How to Track Brand Mentions in AI and Attribute Referral Traffic Accurately

Overview: Why AI Referral Tracking Now Matters

AI assistants are now a real acquisition channel.

Adobe found U.S. web traffic from AI-driven referrals grew 10x between July 2024 and February 2025, and Similarweb reported over 1.1B AI referral visits in June 2025, with about 7% conversion on transactional sites.

If you don’t track brand mentions in AI assistants and attribute referral traffic accurately, you’re already flying blind.

This step-by-step tutorial shows you how to:

• Detect AI-driven referrals in your analytics

• Tag sessions from ChatGPT, Claude, Gemini, Perplexity, and others

• Connect those sessions to specific brand mentions in LLM responses

• Build an AI referral dashboard for your marketing and ecommerce teams

For deeper context and KPI frameworks, see the related pillar guide: AI Recommendation Analytics: Measuring Referral Traffic from LLM Mentions (link internally from your blog).

Prerequisites: What You Need Before You Start

Before you instrument AI referral tracking, confirm these basics.

1. Analytics Stack Readiness

You’ll need:

• Google Analytics 4 (GA4) or equivalent analytics platform

• Admin access to configure:

  • UTM parameters

  • Custom channel groups

  • Events and dimensions

Google Analytics notes that session source/medium and UTM parameters are the core building blocks for attribution.

2. Tag Management & Data Layer

You should have:

• A tag manager (Google Tag Manager or similar)

• A standardized data layer for:

  • Page metadata

  • Campaign params

  • User events (add-to-cart, checkout, etc.)

3. Access to AI Recommendations

To connect brand mentions to traffic, you need visibility into:

• Your brand’s presence in ChatGPT, Claude, Gemini, Perplexity and other engines

• The exact prompts and answers that include your brand

You can gather this via:

• Manual testing and prompt logs

• An AI visibility platform like Era, which tracks:

  • Share of voice in AI answers

  • Rankings and citations

  • Pros/cons and sentiment

  • SKU-level inclusion in shopping carousels

Step 1: Define AI Referral Sources and Naming Conventions

Start by clearly defining how you’ll recognize AI referrals.

This makes later analysis and dashboarding consistent.

1.1 List Your AI Assistants

Create a simple table of AI engines you care about:

• ChatGPT (web + apps)

• Claude

• Gemini (Search + AI Mode)

• Perplexity

• Other AI search and shopping agents relevant to your market

Similarweb’s 2026 data shows ChatGPT’s share of gen AI web traffic dropped from ~76% to ~53%, while Gemini rose to ~27–28% and Claude to ~9%—multi-model tracking is now mandatory.

1.2 Standardize UTM source and medium

Use clear, machine-readable names so your reports are unambiguous.

Recommended conventions:

• utm_source=chatgpt.com

• utm_source=claude.ai

• utm_source=gemini.google.com

• utm_source=perplexity.ai

• utm_medium=ai_referral

OpenAI explicitly documents that ChatGPT search referrals include utm_source=chatgpt.com, so aligning with this convention avoids confusion.

1.3 Create an AI Campaign Taxonomy

Define how you’ll use utm_campaign and utm_content:

• utm_campaign=ai_recommendation for organic AI mentions

• utm_campaign=ai_integration for embedded tools or plugins

• utm_content=

  • brand_mention for general brand recommendations

  • sku_12345 for specific product/SKU recommendations

  • query_best_running_shoes for prompt-level detail

This structure lets you connect traffic to specific brand mentions and queries later.

Step 2: Detect and Tag AI-Driven Sessions in GA4

Now instrument GA4 so AI referrals are reliably detected and tagged.

2.1 Capture UTM Parameters on Landing

Ensure your analytics and tag manager capture:

• session_source / medium

• first_user_source / medium

• All UTM parameters

In GA4:

• Confirm session source/medium and first user source/medium are exposed in your standard reports

• Validate that visits from ChatGPT links show source=chatgpt.com when UTMs are present

Ahrefs notes that AI traffic is still underestimated, with some visits appearing as “direct,” so consistent UTM tagging is essential.

2.2 Create a Custom Channel Group for AI Traffic

Use GA4’s custom channel groups to classify AI referrals.

Steps:

1. Go to Admin → Data Settings → Channel Groups

2. Create a new channel group: AI Assistants

3. Add rules such as:

   • If source equals chatgpt.com → Channel: AI – ChatGPT

   • If source equals claude.ai → Channel: AI – Claude

   • If medium equals ai_referral → Channel: AI – Other

This makes AI traffic visible alongside Paid, Organic, and Direct in every report.

2.3 Track AI Referral Events

Create a custom event that fires for every AI referral session.

Example using GA4 + GTM:

• Trigger: Pageview where:

  • source in ['chatgpt.com', 'claude.ai', 'gemini.google.com', 'perplexity.ai']

  • OR medium equals ai_referral

• Event name: ai_referral_session

• Event parameters:

  • ai_engine (ChatGPT, Claude, Gemini, Perplexity)

  • ai_query (if available from URL or additional tagging)

  • ai_recommendation_type (brand_mention, sku_recommendation, comparison, etc.)

This event is the backbone of your AI referral dashboard and funnels.

Step 3: Instrument AI Referral Links and Deep Links

Attribution depends on the links AI assistants use.

You can’t control every engine, but you can influence and instrument many of them.

3.1 Optimize Links in AI Integrations and Tools

If you run:

• ChatGPT or Claude apps/plugins

• Gemini or Perplexity integrations

Make sure outgoing links include:

• utm_source set to the host assistant (e.g., chatgpt.com, claude.ai)

• utm_medium=ai_referral

• utm_campaign and utm_content aligned with your taxonomy

Example URL:

https://yourbrand.com/running-shoes?utm_source=chatgpt.com&utm_medium=ai_referral&utm_campaign=ai_recommendation&utm_content=sku_98765

https://yourbrand.com/running-shoes?utm_source=chatgpt.com&utm_medium=ai_referral&utm_campaign=ai_recommendation&utm_content=sku_98765

3.2 Encourage AI Assistants to Use Trackable URLs

For assistants you don’t directly control, you can:

• Ensure your canonical URLs support UTM parameters cleanly

• Use well-structured landing pages that AI engines are likely to cite

• Document your preferred tracking URLs in developer docs or API responses when you expose product data

Cloudflare found that AI bots generate massive crawl volumes (e.g., 38,000 crawls per visitor for Anthropic in July 2025), so clean, structured URLs increase the chances that assistants link predictably.

3.3 Handle “Direct” AI Traffic Edge Cases

Ahrefs reports that AI chatbots account for about 0.1% of traffic across ~35K sites and that this is likely understated because some AI traffic appears as direct.

Mitigation tactics:

• Analyze spikes in direct traffic that correlate with:

  • AI crawling activity

  • Known campaign pushes involving AI

• Use landing page grouping to identify pages predominantly reached via AI mentions

• Treat these as “probable AI referrals” and monitor them separately

This isn’t perfect attribution, but it gives you directional insight beyond standard channels.

Step 4: Connect AI Brand Mentions to On-Site Behavior

Tracking sessions is only half the equation.

You need to tie those sessions back to specific brand mentions and decision-stage queries.

4.1 Capture Prompt and Answer Context (Where Possible)

For AI surfaces you own (e.g., plugins, apps, shopping agents):

• Log:

  • User prompts (e.g., “best wireless earbuds under $200”)

  • Model answers that include your brand or SKUs

  • The rank and role of your brand in the answer (primary recommendation vs one of many)

Store this in:

• Your own database

• A data warehouse (BigQuery, Snowflake)

Connect each recommendation to:

• A landing URL

• A UTM-tagged click event

4.2 Use an AI Visibility Layer to Track Mentions Across Engines

Manual logging won’t scale across ChatGPT, Claude, Gemini, Perplexity, and emerging agents.

Platforms like Era provide:

• Multi-model monitoring of your brand’s share of voice

• Answer-level analytics:

  • Where your brand appears

  • How it’s positioned (pros, cons, sentiment)

  • Which SKUs and merchants are recommended

• Region and language filters so you can see differences in U.S., EU, etc.

You can then:

• Export Era’s AI recommendation data via API

• Join it to your GA4 events by:

  • ai_engine

  • ai_query

  • sku or product IDs

This creates a closed-loop view from AI mention → click → on-site behavior → revenue.

4.3 Build Decision-Criteria Attribution

Ahrefs found that AI citations don’t mirror Google rankings:

• 28% of ChatGPT’s most-cited pages have zero Google organic visibility

• 80% of sources cited across AI platforms are not in Google

This means AI assistants use different evidence than traditional SEO.

To interpret performance:

• Map AI decision criteria for your category:

  • Price and promotions

  • Availability and shipping speed

  • Trust signals (reviews, ratings, guarantees)

  • Specs and compatibility

• Annotate each AI recommendation with these criteria

• Compare on-site behavior:

  • Do AI referrals driven by “best value” convert differently than those driven by “premium quality”?

Over time, this helps you prioritize which evidence types to strengthen in your catalogue and content.

Step 5: Build an AI Referral Dashboard for ChatGPT, Claude, Gemini & More

With data flowing, you can build a practical AI referral dashboard.

5.1 Core Metrics to Include

Align your dashboard to the KPI stack emerging across AI visibility tools.

At minimum, track:

• AI referral sessions

  • Per engine (ChatGPT, Claude, Gemini, Perplexity)

  • Per region and device

• Share of AI traffic

  • % of total sessions from AI assistants

• Conversion metrics

  • Add-to-cart, checkout, revenue per session

  • Compare against other channels (Organic, Paid, etc.)

• Brand mention metrics (via Era or your logs)

  • Number of answers where your brand is recommended

  • Position (1st, 2nd, etc.) in recommendation lists

  • Sentiment and pros/cons

5.2 Example GA4 Exploration Layout

Use GA4’s Explore workspace to build a dedicated AI Referral Exploration.

Recommended tabs:

1. Traffic Overview

   • Rows: ai_engine, country

   • Columns: sessions, users, ai_referral_session events

2. Conversion by Engine

   • Rows: ai_engine

   • Metrics: purchases, revenue, conversion_rate

3. Landing Page Performance

   • Rows: landing_page

   • Filters: channel=AI – ChatGPT, AI – Claude, etc.

5.3 Join AI Recommendation Data in a Warehouse

For deeper attribution, push your GA4 data to BigQuery and join it with AI recommendation logs.

Data model example:

• Table ai_recommendations (from Era/API):

  • timestamp

  • ai_engine

  • ai_query

  • brand

  • sku

  • position

  • sentiment

• Table ai_sessions (from GA4 export):

  • session_id

  • ai_engine

  • utm_content

  • landing_page

  • revenue

Join on:

• ai_engine

• sku or utm_content where you encode SKU

• Time windows (e.g., recommendation within 24h of session)

This lets you answer questions like:

• “Which Claude queries produce the highest revenue per session?”

• “Where does Gemini recommend competitors over us, and what’s the impact?”

Step 6: Replace Legacy SEO Dashboards with AI-Focused Reporting

Traditional SEO dashboards don’t reflect how AI models surface brands in conversation.

Pew found that when Google users see an AI-generated summary, they click traditional search results in 8% of visits vs 15% when no summary appears.

Users rarely click cited sources inside the summary.

This shifts discovery into AI answer layers, not classic SERPs.

6.1 Expand Beyond Rankings and Organic Sessions

Your core SEO dashboards should now include:

• AI share of voice in ChatGPT, Claude, Gemini, Perplexity

• AI referral traffic and revenue alongside organic search

• AI citations vs Google rankings (to see the disconnect Ahrefs highlights)

6.2 Introduce AI Visibility KPIs

Borrow from AI visibility platforms like Era:

• Answer presence rate: % of relevant AI queries where your brand appears

• Recommendation rank: average position in AI recommendations

• Evidence completeness: coverage of key decision criteria in your catalogue and content

Tie these directly to:

• SKU-level performance

• Region-specific AI visibility

6.3 Operationalize AI GEO/AEO Programs

Use your new dashboard to drive ongoing Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) programs:

• Identify SKUs that appear frequently in AI answers but underperform on-site

• Fix:

  • Product page specs and clarity

  • Pricing and shipping competitiveness

  • Review volume and rating displays

• Feed improvements back into AI via:

  • Structured data and feeds

  • Updated content and landing pages

Era’s GEO plans and content autopilot can help automate:

• Query discovery across AI assistants

• Daily AI-optimized articles published directly to your CMS

• Technical GEO tasks aligned with how LLMs consume data

This moves AI visibility from an experiment to a predictable, revenue-linked program.

FAQ: Tracking Brand Mentions in AI and Referral Traffic

1. How can I tell if traffic is coming from ChatGPT or Claude?

You can detect AI-driven referrals primarily through UTM parameters and source/medium.

OpenAI says ChatGPT referrals include utm_source=chatgpt.com, and you can align Claude, Gemini, and Perplexity similarly.

Create a custom channel group in GA4 that classifies any session with source equal to these domains or medium=ai_referral as AI traffic.

2. What if AI referral visits show up as “direct” in analytics?

Ahrefs’ research suggests some AI traffic is misclassified as direct.

To mitigate this:

• Use UTM parameters wherever you control links

• Monitor landing pages that see unexplained direct traffic spikes following AI crawling or campaigns

• Treat these clusters as probable AI referrals and track them in a separate view

It won’t be perfect, but it improves visibility beyond default reports.

3. Are AI brand mentions and AI referral traffic the same thing?

No.

Brand mentions measure how often and how prominently AI assistants recommend your brand.

Referral traffic measures how many users click through to your site.

Ahrefs found that 28% of ChatGPT’s most-cited pages have zero Google organic visibility and 80% of AI-cited sources don’t appear in Google, so AI visibility is a distinct layer.

You need both metrics to understand the full picture.

4. Which tools help track brand mentions in AI assistants?

You can combine:

• Analytics platforms (GA4) for traffic and conversions

• AI visibility platforms like Era to:

  • Monitor brand mentions in ChatGPT, Claude, Gemini, Perplexity

  • Track share of voice, rankings, sentiment, and pros/cons

  • Provide SKU-level and merchant-level tracking for ecommerce

Together, they create a closed-loop AI recommendation analytics stack.

5. How does this tie into GEO/AEO and my broader SEO strategy?

Google says AI Overviews and AI Mode are grounded in core ranking systems and don’t require special optimization beyond strong SEO fundamentals.

However, AI assistants use additional signals: structured data, catalogue hygiene, reviews, and brand authority.

GEO/AEO focuses on exposing machine-readable evidence across these surfaces.

By tracking AI referrals and brand mentions, you can identify which evidence gaps to fix and prioritize GEO work that actually moves revenue and P&L, not just rankings.

By following these steps—defining AI sources, tagging sessions, instrumenting AI links, connecting mentions to behavior, and building AI-focused dashboards—you’ll be ready for the era where AI answer engines and shopping agents become the new front door for product discovery.

And with platforms like Era providing multi-model AI visibility and optimization, you can move from guessing about AI to operating it as a measurable, optimizable channel.