AEO Attribution Methodology

Any CMO can tell a CFO that AEO is working. The CMO who keeps their budget tells a CFO how it's working — the model, the assumptions, the math. This page is that math.

It is a deeper read than the AEO Pipeline Attribution page and the CFO Guide, intended for RevOps leads, attribution analysts, and anyone building the model themselves. The three pillars: stage-weighted probability, the first-touch-versus-multi-touch decision, and lookback window calibration.

Pillar 1. Stage-weighted probability

Pipeline dollars at face value are indefensible. A $100,000 opportunity in Discovery is not the same as a $100,000 opportunity in Commit. A CFO knows this; the attribution model must honor it.

Stage-weighted probability multiplies each opportunity's dollar value by its deal-stage probability before summing. The result is a probability-adjusted pipeline number that accounts for how likely each opportunity actually is to close.

The CFO question this preempts is the one every CFO eventually asks: "those deals aren't closed yet — how do I know they will?" The answer is, we've already discounted them by the same stage probabilities your pipeline coverage model uses. This isn't a new assumption; it's your assumption, applied to AEO.

Where stage probabilities come from

Do not invent them. Use the exact same stage probabilities that appear in your existing HubSpot deal stage configuration — the ones your RevOps and finance teams already agreed on for quarterly forecasting. Consistency across channels is what makes the comparison meaningful.

Typical B2B SaaS stage weights (for reference; yours may differ):

• Appointment Scheduled — 2%
• Qualified to Buy — 10%
• Discovery — 20%
• Decision Maker Bought In — 40%
• Contract Sent — 60%
• Commit / Legal Review — 80%
• Closed Won — 100%
• Closed Lost — 0%

What stage-weighted does not do

Stage-weighted pipeline is still pipeline, not revenue. It is a forecast. Closed-won revenue is a fact. The CFO memo reports both — stage-weighted pipeline as a forward-looking number, closed-won ARR from AEO-influenced deals as the hard number. CAC payback math uses the latter.

Pillar 2. First-touch versus multi-touch AEO

A buyer's journey through AI answers rarely involves a single citation. The typical shape (from customer interviews and our own product data): an initial ChatGPT query that introduces the category, a Perplexity check a few days later to corroborate, a blog post read, a quiet period of 2–6 weeks, an AI Overviews query with a more specific intent, and finally a form fill. How you distribute credit across those touches is the first-touch-vs-multi-touch decision.

First-touch

Give 100% of the attribution credit to the first AEO touchpoint within the lookback window. Simple, auditable, and it matches how buyers describe their discovery ("I first heard about this on ChatGPT").

Use first-touch when: Defending budget to a CFO. Reporting to the board. Answering the question "what channel introduced these buyers to our category." CFOs understand first-touch intuitively and it resists the "but they also saw a paid ad" rebuttal.

First-touch weakness: It over-credits the engine that happened to be sampled first, and under-credits the engine that closed the loop. Use in parallel with multi-touch for internal prioritization, not in place of it.

Multi-touch with time-decay

Distribute the credit across all AEO touchpoints on the contact record within the lookback window, weighted by recency. The most recent touch gets the largest share; earlier touches get exponentially less.

A typical time-decay curve: the most recent touchpoint gets 40%, the next 25%, then 15%, 10%, 6%, 4%. Adjust the decay constant based on your median sales cycle — faster cycles concentrate credit on recent touches, longer cycles distribute more evenly.

Use multi-touch when: Prioritizing prompts and content. Deciding which engines to over-invest in. Answering the question "which of our 40 tracked prompts produced the most pipeline this quarter."

Multi-touch weakness: Harder to explain to a CFO without a 15-minute math detour. Don't lead with it in a renewal review.

The mature setup: both, in parallel

First-touch in the monthly CFO PDF. Multi-touch in the weekly operational dashboard. The two numbers will differ; that's fine — each answers a different question. Document which one is reported where and why, so nobody gets ambushed in a meeting.

Lantern ships first-touch in the monthly PDF and multi-touch with 90-day time-decay in the product dashboard. Both are configurable — customers with longer sales cycles use 180-day lookbacks and different decay constants.

Pillar 3. Lookback windows

The lookback window is the most-contested assumption in AEO attribution — and the one most likely to be wrong if copied from traditional digital attribution.

30-day lookback is too short for AEO. The ChatGPT query that first introduces a buyer to your category typically happens 4–8 weeks before they fill out a form. A 30-day window would miss most first-touches.

60-day lookback catches most mid-market B2B SaaS deals. Use this only if your category has short sales cycles (under 30 days) and most contacts convert quickly after discovery.

90-day lookback is the default for B2B SaaS with sales cycles of 30–90 days. Catches the typical discovery-to-form-fill gap plus the sales cycle itself. This is Lantern's default.

180-day lookback is appropriate for enterprise SaaS, regulated industries (fintech, healthcare, legal), or any category where decision committees are large and sales cycles run 90–180 days. The trade-off: more noise, more cross-channel overlap, but you capture the real first touch.

Source property definitions

The attribution math is only as good as the source-property rules. What counts as an "AEO touchpoint"? Lantern's definition, which maps cleanly to CFO-defensible claims:

1. A self-reported source field match. Contact fills out "How did you first hear about us?" and the response contains one of: ChatGPT, GPT, Perplexity, Claude, Gemini, AI search, AI Overviews, an AI.
2. A UTM-tagged landing session. Session lands with utm_source=ai-search on any known AI-referred landing page.
3. A Direct session on a citation-known page within N minutes of a known citation serve. Session lands with no referrer on a page that was cited by a tracked engine within the past hour. Lower-confidence signal; typically gets half-credit.

Each touchpoint source is logged separately. Your model decides how to combine them. Lantern uses OR-logic for contact-level AEO flag (any of the three sources → aeo_influenced = yes), and weighted credit at the touchpoint level (self-reported = 1.0, UTM-tagged = 0.7, proximity = 0.4).

Why not just use last-click

A closing thought that matters more than it should. Last-click attribution structurally under-credits AEO. Here's why:

• AEO is a discovery channel. Buyers encounter your brand in answer engines early, before they're ready to convert.
• Between the AEO touchpoint and the conversion, buyers typically revisit via branded search, direct entry, or a paid retargeting ad.
• Last-click credits the final channel, which — for AEO-discovered buyers — is almost always not AEO.

The operational consequence: under last-click, AEO-influenced pipeline shows up as $0. The channel looks unprofitable. The budget gets cut. The pipeline it was creating evaporates over the next 2–3 quarters. Two quarters after the cut, a new CMO arrives wondering why pipeline is down.

First-touch with a lookback window — or a multi-touch model with any credit to the discovery position — is the only defensible approach for AEO.

Handling ambiguity honestly

Every attribution model has ambiguity. The CFO-safe response is to disclose it, not hide it. Three examples:

• Self-reported source is wrong sometimes. A buyer writes "ChatGPT" but actually found you on LinkedIn. Acceptable error rate: treat self-reported source as ~85% accurate. Cross-check with UTM data quarterly.
• Citations and conversions don't line up perfectly. You get cited by ChatGPT in week 2, then not again; the deal closes in week 10. Multi-touch credits the week-2 citation; first-touch credits the earliest AEO source field match. Both are defensible; pick one for the CFO report and be consistent.
• Proximity signals are low-confidence. Direct traffic within N minutes of a citation serve is probability, not proof. Weight accordingly — Lantern uses 0.4 — and document the weight in your methodology page.

Show the math. Disclose the assumptions. Let the CFO ask questions. That's the difference between a number that survives a renewal review and a number that doesn't.

Closed-won and CAC payback

Stage-weighted pipeline is the forward-looking number. Closed-won is the hard number. AEO CAC payback uses closed-won.

Target: under 12 months to survive a CFO review. Benchmark alongside paid CAC payback and SEO CAC payback on the same ACV assumption and the same gross margin — channel comparisons are meaningful only on the same denominator.

Methodology disclosure checklist

Any AEO attribution number presented to the CFO should be paired with the following disclosures on the same page:

• Stage-weighted probability model (with the HubSpot deal stage configuration inline)
• Lookback window in days
• First-touch vs multi-touch (and which time-decay constant if multi)
• Source-property definitions (self-reported match rules, UTM match rules, proximity rules and weights)
• Gross margin assumption
• Sample size (number of AEO-influenced opps in the reporting period)
• Known limitations (self-reported error rate, proximity-signal confidence)

A CFO who can inspect every assumption usually signs the renewal. A CFO who can't, doesn't.

How Lantern implements this

Lantern ships the full methodology above as the default — 90-day lookback, first-touch in the monthly PDF, multi-touch with time-decay (λ = 0.023) in the weekly dashboard, stage-weighted probability pulled from the customer's own HubSpot deal stage configuration, source-property weights at 1.0/0.7/0.4 for self-reported/UTM/proximity. All configurable for Enterprise customers.

$99/mo for the standard methodology. Enterprise for bespoke lookback windows, custom decay constants, and co-built CFO memo templates. No per-prompt pricing. No per-seat pricing.