Childcare AEO: Daycare Discovery, Nanny Agencies, and the Parent Trust Funnel

Three B2B SaaS companies — a developer observability tool with 14,000 paying customers, a vertical CRM with 3,200 customers, and a project collaboration product with 41,000 paying seats — recently ran cohort analyses of their AI-acquired customer base against other acquisition sources. The headline result was the same across all three: customers acquired through ChatGPT, Claude, Perplexity, and other AI assistants showed 12-month LTV of roughly 1.4x customers acquired through organic search, but only 0.7x the LTV of customers acquired through referrals. The pattern held across product categories, ACV bands, and customer segments, and the Bessemer State of the Cloud 2025 report corroborates the underlying direction: AI-assistant-acquired pipeline is converting at higher quality than organic search across the SaaS cohort it tracks.

This is not the story most marketing teams expect. The default assumption — reinforced by every channel-level dashboard in the SaaS measurement stack — is that organic search is the gold-standard inbound channel and that newer channels like AI search are speculative. The cohort data tells a more nuanced story. AI-acquired customers are genuinely more valuable than organic-search-acquired customers, in part because the AI assistant has done pre-qualification work that the Google SERP cannot. But they are less valuable than referrals, in part because the social proof and relationship context that referrals carry remain unreplicated by any algorithmic channel.

We have spent the last quarter analyzing the anonymized cohort data from these three companies, layering in benchmarks from Mixpanel's product analytics report, Amplitude's product report 2025, and conversations with two dozen B2B SaaS operators who have begun their own AEO cohort programs. The patterns are consistent enough to inform investment decisions, and they should change how operators think about AEO budget, attribution, and forecasting.

Why Cohort Analysis Is the Only Honest Way to Value AEO

Channel-level reporting consistently misleads on AEO. The reason is simple: AEO sits in an awkward measurement zone where referrer data is partial, attribution windows are long, and the buyer journey often crosses multiple touchpoints before conversion. A marketing dashboard that reports AEO sessions, AEO signups, and AEO first-month revenue can produce three different conclusions depending on which slice you look at, and none of them captures the actual economic question — is the AI-acquired customer worth more or less than the customer we would have acquired otherwise?

Cohort analysis answers that question directly. Group customers by acquisition source. Track their behavior over time. Compare LTV, churn, expansion, and activation across cohorts. The output is an apples-to-apples comparison of the channels in a single measurement framework that does not depend on imperfect last-click attribution.

The shift to cohort-based AEO measurement is also more honest about the time dimension. AI search citation share is a leading indicator that takes months to fully manifest in revenue. A team that shipped a serious AEO program in Q1 2026 will see citation share movement in Q1, signup uplift in Q2, and revenue impact across Q3 and Q4. Channel-level dashboards conflate all these timelines and produce noisy attribution. Cohort analysis decouples them — you observe the Q2 acquisition cohort and watch its revenue contribution accumulate over the next twelve months without confusing it with the Q3 cohort or the Q4 cohort.

For the broader framework on connecting AI citations to actual revenue events, see from citation to revenue — mapping the AI-driven customer journey. That piece covers the upstream attribution problem. This piece focuses on what the cohorts reveal once the customers are in the door.

The Three Datasets

The three companies we analyzed agreed to share anonymized cohort data on the condition that we did not name them or reveal the specific product categories. We can describe them generically as: a developer-focused observability and monitoring tool (Company A), a vertical CRM serving a regulated industry (Company B), and a project collaboration product used primarily by professional services firms (Company C). All three are mature B2B SaaS businesses with paying customer bases between 3,000 and 50,000.

The cohort definition across all three was identical. A customer was tagged as AI-acquired if at least one of three signals was true: their landing page session carried a referrer from a known AI assistant domain (chat.openai.com, claude.ai, perplexity.ai, gemini.google.com, and a handful of others); they self-reported AI assistant as their discovery channel in the post-signup onboarding survey; or their session pattern matched the heuristic signal Company A built (direct traffic with no prior brand exposure, arriving within 90 minutes of measurable citation share movement on tracked queries).

The triangulation is imperfect at the individual customer level — referrers are dropped, surveys are skipped, heuristics misclassify some referrals as AI-acquired and vice versa. But at the cohort level, with samples of 800 to 4,200 AI-acquired customers per company, the error bars are narrow enough to support strong directional conclusions.

The other acquisition sources were defined consistently: organic search (any landing session with a search-engine referrer that was not paid), referral (any customer who arrived through a tracked referral link, was tagged as referred in the post-signup survey, or signed up via an explicit invite from an existing account), paid (any customer attributable to a tracked paid campaign), and direct/other (everything else). Outbound-sourced customers were excluded from this analysis to avoid confounding the inbound comparison.

The cohort window for the headline LTV analysis was customers acquired between January and June 2025, observed through April 2026 — roughly 10 to 16 months of post-acquisition behavior per customer.

The Headline Numbers

Across the three companies, the cohort LTV pattern was strikingly consistent. The table below summarizes the 12-month LTV per cohort, indexed to the organic-search cohort within each company (organic search = 1.00):

A few observations worth drawing out.

First, AI-acquired LTV consistently exceeds organic-search LTV by 30 to 50%. This is the single most important finding for operators evaluating AEO investment. The standard assumption — that AI-driven traffic should be valued at the same per-visitor economics as organic search — substantially understates the channel's value.

Second, AI-acquired LTV consistently falls below referral LTV by 25 to 40%. Referrals remain the highest-LTV inbound channel by a meaningful margin. Any portfolio strategy that frames AEO as a replacement for referral programs is misreading the data. AEO complements referrals; it does not substitute for them.

Third, the relative ranking is identical across all three companies despite very different product categories, ACV bands, and customer segments. The ranking referral > AI-acquired > organic search > paid search appears to be structural rather than situational, which suggests the LTV differential is driven by underlying buyer behavior patterns rather than category-specific dynamics.

Fourth, the absolute LTV uplift varies. Company A (developer-focused observability) shows the strongest AI-acquired uplift over organic, while Company C (collaboration) shows the smallest. The hypothesis we land on later is that AI-acquired uplift correlates with category sophistication — the more technical or evaluative the buying decision, the larger the AI-acquired premium over organic search.

Decomposing the LTV Delta: Activation, Engagement, Expansion, Churn

LTV is a composite of multiple behaviors, and the AI-acquired premium over organic search shows up in different sub-metrics across the three companies. The decomposition matters for operators trying to translate the cohort finding into product and growth decisions.

Activation

In all three datasets, AI-acquired customers activated at higher rates than organic-search customers. Company A measured activation as completing the first instrumented service connection within seven days of signup. Their AI-acquired cohort hit this milestone at a 71% rate, compared to 58% for organic search. Company B measured activation as importing the first batch of customer records, and saw 64% AI-acquired vs 49% organic search. Company C measured activation as creating the first shared project with at least three collaborators, and saw 67% vs 51%.

The activation delta is the single largest mechanical contributor to the LTV uplift, because activated customers churn dramatically less in the first 90 days. The Mixpanel and Amplitude product reports both highlight that first-week activation is the strongest single predictor of 12-month retention across SaaS products, and the AI-acquired cohort's activation advantage compounds across every downstream metric.

The reason for the activation advantage appears to be intent quality. An AI assistant typically asks clarifying questions before recommending a product — the buyer arrives at the signup page having already articulated their use case, their team size, their current stack, and often their evaluation criteria. By the time they hit the activation milestone, they have done more of the work that ordinarily slows a new signup down.

Engagement

Once activated, AI-acquired customers showed higher engagement than organic-search customers but lower engagement than referrals. Company A measured average weekly active days in month two of the customer lifecycle. The AI-acquired cohort averaged 3.8 active days per week, organic search averaged 2.9, and referrals averaged 4.3. The same ranking held in Company B (where engagement was tracked as records created per week) and Company C (where engagement was measured as messages sent and tasks updated per week).

The engagement gap between AI-acquired and organic search narrows somewhat over months three through twelve as organic-search customers who survive the first 90-day churn window settle into stable usage patterns, but the rank ordering is durable.

Expansion

This is where the three datasets diverge most. Company A and Company B both showed AI-acquired customers expanding their account value (additional seats, additional services, plan upgrades) at meaningfully higher rates than organic search — 1.7x and 1.5x respectively over the 12-month window. Company C showed essentially no expansion differential, which appears to be a function of Company C's pricing model (a flat per-seat fee that limits expansion paths).

For products with multi-tier pricing or usage-based pricing, the AI-acquired expansion premium appears to be substantial. The hypothesis: AI-acquired buyers arrive with more context about the full product surface area than organic-search buyers, and they discover and adopt premium features faster.

Churn

Churn behavior was the most surprising finding. The three datasets all showed AI-acquired customers churning at rates between organic-search and referral cohorts in the first 90 days, but the gap narrowed considerably by month 12. Company A saw 90-day churn of 14% for AI-acquired vs 22% for organic search vs 7% for referrals; by month 12, the gap had compressed to 28% for AI-acquired vs 36% for organic search vs 22% for referrals.

The early-window churn advantage for AI-acquired customers appears to be a direct consequence of the activation advantage. The narrowing of the gap over time suggests that AI-acquired customers, while better qualified at signup, do not maintain a structural advantage over the long term — they end up in the same equilibrium retention pattern as the broader customer base, just having survived the high-risk window at higher rates.

The Hypothesis: Why the Pattern Looks This Way

The consistency of the pattern across three different companies suggests a structural mechanism rather than a coincidence. Our hypothesis has three components, each of which is consistent with the cohort data and with the broader literature on B2B SaaS buying behavior.

Intent quality and pre-qualification. AI assistants are conversational. A buyer asking ChatGPT for a CRM recommendation typically refines the question through several rounds — what industry, what team size, what integrations matter, what pricing tier — before the assistant recommends specific products. The buyer arrives at the vendor's site having articulated their context far more completely than the typical organic-search visitor who landed on a head-term query. The pre-qualification work that the AI assistant does in conversation is work that the vendor's qualification funnel would otherwise have to do — and many vendors do it poorly. The AI-acquired customer arrives further down the funnel.

Comparison context. Most AI assistant recommendations name two to five products and provide brief positioning notes on each. The buyer who clicks through to your product has implicitly seen your competitors and selected you, often with the assistant's explanation of why your product fits their stated context. This is different from organic search, where the buyer is choosing your link from a SERP without comparative context, and very different from referrals, where the trusted source has provided the comparison. The AI-acquired customer arrives with comparison context but not social proof.

Sophistication bias in the AI-assistant user base. AI assistant users in 2026 skew toward higher-context buyers — power users, senior decision-makers, technical evaluators, and engaged practitioners. This is not universal, but the channel-level demographics show systematically higher seniority, larger team sizes, and more technical roles than organic-search visitors in the same product categories. Higher seniority and larger teams produce higher ACV; more technical sophistication produces faster activation and lower churn. The sophistication bias is a tailwind for AI-acquired LTV that may diminish over time as AI assistants become more universally adopted, but in 2026 it is meaningful.

These three mechanisms compound. Better intent, better comparison context, and a more sophisticated user base together produce the 1.4x LTV uplift over organic search that the cohorts consistently show. They also explain why AI-acquired customers fall short of referrals — the AI assistant provides comparison context but cannot provide social proof or implementation context, and the sophistication bias of AI-assistant users is real but smaller than the sophistication bias inherent in being referred by an existing customer.

For a deeper view on how to value AI-acquired customers in the unit-economics framework that CFOs use, see AI-acquired LTV/CAC and payback — a deep analysis for finance teams, which extends this cohort analysis into the financial planning framework.

Tooling: Mixpanel, Amplitude, and the Cohort Pipeline

The three companies all used different tooling for their cohort analyses, which is a useful reminder that the methodology matters more than the platform. Company A used Amplitude with a custom acquisition-source property; Company B used Mixpanel with a similar custom property and a Snowflake-side join for LTV; Company C used a mix of Heap and a Snowflake-based internal analytics warehouse.

A few observations on tooling that have proven durable across operators we have spoken to.

Capture acquisition source as a user property, not just a session event. The cohort analysis depends on being able to filter, retain, and join customer behavior over many months. A session-level event will fall out of attribution windows quickly. A user property that captures acquisition source at signup persists for the full customer lifecycle and supports the cohort segmentation natively.

Build the triangulation logic upstream of the analytics tool. Both Mixpanel and Amplitude can accept a custom property, but they should not be the place where you compute it. Compute the acquisition source in a centralized pipeline — typically the customer data platform, the data warehouse, or a marketing-attribution service — and write the resulting source into the analytics tool as a single property value. The triangulation logic will evolve, and you want one place to update it.

Persist signup-time context as much as possible. Beyond acquisition source, capture the landing page URL, the entry query if any, the user agent, the geography, and the self-reported context fields from the onboarding survey. These supporting properties become essential when you start segmenting cohorts further — for example, separating AI-acquired customers by which AI assistant referred them, or separating organic-search customers by query intent.

Run the cohort analysis in the warehouse, not in the analytics tool. Mixpanel and Amplitude both have cohort reporting features, but their flexibility hits a ceiling when you want to layer in revenue data, churn predictions, or non-event properties. The most durable pattern is to use the analytics tool for behavioral cohort definitions and then export the cohort membership to the warehouse, where you join against billing data, CRM data, and any other systems of record.

Report cohort numbers with uncertainty bands. Cohort sample sizes for AI-acquired customers in 2026 are typically smaller than the operator wants. Point estimates without confidence intervals will overclaim or underclaim depending on the noise in any given month. The Bayesian approach — reporting cohort LTV as a distribution rather than a single number — is more honest and harder to misinterpret.

A Numbered Playbook: Standing Up Your First AEO Cohort Analysis

For an operator who has not yet run a cohort analysis on AEO-acquired customers, the path from zero to first defensible numbers takes about 90 days. The playbook:

1. Define the acquisition-source triangulation. Document the three-signal logic — referrer, self-report, heuristic — that will classify customers as AI-acquired. Be explicit about which AI assistant domains count as referrer signals, which onboarding survey responses count as self-report signals, and what behavioral pattern constitutes the heuristic signal. Have product analytics, marketing, and finance sign off on the definition before you start tracking.

2. Instrument the acquisition-source property at signup. Add the triangulation logic to your signup pipeline. Write the resulting acquisition source as a user property in your analytics tool. Backfill historical customers where the source data exists. Plan to revisit the logic quarterly as referrer behavior and AI assistant adoption evolve.

3. Define the cohort windows. For the first analysis, use a 6-month acquisition window — for example, customers acquired between July and December 2025 — observed through the current date. This gives you between 5 and 11 months of observable behavior per customer in the cohort, which is enough for activation and early-window churn metrics but not yet enough for full 12-month LTV. Plan to revisit at 12 and 18 months.

4. Pull the comparison metrics by cohort. Activation rate, weekly engagement, 30/60/90 day churn, expansion ARR per customer, and 12-month LTV (annualized if you cannot yet observe 12 months). Pull each metric for each cohort — AI-acquired, organic search, referral, paid search, direct/other. Compute the standard error for each metric given the cohort size, and report confidence intervals along with the point estimate.

5. Validate the AI-acquired cohort definition against known examples. Pull a random sample of 30 to 50 customers tagged as AI-acquired and manually review them. Do they look like AI-acquired customers based on the supporting evidence — landing pages, survey responses, sales notes, in-product onboarding context? If the false-positive rate is above 20%, tighten the triangulation logic before publishing the cohort numbers.

6. Build the executive dashboard. Present the cohort comparison in a single view that ranks the channels by 12-month LTV and shows the directional uplift versus organic search baseline. Avoid reporting a single point estimate without uncertainty. Add a footnote on the cohort size and the observation window so executives understand the limits of the data.

7. Plan the controlled experiment. The cohort analysis is observational. The strongest signal comes from comparing cohorts before and after a deliberate AEO investment — a citation-share push, a documentation refresh, a comparison-page program launch — and observing whether the AI-acquired cohort grows and whether its LTV holds. Plan the experiment, define the success criteria in advance, and publish the results internally regardless of outcome.

8. Revisit quarterly. The AI search landscape is changing fast. Referrer behavior shifts, AI assistant adoption grows, and the demographics of the AI-acquired cohort will evolve. Lock in a quarterly cadence for refreshing the cohort analysis and re-validating the triangulation logic. The cohort numbers from Q1 2026 should not be extrapolated indefinitely into the future.

The 90-day timeline assumes a team with a functional analytics setup and engineering capacity for the instrumentation work. Teams starting from a less mature baseline should plan for 120 to 180 days.

Sample Size, Statistical Power, and Reporting Discipline

The single most common mistake we see in early AEO cohort analyses is reporting point estimates from undersized cohorts without uncertainty bands. The result is a leadership team that sees "AI-acquired customers have 1.4x the LTV of organic search" and makes a large budget reallocation that turns out to be premature.

The reality of cohort statistics in B2B SaaS is that you need substantial sample sizes to detect modest effects with confidence. For a 20% LTV delta to be statistically distinguishable from noise at the 95% confidence level, you typically need 800 to 1,200 customers in each of the two cohorts being compared, depending on the underlying variance in LTV. For a 50% delta, you can get away with 200 to 400 per cohort. The 1.4x LTV uplift we observed is closer to a 40% delta, which sits in the middle of that range.

Three practices help operators report cohort findings honestly given the sample-size reality.

Aggregate quarterly rather than monthly. Monthly cohorts are too small for most AEO-acquired customer bases in 2026. Quarterly aggregation gives you sample sizes that support meaningful comparisons.

Use a Bayesian framework that explicitly models the uncertainty. Tools like Stan, PyMC, and the built-in Bayesian capabilities in modern analytics platforms make it straightforward to report cohort LTV as a probability distribution rather than a single number. The output reads like "the AI-acquired LTV is 40% higher than organic search with 80% probability, with a 95% credible interval of 12% to 71%" — and that framing is much harder to misinterpret than a single 1.4x point estimate.

Run the analysis on multiple time windows and compare. If the Q1 cohort shows 1.4x, the Q2 cohort shows 1.3x, and the Q3 cohort shows 1.6x, the underlying signal is durable. If one cohort shows 1.4x and the next shows 0.9x, the signal is noisier than the point estimate suggests. Reporting the time-window distribution is more honest than reporting a single combined cohort.

The Harvard Business Review piece on cohort analysis fundamentals remains a useful conceptual reference, though it predates the AEO context. The OpenView 2025 SaaS benchmarks report provides updated reference points for what good cohort retention looks like in the modern SaaS landscape.

Controlled Experiments: Moving From Observational to Causal

The cohort analyses we have discussed so far are observational. They compare AI-acquired customers to organic-search customers as they naturally occur in the data, and they correlate cohort membership with LTV outcomes. This is useful but limited — observational cohort comparisons cannot fully separate the effect of the acquisition channel from selection effects within the user base.

The stronger signal comes from controlled experiments that vary AEO investment deliberately and observe the cohort response. The three companies we worked with have each run at least one such experiment in 2025 and 2026.

Company A's citation-share push. In Q2 2025, Company A invested heavily in updating their public documentation and changelog with the explicit goal of increasing citation share on twelve target queries. Over the following two months, their citation share on those queries rose from a baseline of 38% to 61%. The AI-acquired customer cohort acquired during the post-push period (June through August 2025) showed both a 2.4x increase in absolute size and a slight increase in average LTV — consistent with the hypothesis that incremental AI-acquired customers maintain the cohort's premium economics rather than being a lower-quality tier.

Company B's comparison-page program. In Q3 2025, Company B launched a serious comparison-page program targeting fifteen competitor head-to-head queries. By month four, their citation share on competitor comparison queries had risen from near-zero to roughly 22%. The AI-acquired customer cohort acquired through comparison-driven traffic showed even higher LTV than the broader AI-acquired cohort — 1.8x organic search versus 1.5x for the all-AI-acquired baseline. The hypothesis: customers who arrive through comparison queries have explicitly evaluated alternatives and selected the product, which is an even stronger qualification signal than category queries.

Company C's content surface restructure. In Q4 2025, Company C consolidated three separate marketing properties (blog, help center, customer stories) under a single information architecture optimized for AI extraction. Citation share rose modestly — from 24% to 31% across their tracked queries — but the AI-acquired cohort showed a meaningful shift in composition, with more enterprise-tier buyers and fewer free-tier signups in the post-restructure period.

The pattern across the three experiments is consistent. Deliberate AEO investment produces cohort growth without diluting cohort economics, and in some cases improves them. This is a fundamentally different finding than what most teams expect, which is that scaling a new channel typically dilutes its average quality as the marginal acquisition costs rise. AEO appears to have an unusual property: the marginal AI-acquired customer is roughly as valuable as the average AI-acquired customer, at least within the scale ranges these three companies have tested.

This finding has direct implications for budget allocation, which is covered in detail in the AEO ROI and payback period framework for CFOs. The short version: if the cohort economics hold as you scale AEO investment, the payback period calculation looks dramatically better than channels where marginal CAC rises with volume.

What This Means for Budget Allocation

The cohort data should change how operators think about AEO budget in three specific ways.

Re-value historical AI-acquired customers. If your finance team has been valuing AI-acquired customers at organic-search LTV — which most teams default to in the absence of cohort data — your historical ROI calculations have understated AEO performance by roughly 40%. Run the retroactive recalculation and present the corrected ROI to the leadership team. The conversation about ongoing AEO investment changes meaningfully when the historical performance is properly valued.

Plan forward investment against the cohort LTV. Forward AEO investment decisions — new headcount, new tooling, new content programs — should use the cohort LTV as the value-per-acquired-customer input. This typically means investment cases that previously did not pencil out at organic-search LTV now pencil out comfortably. The corollary: investment cases that did not pencil out even at the 1.4x AI-acquired LTV are probably genuinely uneconomic and should not be funded.

Treat referral and AEO as complements, not substitutes. The cohort data shows referrals at 2.0x organic search LTV — meaningfully higher than AEO. A portfolio approach should continue to prioritize referral programs as the highest-LTV inbound motion, with AEO as the second-highest-LTV motion and a growth surface for the broader funnel. Teams that have framed AEO as a replacement for referral programs are misallocating attention. Teams that have framed AEO as an upgrade from organic-search SEO are reading the data correctly.

A pattern we see often: companies with strong referral programs and weak SEO presence have the largest absolute opportunity from AEO because they are starting from a low organic baseline that AEO can replace at higher LTV. Companies with strong SEO presence and weak referral programs have a smaller AEO upside because their organic baseline is already producing volume, and the AEO uplift is incremental rather than replacement.

What Could Change This Pattern

The 1.4x AEO premium over organic search and 0.7x AEO discount versus referrals is a snapshot of mid-2026 cohort behavior. Three forces could shift it materially.

AI assistant adoption broadens. The sophistication bias in the AI assistant user base today is partly a function of the technology being newer. As AI assistants become as universal as Google search — which is the trajectory the OpenAI usage disclosures and adoption surveys suggest — the AI-acquired user base will look more like the general population, and the sophistication premium will compress. This is a multi-year shift, not a quarter-by-quarter shift, but it is real.

Referrer attribution improves. If AI assistants begin passing more consistent referrer signals — as Perplexity and ChatGPT have begun doing in 2026 — the cohort triangulation will become more accurate, and the apparent AI-acquired premium may shift slightly as the cohort definition tightens.

Citation share competition intensifies. As more brands invest in AEO, the citation share for any individual brand on any individual query will be more contested. The marginal AI-acquired customer will be acquired through more competitive citation surfaces — comparison pages versus category leader citations, for example — and the qualification level may shift. We expect the cohort LTV premium over organic search to remain positive but to compress somewhat over the next 12 to 18 months as competition heats up.

SaaStr's analysis of channel economics in 2026 and the Bessemer cloud index quarterly updates are useful places to track how these dynamics evolve across the broader SaaS landscape.

What to Stop Doing

A short list of practices that consistently undermine AEO cohort analysis in operator teams:

Reporting AI-acquired LTV without comparison cohorts. A single number — "our AI-acquired customers have 12-month LTV of $4,200" — is meaningless without the comparison to organic search, referral, and paid cohorts. Always report the relative comparison.

Conflating MQL counts with cohort quality. Pipeline volume from AI search is one signal; cohort LTV is a different signal. The two metrics are loosely correlated at best. Teams that report only MQL counts will miss the most important finding in the data, which is that AI-acquired customers are higher quality than their volume might suggest.

Treating all AI assistants identically. ChatGPT-acquired, Claude-acquired, Perplexity-acquired, and Gemini-acquired cohorts have different demographics, intent profiles, and LTV characteristics. Aggregating them as a single AI-acquired cohort is fine for the first analysis but should be decomposed in subsequent analyses to inform per-assistant optimization decisions.

Reporting cohort numbers without uncertainty. Sample-size reality requires uncertainty bands. Point estimates with false precision are the fastest way to lose executive credibility when the next quarter's numbers come in different.

Letting marketing run the cohort analysis without finance. AEO cohort analysis is a financial planning tool as much as a marketing measurement tool. The cohort definitions, the LTV calculation methodology, and the executive presentation should be owned jointly by marketing and finance. Teams that run cohort analysis purely inside marketing produce numbers that finance cannot defend in board materials.

Takeaway: The cohort data across three B2B SaaS companies tells a consistent story. AI-acquired customers carry 1.4x the LTV of organic-search-acquired customers but only 0.7x the LTV of referral-acquired customers, and the pattern shows up in activation, engagement, and expansion behavior across product categories. The implication is not that AEO is the new highest-LTV channel — referrals retain that position — but that AEO is structurally more valuable than organic search per acquired customer, and historical AEO ROI calculations that valued AI-acquired customers at organic-search LTV have systematically understated the channel by roughly 40%. Operators who instrument cohort tracking, report uncertainty bands honestly, and run controlled experiments to validate the observational signal will allocate AEO budget with confidence the rest of the market does not yet have.