Calculate CAC from Perplexity and ChatGPT Referrals: A Practical Model

By maxaeo.ai ｜ Published 2026-10-09 ｜ Updated 2026-10-09

To calculate CAC from Perplexity and ChatGPT referrals, divide the acquisition costs assigned to AI search by the number of new customers attributable to those platforms. The difficult part is not the formula—it is separating observable referral customers from AI-assisted customers whose sessions appear as direct, organic, or another channel.

This model solves that problem with three outputs: direct referral CAC, evidence-weighted CAC, and platform-level CAC where cost allocation is reliable.

What Counts as CAC From AI Search?

AI search CAC is the cost of acquiring a new customer through discovery or evaluation in an AI answer engine. It should count customers—not sessions, form submissions, qualified leads, or citations—and use the same acquisition-cost policy applied to your other channels.

For a basic calculation:

Direct AI referral CAC = AI search acquisition costs ÷ new customers with an observed AI referral

Include costs that would disappear if the AI search program stopped:

• Content creation and updates assigned to answer engine optimization
• GEO or AEO strategy and research
• Analytics implementation and attribution maintenance
• AI visibility monitoring software
• Agency, contractor, or allocated employee costs
• Relevant sales costs, if your company includes sales expenses in channel CAC

Do not automatically charge the program for an entire SEO team, website redesign, or content library. Allocate shared costs using documented labor hours, content usage, or another consistent rule.

How Do You Identify ChatGPT and Perplexity Customers?

Start with session-level referral evidence, then connect identified visitors to CRM opportunities and closed customers. GA4 can report the source and medium associated with a session, but the analytics record alone does not prove that the session produced a new customer.

Create an “AI referrals” channel containing known source values such as:

chatgpt.com
chat.openai.com
perplexity.ai

In GA4, review Traffic acquisition using the Session source / medium dimension. Google defines this dimension as the source and medium associated with a new session in its Traffic acquisition documentation. (support.google.com)

Pass the original source, landing page, and timestamp into hidden form fields or your customer data platform. Then preserve them through:

1. Form submission or signup
2. Lead and account creation
3. Opportunity creation
4. Closed-won status
5. New-customer validation

Exclude employees, test visits, existing customers, bots, duplicate accounts, and renewals. A crawler request is not a referral session, and a referral session is not a customer until the CRM confirms it.

Why Direct Referral CAC Is Usually Incomplete

Direct referral CAC is a lower-confidence denominator because some AI-influenced visits arrive without a usable referral source. When traffic-source information is unavailable, Google Analytics can classify the session as (direct) / (none), including cases involving redirects, missing parameters, or tracking interference. (support.google.com)

AI influence can also precede a later branded search, direct visit, sales conversation, or return session. Therefore, report three evidence tiers instead of forcing every customer into one attribution bucket.

This avoids two common errors: treating every direct customer as AI-influenced and claiming every AI mention generated revenue.

A more complete attribution design is covered in the AI search attribution model for enterprise SaaS.

How Do You Calculate Evidence-Weighted AI CAC?

Evidence-weighted CAC adds only the defensible share of assisted customers to the direct-customer denominator. Each assisted conversion receives a confidence weight based on the strength of its supporting evidence.

Use these formulas:

Weighted assisted customers =
Σ (assisted customer × confidence weight)

Evidence-weighted customers =
direct referral customers + weighted assisted customers

Evidence-weighted AI CAC =
AI search acquisition costs ÷ evidence-weighted customers

A practical weighting policy might assign:

• 0.75: Buyer names the AI engine and the CRM journey supports the claim
• 0.50: Buyer selects “ChatGPT or another AI assistant,” with matching timing
• 0.25: Sales notes mention AI research, but no platform or date is confirmed
• 0.00: Only aggregate traffic, citations, or brand-search growth is available

These weights are an internal accounting policy, not universal benchmarks. Approve them before reviewing results, apply them consistently, and show both weighted and unweighted counts.

For pipeline measurement before customers close, use a separate CRM framework for attributing pipeline to AI search engines. Do not substitute pipeline value for the customer count in CAC.

Worked Example: A Quarterly AI Search Cohort

Consider an illustrative SaaS program that spends $30,000 in one quarter and acquires 12 customers with observable ChatGPT or Perplexity referrals. Its direct referral CAC is $2,500, before assisted conversions are considered.

Direct AI referral CAC = $30,000 ÷ 12 = $2,500

Suppose another 10 new customers report using an AI assistant, but have no preserved referral. After applying the predefined evidence rules, they represent five weighted customers.

Evidence-weighted AI CAC = $30,000 ÷ (12 + 5)
                         = $1,764.71

Report both figures. The first is more observable; the second represents a broader but explicitly modeled view.

Can You Calculate Separate ChatGPT and Perplexity CAC?

Calculate platform-level CAC only when costs can be assigned to each engine using a defensible allocation method. Dividing total spend by each platform’s customers would charge the same dollars twice and produce misleading comparisons.

Engine-specific allocation may be reasonable when you track:

• Dedicated content or experiments for each platform
• Platform-specific monitoring and analyst hours
• Landing pages associated with identifiable referral sources
• Separate agency deliverables or research projects

If most work supports both engines, keep one combined AI search CAC. Report ChatGPT and Perplexity customer counts, conversion rates, average contract values, and payback periods as diagnostic metrics rather than manufacturing separate CAC figures.

Visibility data can explain why those metrics change. MaxAEO monitors mentions, citations, recommendations, sentiment, and competitive positioning across eight AI engines with daily updates. Its visibility data should be treated as explanatory evidence—not automatically counted as acquisition. The daily AI search tracking workflow provides a process for connecting these leading indicators to business reporting.

What Should an AI Search CAC Dashboard Include?

A useful dashboard separates financial outcomes from visibility and attribution-quality indicators. Executives should see the resulting economics, while channel operators need enough diagnostic detail to identify why CAC moved.

Track these fields by monthly or quarterly acquisition cohort:

• Direct and weighted new customers
• Direct and evidence-weighted CAC
• ChatGPT and Perplexity referral sessions
• Visitor-to-lead and lead-to-customer conversion rates
• Revenue, gross margin, and CAC payback period
• First-touch, last-touch, and assisted customer counts
• Self-reported attribution completion rate
• Percentage of customers with corroborating evidence
• Brand mention rate, recommendation position, and cited sources
• Cost-allocation assumptions and any policy changes

Do not retroactively change weights to improve the result. When attribution rules change, recalculate historical periods or mark the break clearly. For a wider financial view, connect CAC to the pipeline model for calculating AEO ROI.

Common Questions

Should leads be used instead of customers?

No. Dividing costs by leads calculates cost per lead, not customer acquisition cost. Use closed, validated new customers for CAC and report lead economics separately.

Should AI visibility software count as an acquisition cost?

Yes, when it supports the AI search acquisition program. If the software also supports reputation management or research, allocate only the relevant portion.

How often should AI referral CAC be calculated?

Monthly reporting can reveal direction, but quarterly cohorts are often more stable for B2B SaaS because sales cycles delay customer outcomes. Keep the attribution window consistent.

What if no AI referral customers have closed?

Report costs, referral sessions, qualified pipeline, and visibility indicators, but mark CAC as unavailable. Dividing by zero or substituting leads would create a false result.

What is the most defensible way to calculate CAC from Perplexity and ChatGPT referrals?

Use CRM-confirmed direct customers as the primary denominator, publish assisted customers as a separately weighted calculation, and disclose every included cost and attribution rule. This produces a number finance teams can audit without overstating AI search’s contribution.