual prompt-tracking log structure
Field	What to record
Priority prompt	The exact phrase a customer might type or ask
Platform	ChatGPT, Perplexity, Gemini, Copilot, Google AI Overview
Brand appearance	Whether your business is named at all, yes or no
Citation	Whether a link to your site is included alongside the mention
Competitor appearance	Which named competitors appear in the same answer
Result changes	Notable shifts since the last check — new source cited, dropped mention, changed wording

Run this check monthly at minimum, and re-run individual prompts after significant content changes to see whether they move the needle. Over several cycles this log becomes far more valuable than any single snapshot, because it shows direction of travel rather than a one-off result.

Which metrics matter most?

AI referral traffic — sessions from known AI platform domains, tracked as a trend rather than a single figure.
Conversion rate of AI-referred sessions compared with other channels, indicating lead quality rather than just volume.
Qualified leads specifically flagged as AI-sourced in your CRM, not just raw enquiry counts.
Assisted conversions, captured through survey responses and CRM notes where AI search played a role but was not the last click.
Brand-search growth in Search Console, as a proxy for citation-driven awareness.
Share of voice in your prompt-tracking log against named competitors, tracked over time.

Common AI search measurement problems

Treating "Direct" traffic growth as unrelated to AI search when it may well be the effect of unlinked citations.
Missing AI referral sessions because they are absorbed into a generic "Organic Search" or "Referral" channel by default.
Expecting a single dashboard to explain the whole picture, rather than combining analytics with manual monitoring.
Judging AI Overview performance against standard click-through benchmarks, when impression-only visibility is the more relevant signal.
Not passing UTM or source data through to the CRM, so marketing and sales end up with two disconnected pictures.
Checking prompts once and drawing conclusions, rather than tracking change over multiple cycles.

A simple monthly reporting framework

Week 1: Pull the GA4 AI-referral exploration report — users, sessions, landing pages, engagement and conversions.
Week 1: Check Search Console AI Overview appearance data and branded search volume trends.
Week 2: Run the manual prompt-tracking log across your priority prompt list on each platform.
Week 3: Reconcile CRM lead-source data and any survey responses flagging AI search as an influence.
Week 4: Compile the combined report — AI referral traffic, conversion rate, qualified leads, assisted conversions and share of voice — and compare against the previous month.

Frequently asked questions

Can GA4 track ChatGPT traffic automatically?+

GA4 can capture sessions from chatgpt.com as a source without any extra setup, but it will usually classify them under a generic channel group rather than a distinct AI Referral category. You need to build a custom channel group or exploration report to see them clearly rather than buried inside Organic Search or Referral.

Why does my AI referral traffic look so much smaller than I expect?+

Because a large proportion of AI-influenced visits arrive as direct or branded organic traffic rather than a tracked referral, particularly when a citation carries no link or when research and enquiry happen on different devices. AI referral traffic in GA4 is a floor, not the full picture.

Is there a tool that shows exactly when my business is cited in ChatGPT?+

Not reliably and not for free. Several paid AI-visibility monitoring platforms have emerged that sample prompts across models automatically, but none offer complete or guaranteed coverage. A manual prompt-tracking log remains a sound, low-cost complement or starting point.

Should I use UTM parameters on links inside AI answers?+

You cannot control links generated dynamically by an AI model when it cites your page, but you can add UTM parameters to any link you supply directly — profile bios, directory listings, press mentions — increasing the chance that any resulting AI-sourced click is attributed correctly.

How do I know if an AI Overview is actually sending me traffic?+

Use Google Search Console's search appearance filter for AI Overview results and compare impressions and clicks for those queries against your standard organic figures. Expect impressions to far outweigh clicks, since AI Overviews satisfy many queries without a click at all.

What is the single most useful lead-tracking change to make first?+

Add a specific 'AI search' option to your CRM's lead-source picklist and a short 'how did you first hear about us' question to your enquiry form. This single change usually surfaces more AI-influenced leads than any amount of additional analytics configuration.

How often should I check my priority prompt list?+

Monthly is a sensible baseline for most businesses, with additional checks after significant content changes or new page launches so you can judge whether the change actually shifted your visibility.

Conclusion

There is no single dashboard that fully explains AI search performance, and treating the absence of one as a tracking failure leads to bad decisions. Reliable measurement comes from combining GA4 referral and conversion reporting, disciplined UTM and CRM source tracking, and a manual prompt-monitoring log run on a fixed schedule. Together they cover the traffic you can see, the leads you can attribute, and the citations that never produce a click at all. For a deeper look at the content side of this, see our guide to how you do SEO for ChatGPT.

If you would like help building an AI search dashboard linked directly to enquiries and revenue, rather than a collection of disconnected reports, get in touch and we will set one up around your actual sales process.