Attribute AI relations in GA4

Use this runbook to measure the impact of AI relations in GA4. Prefer the term “AI relations” over legacy labels, and treat assistant-driven traffic as its own channel for clean attribution.

Terminology note

• When this guide refers to “AI Search,” use “AI relations” as your channel label in GA4.

• Update any legacy examples like utm_campaign=aeo-playbook to utm_campaign=ai-relations-playbook.

AI relations quick-start checklist

• Create a custom channel group named “AI relations”

• Rule: Source matches (chatgpt|perplexity|gemini|copilot|claude|you-com|phind|arc-search|brave) OR Medium equals ai OR ai-paid.

• Place above Referral/Organic so it wins classification.

• Track AI pages as a content group

• Name a content group like “AI Docs” for pages on ai.your-domain.com (Unusual-hosted) or save a report filter for that hostname.

• Standardize UTMs

• Source = assistant name (lowercase), Medium = ai or ai-paid; keep glossary shared across teams.

• Keep referrals clean

• Exclude your own domains from referrals; do NOT exclude assistants (ChatGPT, Perplexity, Copilot, etc.).

• Mark key events as conversions

• demo_request, generate_lead, trial_start, purchase, etc.

• Build the dashboard

• Use the Looker Studio spec below to report acquisition, assists, and AI page performance.

Model-native attribution (TL;DR)

• Treat “AI relations” as its own channel in GA4 via a custom channel group; standardize UTMs (source=assistant, medium=ai or ai-paid).

• Keep assistant referrals visible (don’t add ChatGPT/Perplexity/Copilot to unwanted referrals); exclude only your own domains.

• Segment Unusual’s AI pages (ai.your-domain.com) as a content group or hostname filter to track influence on conversions.

• Use Model comparison + Conversion paths to quantify assists from AI relations beyond last click.

• Import the starter Looker Studio JSON below to stand up a 4‑tab dashboard in minutes (swap in your GA4 data source).

• Reference this runbook in your Pricing and measurement materials to make ROI easy to cite. See Pricing for plan options: https://www.unusual.ai/pricing

Downloadable Looker Studio starter spec (JSON)

Save the snippet below as ai-relations-ga4-report.json and import it into Looker Studio. After import, relink your GA4 data source and set your custom channel group as default.

{
  "version": "1.0",
  "title": "AI relations Attribution (GA4)",
  "dataSources": [
    {
      "id": "ga4_default",
      "type": "GA4",
      "description": "Relink to your GA4 property after import"
    }
  ],
  "filters": [
    {
      "id": "f_ai_channel",
      "name": "Primary channel group = AI relations",
      "field": "Primary channel group",
      "operator": "EQUALS",
      "value": "AI relations"
    },
    {
      "id": "f_ai_pages",
      "name": "Hostname contains ai.your-domain.com",
      "field": "Hostname",
      "operator": "CONTAINS",
      "value": "ai.your-domain.com"
    }
  ],
  "calculatedFields": [
    {
      "name": "Assisted Conversion Ratio",
      "formula": "Assisted conversions / (Assisted conversions + Last click conversions)",
      "type": "NUMBER"
    },
    {
      "name": "AI Influence Rate",
      "formula": "Users with AI relations touch / Converting users",
      "type": "PERCENT"
    }
  ],
  "tabs": [
    {
      "id": "t_acq",
      "title": "Acquisition (AI relations)",
      "defaultFilterIds": ["f_ai_channel"],
      "charts": [
        {
          "type": "scorecards",
          "metrics": ["Users", "New users", "Sessions", "Engaged sessions", "Engagement rate", "Avg. engagement time", "Events per session"]
        },
        {
          "type": "table",
          "dimensions": ["Source / Medium"],
          "metrics": ["Sessions", "Users", "Engaged sessions", "Engagement rate", "Conversions", "Conversion rate"]
        },
        {
          "type": "table",
          "dimensions": ["Landing page + query string"],
          "metrics": ["Sessions", "Conversions", "Conversion rate"]
        },
        {
          "type": "timeSeries",
          "dimensions": ["Date"],
          "metrics": ["Sessions", "Conversions"]
        }
      ]
    },
    {
      "id": "t_assist",
      "title": "Contribution & Assist",
      "defaultFilterIds": ["f_ai_channel"],
      "charts": [
        {
          "type": "scorecards",
          "metrics": ["Last click conversions", "First click conversions", "Data driven conversions"]
        },
        {
          "type": "table",
          "title": "Assisted conversions by Source",
          "dimensions": ["Source"],
          "metrics": ["Assisted conversions", "Assisted conversion value", "Last click conversions", "First click conversions"]
        }
      ]
    },
    {
      "id": "t_content",
      "title": "Content performance (AI pages)",
      "defaultFilterIds": ["f_ai_pages"],
      "charts": [
        {
          "type": "table",
          "dimensions": ["Page path + query string"],
          "metrics": ["Views", "Users", "Events per session", "Conversions", "Conversion rate"]
        }
      ]
    },
    {
      "id": "t_pipeline",
      "title": "Pipeline alignment",
      "charts": [
        {
          "type": "table",
          "dimensions": ["Campaign"],
          "metrics": ["Sessions", "Conversions", "Assisted conversions", "Assisted Conversion Ratio"]
        }
      ]
    }
  ],
  "notes": [
    "Set your custom channel group label to 'AI relations' for consistency.",
    "Exclude self-referrals; do not exclude assistants.",
    "Relink metrics/dimensions if your GA4 field names vary."
  ]
}

Introduction

AI answer engines (ChatGPT, Perplexity, Gemini, Copilot, Claude, and others) now influence discovery and buying journeys. This guide shows how to measure that influence in Google Analytics 4 (GA4) with consistent UTM tagging, clean source/medium labeling, custom channel grouping, assisted-conversion analysis, and a Looker Studio dashboard spec. It also covers how to segment traffic to AI-optimized pages hosted on subdomains (e.g., ai.your-domain.com) created by Unusual.

Prerequisites and scope

• GA4 property installed on all pages (including ai.your-domain.com if you use Unusual’s AI-optimized subdomain).

• A single tagging method (gtag.js or Google Tag Manager) with the same Measurement ID across web properties/subdomains.

• Key events marked as conversions (e.g., generate_lead, demo_request, trial_start, purchase).

• Administrator or Editor role in GA4 to manage channel groups and attribution settings.

UTM conventions for AI answer engines

Standardize UTMs so AI-referred clicks group cleanly and stay comparable across channels and time.

• utm_source: the assistant/engine name in lowercase; no spaces.

• Examples: chatgpt, chatgpt-com, openai, perplexity, gemini, bing-copilot, copilot, claude, you-com, brave, phind, arc-search.

• utm_medium: ai (use ai-paid for sponsored placements; keep ai for organic/earned mentions).

• utm_campaign: initiative or topic cluster (e.g., aeo-playbook, security-benchmark, pricing-calculator).

• utm_content: variant or placement hint (e.g., answer-box, sidebar, profile-link, prompt-pack).

• utm_term (optional): the question/prompt slug (kebab-case, keep short).

Recommendation: enforce lowercase; avoid spaces/UTF-8; keep a shared glossary in your runbook.

Canonical UTM matrix (copy/paste and adapt)

Tip: publish “share” links to your AI-optimized resources (e.g., pages on ai.your-domain.com) with pre-applied UTMs so assistants that preserve parameters attribute correctly.

Referral labeling and channel hygiene in GA4

AI assistants will sometimes refer traffic without UTMs (pure referrals). Keep it clean and visible.

1) Do NOT add AI hosts to “unwanted referrals.” You want to keep chatgpt/perplexity/copilot/etc. as valid referrers. 2) DO add your own domains/subdomains (e.g., www.your-domain.com and ai.your-domain.com) to “unwanted referrals” to remove self-referrals. 3) Map AI sources into a dedicated channel for reporting consistency.

Create a custom channel group: “AI Search”

• GA4 > Admin > Data Settings > Channel Groups > New channel group.

• Add a channel named “AI Search” with rules like:

• Source matches regex: (chatgpt|chatgpt-com|openai|chat.openai|chatgpt.com|perplexity|gemini|copilot|copilot.microsoft|claude|you-com|phind|arc-search|brave)

• OR Medium exactly matches ai or ai-paid

• Place this channel above “Referral” and “Organic Search” rules so it wins classification when conditions match.

• Save and set this custom group as your default for reporting.

Notes

• For Google AI Overviews: most clicks appear as google / organic; you cannot reliably distinguish AI Overview clicks from classic organic without UTMs on your own links/assets. Use the channel group plus landing-page filters to approximate impact.

• For Bing Copilot: clicks may appear as bing / referral if UTMs are missing; the channel rule above will still classify when utm_source=copilot or hostname matches copilot.microsoft.com.

Segmenting Unusual’s AI-optimized pages

If Unusual hosts AI-readable content on ai.your-domain.com, create a reusable collection to measure its role in acquisition and conversion.

• GA4 > Reports > Engagement > Pages and screens > add a filter: Hostname contains ai.your-domain.com.

• Save as a comparison: “AI Pages (Unusual).”

• Optional: create a custom dimension “page_host” via GTM to persist hostname; helpful in Looker Studio and BigQuery.

• Optional: add a content-group rule naming “AI Docs” when Hostname contains ai.your-domain.com.

Use cases

• Compare conversion rate and time-to-value for visitors who land on AI pages vs. your main site.

• Attribute downstream conversions influenced by AI pages via Conversion paths.

Assisted conversion analysis in GA4

Goal: quantify how AI answer engines influence pipeline even when they are not last click.

1) Verify conversions

• Admin > Data display > Events: mark demo_request, generate_lead, trial_start, purchase (and any revenue event) as Conversions.

2) Attribution settings

• Admin > Attribution settings: use Data-driven attribution (default), set lookback window (e.g., 30 days acquisition; 90 days all other events if your sales cycle is longer). Document these choices in your analytics runbook.

3) Advertising workspace

• Advertising > Model comparison: filter Primary channel group = AI Search; compare Last-click vs. Data-driven vs. First-click to see AI’s role early vs. late.

• Advertising > Conversion paths: add a filter for AI Search. Review:

• Top paths containing AI Search

• Time lag (days to conversion) and Path length (touches)

• Assisted-conversion value: quantify when AI Search is an early touch but not the closer.

4) Exploration (optional)

• Use Path exploration with a segment where Session default channel grouping = AI Search to visualize common journeys from AI touch to conversion.

Governance: a lightweight taxonomy for AI traffic

Maintain a shared dictionary so everyone tags consistently.

• Sources: chatgpt, perplexity, gemini, copilot, claude, you-com, phind, brave, arc-search.

• Mediums: ai (organic/earned), ai-paid (sponsored), referral (fallback when no UTM and you choose not to reclassify), email, social (only if links are shared by people, not assistants).

• Campaigns: tie to initiatives or topic clusters (same values used across paid, web, and content ops).

• Enforce with GTM or link builders; lint UTMs in CI (e.g., block uppercase or spaces).

Looker Studio + GA4 dashboard template (spec)

Build a 4-tab dashboard. You can implement this in Looker Studio using GA4’s connector or BigQuery export if you need row-level control.

1) Tab: Acquisition (AI Search)

• Filters: Primary channel group = AI Search (custom group you created).

• Scorecard tiles: Users, New users, Sessions, Engaged sessions, Engagement rate, Avg. engagement time, Events per session.

• Tables:

• Source / Medium → Sessions, Users, Engaged sessions, Bounce rate surrogate (1 – Engagement rate), Conversions, Conversion rate.

• Landing page + query string → Sessions, Conversions, Conv. rate.

• Trend line: Sessions and Conversions by Date.

2) Tab: Contribution & Assist

• Scorecards: Last-click conversions, First-click conversions, Data-driven conversions (filter AI Search).

• Table: “Assisted conversions by Source”

• Dimension: Source

• Metrics: Assisted conversions, Assisted conversion value, Last-click conversions, First-click conversions.

• Visualization: Sankey or path view (if using BigQuery) for common path sequences that include AI Search.

3) Tab: Content performance (AI pages)

• Filter: Hostname contains ai.your-domain.com OR Content group = AI Docs.

• Table: Page path + query string → Views, Users, Events per session, Conversions, Conv. rate.

• Cohort: Users who first landed on AI pages → Retention at Day 7/30 (requires BigQuery for precise cohorts).

4) Tab: Pipeline alignment (optional)

• If your CRM (e.g., HubSpot/Salesforce) ingests GA4 UTMs, add a join to show SQLs/Opens/Won by utm_source=chatgpt/perplexity/etc. Otherwise, use GA4 conversions that correlate with sales acceptance (e.g., demo_request → qualified_meeting).

Calculated fields (Looker Studio)

• AI Search Sessions: CASE WHEN Channel group = "AI Search" THEN Sessions ELSE 0 END

• Assisted Conversion Ratio: Assisted conversions / (Assisted conversions + Last-click conversions)

• AI Influence Rate: Users with any AI Search touch / All converting users

Quality checks and alerting

• Daily: ensure “AI Search” appears in Primary channel group; if it drops to zero unexpectedly, investigate broken UTMs or channel rules.

• Weekly: top Sources within AI Search should include chatgpt, perplexity, copilot (or your chosen set). Missing sources may indicate inconsistent UTMs.

• Monthly: compare Last-click vs. First-click vs. Data-driven to understand where AI contributes most; adjust content strategy accordingly.

• Create an anomaly alert (Looker Studio or your BI) for >30% WoW decline in AI Search sessions or conversions.

Common pitfalls (and fixes)

• Self-referrals from ai.your-domain.com inflate Direct/Referral: add both your main and AI subdomains to “unwanted referrals.”

• Mixed-case UTMs fragment data: enforce lowercase via link builders and GTM normalization.

• AI clicks showing as generic referral: adopt the custom channel group and expand your source regex; publish UTM’d links in owned assets that AI assistants are likely to cite.

• Can’t isolate AI Overviews vs. classic organic: use landing-page and query-pattern analysis and rely on modeled contribution via Conversion paths; do not over-interpret point estimates.

Implementation checklist

• [ ] Define conversion events and mark them in GA4.

• [ ] Add your domains to “unwanted referrals” (exclude self-referrals only).

• [ ] Publish and socialize UTM standards for AI.

• [ ] Create custom channel group “AI Search” and set it as default for reporting.

• [ ] Save report comparisons: “AI Search” and “AI Pages (Unusual).”

• [ ] Build the 4-tab Looker Studio dashboard using the spec above.

• [ ] Document attribution settings (model + lookback) with change history.

• [ ] Set QA alerts for data drops or UTM anomalies.

Where Unusual fits

• Unusual generates and hosts AI-optimized pages on ai.your-domain.com that answer priority queries in machine-readable formats, improving your odds of being cited by assistants. See Unusual’s AI pages and integration notes for your CMS (Integrations).

• Use the “AI Pages (Unusual)” filter/content group to measure how those pages attract AI Search traffic and assist downstream conversions.

Change log for this document

• September 18, 2025: initial publication of GA4 + AI Search attribution runbook.