How to measure the traffic you receive from ChatGPT

Don't lose focus with metrics that don't contribute, focus on the metrics that cite you and really drive traffic to business areas.

The practical order (from most reliable to least) is usually:

1. UTM utm_source=chatgpt.com when it exists. OpenAI indicates that ChatGPT adds it to reference URLs from ChatGPT Search results, making attribution easier.
2. Referrer/source with OpenAI domains (for example, chatgpt.com, historical variants or openai patterns).
3. Landing page inferences + temporary patterns, and partial validation with logs.

How ChatGPT appears in your analytics (domains, UTMs and referrer)

In analytics, three “signals” usually appear to identify sessions from ChatGPT:

• Reference domain (source/referrer): the user clicks and your analytics receive a source like chatgpt.com (or another variant).
• UTMs: the link comes with campaign parameters (the ideal is utm_source=chatgpt.com).
• HTTP Referer header (referrer): the browser/app sends the origin of the navigation; GA4 translates it to source/medium dimensions.

Important: You won't always see UTMs. It depends on how the link was generated/served, the environment (app, browser), and whether the referrer is maintained or “stripped” for privacy or because of the sharing flow.

utm_source=chatgpt.com: when will you see it and why is it key

When traffic comes with utm_source=chatgpt.com, you have the cleanest markup for attribution because:

• the session source is explicitly identified;
• you can filter and group with less noise;
• it's easier to compare periods and clusters without depending on how GA4 classifies “referral”.

OpenAI explains that, in referrals from ChatGPT Search results, ChatGPT It automatically includes utm_source=chatgpt.com in the reference URLs, precisely to facilitate tracking in tools such as Google Analytics.

Operational nuance: if you arrive only with utm_source, the “medium” may vary (for example, appear as (not set) if there is no utm_medium). The important thing is that the font is clear and you can use it as a rule.

chatgpt.com vs chat.openai.com vs “(direct)/(none)”

In practice you can see a mixture of:

• chatgpt.com (current website and access from apps, according to OpenAI).
• alternative domains/formats (for example, historical variants or sharing flows).
• sessions that fall like (direct)/(none) if the referrer is lost or if the user copy/paste the link in another context.

Recommendation for reporting: don't look for a single “perfect source”. Build a cluster that captures:

• UTM utm_source=chatgpt.com when it exists;
• sources containing chatgpt or openai;
• and check out “direct” in combination with landings that typically receive appointments.

Fast measurement in GA4 in 5 minutes (without advanced configuration)

If you just want to answer “How much do I get from ChatGPT?” today, without setting up channels or dashboards:

1. In GA4, enter Reports (Reports).
2. Go to Acquisition → Traffic acquisition (Acquisition → Traffic Acquisition).
3. Change the main dimension to Session source/medium.
4. Use the report search engine and filter by Chatgpt and Openai.
5. Review volume and quality metrics (engagement and conversions).

Base report: Traffic acquisition filtering by Session source/medium

Exact steps:

• Reports → Acquisition → Traffic acquisition
• In the table, change the dimension to Session source/medium
• Search:
  
  • chatgpt.com/...
  • ... you thought.../...
  • any combination where the source points you to ChatGPT/OpenAI

What to look at in that same report (without complicating yourself):

• Sessions: volume.
• Engaged sessions/Engagement rate: minimum quality (if it comes and bounces, you'll see it quickly).
• Conversions (or key events): if that traffic provides value.

Tip: Select a sufficient date range (for example, 28 or 90 days) to avoid conclusions due to noise.

What pages ChatGPT is sending you: Landing pages + filter by source

Here you convert “traffic” into “insight”: what URLs ChatGPT is citing or sharing.

Fast method (without scans):

1. In the same report of Traffic Acquisition, apply a filter/comparison where Session source contain chatgpt or openai.
2. Change (or add in another report) the dimension to Landing page + query string (Landing page + parameters).
3. Sort by sessions or engaged sessions and get your “Top landing pages from ChatGPT”.

What to do with that top (direct action):

• If the landing is informational and doesn't convert: create a MOFU “bridge” (internal links to comparisons, advanced guides, tools, or decision pages).
• If the landing already converts: reinforces updating, clarity and internal links (protects what works).
• If there are several landings on the same topic: possible cannibalization; consolidates on a canonical page and redistributes linked.

Recommended configuration: own channel “AI/LLM referrals” in GA4

For stable month-to-month reporting (especially if you report to clients or management), it's best to create a Custom Channel Group that groups attendee traffic (ChatGPT and others). This way you avoid it being diluted in “Referral”, “Unassigned” or mixed with rare sources.

In GA4, Google documents that this is configured in Admin → Data display → Channel groups.

Create a Custom Channel Group with regex rule (ChatGPT and company)

Route:

• Admin → Data display → Channel groups
• Create a new channel within your custom group (for example, “AI/LLM Referrals”).
• Recommended rule: based on Session source (or Session source/medium).

Simple example (idea, not dogma):

• Condition: Session source Regex matches
• Value: chatgpt|openai|chat\ .openai\ .com|chatgpt\ .com

Keep a controlled list: add new sources when you detect them, but avoid inserting an infinite regex if you don't need one.

Channel priority: Why your “AI channel” should go above Referral

The channels in GA4 are assigned by Order of rules: If your traffic complies with a previous rule (for example, “Referral”), it can stay there and not reach your “AI” channel. That's why:

• Put on your “AI” channel Above of “Referral” within the personalized group;
• reviews historical results and validates that it is not being classified doubly.

What metrics matter (and which are misleading) to evaluate traffic from ChatGPT

The typical mistake is to measure only volume. For ChatGPT it is usually more useful to evaluate:

• session quality (real engagement),
• propensity to convert (direct or assisted conversion),
• value per session (when you have revenue or lead scoring).

And always comparing against a benchmark: organic, general referral or direct.

Minimum KPI set: sessions, engagement, conversions and value per session

A minimum (operating) set for the “AI” channel:

• Sessions: upward or downward trend?
• Engaged sessions/Engagement rate: Do they consume content or do they bounce?
• Conversion rate (or key event): do they do what you expect?
• Value per session (if applicable): Does it compensate even if there is less volume?

How to decide with these KPIs:

• Lots of traffic, little engagement: the content is cited but does not satisfy; review intent and first paragraphs.
• Good engagement, low conversion: MOFU/BOFU bridge is missing (internal linking, next step modules, comparisons).
• Few sessions, high conversion: prioritize protecting those landings and expanding related clusters.

Intent analysis: informational vs comparative vs decision

Segment by landing type (even if it's manual at first):

• Information guides: They usually capture quotes, but they need a bridge.
• Comparisons: they usually capture intent of choice (better for MOFU).
• Decision/BOFU: less volume, more impact if qualified traffic arrives.

The goal is not to force everything to convert, but to understand What role does ChatGPT play in your funnel and where you should reinforce the route.

Validation with logs (when GA4 doesn't give you the whole film)

GA4 will give you a piece of the truth. When the referrer is lost, or the links arrive without UTM, some of the traffic may appear to be “direct”. There the server/CDN logs help validate patterns:

• presence of utm_source=chatgpt.com in the query string;
• refer when it exists;
• and distribution by input URLs.

Use it as a technical audit and as a support to explain “why it doesn't fit 100%”.

What to look for in Logs/CDN: utm_source=chatgpt.com and referrers

Typical filters:

• Querystring contains utm_source=chatgpt.com
• Referrer contains chatgpt.com (when the field exists in your logging)

OpenAI documents the use of utm_source=chatgpt.com for referrals from ChatGPT Search, so it's a reasonable pattern to look for in logs when you want technical evidence.

If you're using CDN (Cloudflare/Fastly) or WAF, check which fields you export: not all pipelines have referrers by default.

Inevitable limitations: links without UTM, referrer stripping and copy/paste

Even with logs there are limits:

• Links without UTM: you can't “invent” the origin.
• Referrer stripping: Some apps/browsers reduce or eliminate referrer.
• Copy/paste: The user shares the link and the attribution chain is broken.
• Privacy/environments: Certain contexts reduce signals by design.

Useful conclusion: measure how trending and per pages/cluster, not like perfect accounting.

Going from measuring to optimizing: how to use this data in your SEO roadmap (with Makeit Tool)

Measuring by measuring doesn't work. The real advantage comes when you turn tracking into a backlog:

• which clusters are being cited,
• which landings attract qualified sessions,
• where the user is lost (without a bridge),
• and what needs updating to maintain “citability”.

With a SEMrush-like suite like Makeit Tool, you can operationalize it: group by clusters, map intent, detect gaps and monitor SERP and competition changes without relying on intuition.

From “Top landing pages from ChatGPT” to backlog: what to improve first

Prioritize with simple criteria:

• High engagement + low conversion: build a bridge to MOFU/BOFU (internal linking, decision modules).
• Stable traffic from ChatGPT: reinforce canonical, update sections and add proofs of work (checklists/tables).
• New pages with potential: expand subtopics (fan-out) and make them self-explanatory.

If you have limited resources: First optimize the 10—20 URLs that already receive sessions from ChatGPT before creating new ones.

Optimization checklist for “citable blocks” on pages that already receive traffic

• Direct answer at the beginning of each section (2—3 sentences).
• Precise and consistent definitions (without unnecessary jargon).
• Tables/checklists when providing a decision or method.
• Date of update when the theme evolves.
• Clear internal links to MOFU/BOFU (without aggressive self-promotion).
• Verified claims: Avoid statistics or statements without a source/date.

Frequently asked questions about how to measure traffic from ChatGPT

Does ChatGPT automatically add UTMs to links?

OpenAI indicates that, in referrals from ChatGPT Search results, ChatGPT automatically adds utm_source=chatgpt.com to the URLs to facilitate analytical tracking. Even so, not all links or flows guarantee UTM: it can vary depending on how the link is delivered or shared, and some of the traffic will still be partially attributable.

Why do I see “chatgpt.com/referral” in GA4 sometimes and sometimes I don't?

Because it depends on two things: if the link comes with UTM (which can set the source) and if the browser/app preserves the referrer. In addition, the ranking by GA4 channel may vary and leave it at “Referral” or “Unassigned”. For consistency, create a custom “AI” channel with rules per Session source.

Can ChatGPT traffic appear as (direct)/(none)?

Yes. If the referrer is lost (by app, browser, privacy) or if the user copies/pastes the link, GA4 can record the session as direct. The practical way to approach it is to review landing pages that usually receive citations and validate patterns with logs (UTM when it exists and referrer when it is registered), without promising full attribution.

How do I identify what content ChatGPT is citing on my website?

In GA4, filter for sources that contain “chatgpt” or “openai” and review the report of Landing page. The URLs that appear as a landing page are your strongest candidates to be referenced or shared. From there, it prioritizes improving those pages with citable blocks, updating and internal links to decision content.

Can ChatGPT Search traffic be separated from links shared by users?

Not always. The UTM utm_source=chatgpt.com helps when it's present (especially in traffic from ChatGPT Search, according to OpenAI), but there's no perfect signal to distinguish all cases. The most realistic thing is to work by patterns: UTM/Referrer when they exist and, if not, inferences through landing pages and temporary windows.

What minimum dashboard should I have to report it monthly?

A minimum dashboard can have: (1) sessions and engaged sessions from the “AI” channel, (2) conversions and conversion rate, (3) top landing pages from “AI”, and (4) month-to-month trend. For stability, create a Custom Channel Group in GA4 and place “AI” above “Referral” to avoid dilution of attribution.

Can I know in GA4 which prompt generated the click from ChatGPT?

Not directly and reliably. GA4 can show you the source/medium, campaign (if there is a UTM) and landing page, but not the content of the conversation or the user's prompt. Operationally, the useful thing is to analyze which pages receive that traffic, what intention they cover and how they behave (engagement/conversion) to optimize the cluster and the decision bridges