Two numbers explain why the budget conversation around LLM marketing strategy tends to go badly.
The first: AI referral traffic accounts for 1.08% of all website traffic across ten major industries, according to Conductor's 2026 benchmarks, drawn from an analysis of more than 3.3 billion sessions.
The second: in Conductor's survey of more than 250 enterprise marketing leaders, 97% reported that AEO had a positive impact on their marketing funnel in 2025, and 94% said they plan to increase AEO investment in 2026. That is a strong signal from an AEO-active sample, not a neutral cross-section of the market: 73% of respondents classified their programs as advanced or very advanced.
A finance team looking at the first number closes the budget. A marketing team looking at the second one opens it. Both are reading real data.
Both can be true because session volume can be a poor proxy for AEO's business impact. Many performance-marketing frameworks are centered on clicks, while AI answers can create exposure, preference, and shortlisting before a website visit—or without one.
That does not make AEO ROI unmeasurable. It means the measurement has to be built deliberately, using the right metrics on the right time horizon. Here is how to think about it.
What is an LLM marketing strategy, and where does ROI fit?
An LLM marketing strategy is a coordinated program for influencing how large language models describe, cite, and recommend a brand when people ask questions inside AI assistants and AI-powered search. It spans four areas: technical readiness so models can crawl and parse your content, on-site content structured for extraction, off-site authority signals across the third-party sources models actually cite, and measurement of how the brand appears in generated answers.
Answer engine optimization is the execution layer. Measurement is the layer that determines whether the work survives its first budget review, and it remains a weak point for many programs.
So the practical question for a marketing leader is not whether to have an LLM marketing strategy. It is how to prove one is working in a channel that was not built to report on itself.
Why the standard ROI math breaks
Three assumptions behind click-centered reporting weaken at once.
The click is disappearing as a unit of measurement
SparkToro's 2026 analysis found that fewer than one third of Google searches still send a click to the open web. The share of searches generating at least one click fell 9.51 percentage points between 2024 and 2026, a 22.9% decline. AI Overviews now appear on more than 20% of searches, and when they do, click-through rates drop by nearly 60%.
Cost per click, sessions per dollar, and traffic-driven content ROI all assume a click exists. Increasingly it does not, and that is a design feature of the surface rather than a temporary condition.
Some AI-driven visits are hard to attribute
When someone clicks a link in ChatGPT and the browser passes a referrer—or the destination URL carries campaign parameters—GA4 can attribute the visit. GA4 now includes an AI Assistants channel in its default channel group for supported sources such as ChatGPT. The gap is not that ChatGPT clicks are inherently invisible. It is that the attribution signal does not survive every path from an AI product to a website.
Visits can still appear as Direct when a referrer is absent, as can happen with copied URLs, some apps and webviews, redirects, or privacy controls. Loamly reported that 14,413 of 20,428 AI-associated visits in its 446,405-visit dataset appeared as Direct, or 70.6%. But Loamly's "dark AI" classification combines deterministic signals with behavioral inference, with confidence scores that vary by detection method. The 70.6% figure is therefore a vendor-specific estimate, not a universal rate of confirmed AI referrals.
Loamly also reported a 10.21% transactional conversion rate for its inferred dark-AI cohort versus 2.46% for non-AI traffic. That suggests a potentially valuable blind spot, but it should not be treated as proof that unobserved AI traffic converts at that rate across other sites.
Google's own AI surfaces create a separate limitation: GA4 includes traffic from AI Overviews and AI Mode within Organic Search, so those visits cannot currently be isolated from standard Google organic traffic in default reporting. GA4 can measure observable AI referrals; it cannot measure the full influence of answers that produce no click or a later return visit.
The industry has named this as the problem
Conductor's enterprise survey found the two biggest AEO challenges of 2025 were producing AI-optimized content at scale and measuring ROI. McKinsey's smaller September 2025 survey of roughly 30 Fortune 500 consumer-brand CMOs put a finer point on the same problem: only 16% said they systematically tracked AI search performance.
That demonstrates a measurement gap; by itself, it does not establish either strong or weak returns. But incomplete measurement is enough to lose a budget cycle.
The reframe: value per session, not sessions
Many published datasets—especially in B2B, SaaS, and other research-heavy buying journeys—show AI-referred visitors converting better than organic visitors. The size of the advantage varies sharply, and the direction is not universal across every industry or conversion type.
Knotch's audience journey tracking, cited in Conductor's benchmarks, found that visitors referred directly from LLMs convert at twice the rate and in one third the number of sessions compared with other traffic sources across its clients. Semrush research published in 2025 reported a 4.4x average across more than 500 high-value digital marketing and SEO topics. Opollo's study of 312 B2B technology firms measured 14.2% for AI referrals against 2.8% for Google organic. Ahrefs, analyzing its own traffic, found that 0.5% of visitors from AI search drove 12.1% of signups.
These figures do not measure the same population or even the same conversion event. Most published studies come from marketing and technology companies measuring their own traffic, which is the population most likely to use AI search in the first place. When AI traffic is around 1% of sessions, small absolute numbers produce volatile percentages. Citation behavior also shifts with model updates. In ecommerce, published lifts are more modest, and one large transactional study cited by Marshal found organic search outperforming ChatGPT referrals by about 13%.
The practical move is to build the business case on your own conversion and revenue-per-session data. If you need a planning assumption before enough first-party data exists, the roughly 2x Knotch result is more conservative than the highest published multiples, but it should be modeled as a scenario with a lower-bound case—not promised as a forecast.
The operating implication matters more than any single multiple. AEO should be evaluated on value per visit as well as volume. If reporting ranks channels only by session count, a small but qualified referral stream will look like a rounding error. Adding conversion rate, revenue per session, and pipeline contribution shows whether it deserves more investment.
A three-tier measurement model
One way an LLM marketing strategy gets killed is being judged with late-stage metrics while it is still producing early-stage results.
Tier 1: presence. Citation rate and share of answer across a defined prompt set, competitive share within that set, sentiment and framing in the answers, and which third-party sources are driving citations in your category. This tier moves first, and none of it appears in GA4.
Tier 2: qualified traffic. Properly attributed AI-referred sessions, engagement depth, and landing page mix. That last one is underrated: it tells you whether AI visitors are arriving at revenue pages or only at blog posts, which is the difference between a visibility win and a pipeline win.
Tier 3: revenue. Conversions from AI-attributed sessions, self-reported attribution on lead forms, and pipeline influence.
The tiers run from leading to lagging indicators, but they do not move in a perfectly fixed sequence. One practitioner benchmark from Goodie suggests first mentions can appear within one to two weeks for well-structured content on an established domain, consistent citation patterns may take eight to twelve weeks, and measurable business impact may take three to six months. Those ranges are useful for planning, not an industry-wide service-level agreement.
The rule that follows: a program in month two should be evaluated primarily on Tier 1 progress and measurement readiness, not Tier 3 alone. Revenue evidence still matters when it exists, but its absence at that stage is not a reliable verdict.
Fix the measurement before you judge the program
Four steps, in order of effort:
Start with GA4's native AI Assistants channel. Check the Session default channel group and Session source/medium dimensions to see which supported assistants already pass identifiable traffic. A custom channel group can still be useful for adding sources that Google's maintained list does not yet recognize or for applying your own naming rules. Neither approach can recover sessions that arrive without a referrer or campaign parameters.
Add self-reported attribution. Put an AI assistant option in the "how did you hear about us" field on lead forms. This can capture influence that technical attribution misses, including no-click discovery and later return visits. It is direct evidence from the buyer, but recall and response bias mean it should be combined with analytics rather than treated as complete ground truth.
Apply UTM discipline where you control the link. Add campaign parameters to links in partner content, newsletters, profiles, and other placements you manage. This distinguishes traffic from the placement itself; it does not prove that a separate AI citation caused the visit.
Use deep Direct as a diagnostic, not attribution. Growth in new users landing directly on interior comparison, FAQ, or article pages can justify investigation, especially when it moves with citation visibility or self-reported AI discovery. It is not proof: bookmarks, shared links, copied URLs, privacy tools, and tracking defects can produce the same pattern.
None of this produces a complete number, and the reporting should say so plainly. A dashboard that states a best estimate, names the methods behind it, and acknowledges the gap is more defensible in a budget review than one claiming a precision the channel does not currently allow.
What the business case for an LLM marketing strategy actually rests on
Three arguments hold up under scrutiny. They are worth making in this order.
Efficiency rather than volume. AEO should not be evaluated on session count alone. For research-heavy journeys, published benchmarks suggest it may compare favorably on cost per qualified visit or revenue per session. The budget case should use the organization's own numbers as soon as the sample is large enough.
Much of the citation environment lives off-site. McKinsey found that, in many cases, a brand's own websites make up only 5% to 10% of the sources AI search references. In an analysis limited to Google AI Overviews, Ahrefs found branded web mentions had the strongest measured correlation with brand visibility at 0.664, versus 0.218 for backlinks. Correlation does not prove causation, and Google AI Overviews do not represent every answer engine. The practical point is narrower: first-party content alone does not control the source environment, and credible third-party presence takes sustained work.
Competitive investment is rising. In Conductor's enterprise sample, respondents allocated an average of 12% of digital budgets to AEO in 2025, 56% made a significant or high investment, and 94% planned to increase investment in 2026. Those intentions do not prove that today's positions will persist or that entry costs must rise. They do show that competitors are allocating resources to a visibility layer that cannot be assessed by referral sessions alone.
There is an honest counterweight, and it is better raised by you than by your CFO. Conductor's own VP of services, looking at the same fractional percentages, cautions that this will not move as fast as the doomsday commentary suggests, and that organic search remains the cornerstone of most digital programs. AEO is not an argument for defunding SEO. It is an argument for a separate line with separate KPIs and a separate clock.
A defensible ask therefore has five parts: a named prompt set, a documented baseline, a two to three quarter horizon, tiered reporting against that horizon, and a decision point agreed in advance.
Three things not to promise
Do not promise complete channel-wide ROAS. You can report return from identifiable AI-referred sessions and controlled campaigns. What you cannot honestly roll into one precise figure is the additional influence from no-click answers, missing referrers, and later return visits. Label directly attributed return separately from modeled or self-reported influence.
Do not reverse-engineer revenue from citation counts. A citation is not a session and a mention is not a lead. Modeled revenue built on citation volume is a number with no denominator, and it will not survive its first serious review.
Do not measure only where measurement is easy. ChatGPT accounted for 87.4% of identified AI referral traffic in Conductor's ten-industry dataset, so a framework anchored only on observed referrals will be heavily weighted toward ChatGPT. Other platforms may still shape category perception without sending comparable measurable traffic, which is why presence metrics should complement referral data.
The bottom line
The case for an LLM marketing strategy should not rest on a universal conversion multiple. Published referral studies suggest that AI visitors can be unusually valuable in research-heavy journeys, while enterprise marketers with mature programs report positive funnel impact. Neither finding guarantees the same return for every company.
The practical question is whether your organization can measure observable revenue, leading visibility indicators, and the remaining attribution gap well enough to make disciplined funding decisions. Many organizations are still building that capability.
The brands that solve measurement first will not just report AEO better. They can make better spending decisions while competitors are still working from a less complete picture.
Sentient AEO helps brands build and measure AI search visibility across ChatGPT, Google AI Overviews, and Perplexity, with additional coverage for Gemini, Google AI Mode, Grok, and Claude. If you're trying to understand where your brand stands in the AI answer layer, get in touch with us for an AEO audit: [email protected]
Citations
-
AI referral traffic accounts for 1.08% of all website traffic across 10 industries; 87.4% of AI referral traffic comes from ChatGPT; Google Analytics reports AI Mode, AI Overview, and organic data together; LLM-referred visitors convert at twice the rate in one-third the sessions (Knotch) — Conductor, The 2026 AEO / GEO Benchmarks Report: https://www.conductor.com/academy/aeo-geo-benchmarks-report/
-
97% of CMOs and digital leaders reported positive AEO impact in 2025; 94% plan to increase investment in 2026; 56% made a significant or high investment; enterprises allocated an average of 12% of digital budgets to AEO in 2025; top challenges were content at scale and measuring ROI — Conductor, The State of AEO / GEO in 2026: CMO Investment Report: https://www.conductor.com/academy/state-of-aeo-geo-report/
-
Fewer than one third of Google searches send a click in 2026; click-generating share fell 9.51 percentage points since 2024, a 22.9% decline; AI Overviews appear on more than 20% of searches and reduce CTR by nearly 60% — SparkToro via Search Engine Land, June 2026: https://searchengineland.com/google-zero-click-searches-2026-study-479717
-
In Loamly's 446,405-visit dataset, its detection system classified 14,413 of 20,428 AI-associated visits as dark AI; that inferred cohort had a 10.21% transactional conversion rate versus 2.46% for non-AI traffic — Loamly, The AI Traffic Attribution Crisis: https://www.loamly.ai/blog/ai-traffic-attribution-crisis
-
Loamly distinguishes deterministic human referrals from "Likely from AI" visits identified through behavioral signals — Loamly, How AI Website Traffic Detection Works: https://www.loamly.ai/docs/features/ai-traffic-detection
-
GA4's default channel group includes AI Assistants when the referrer matches Google's maintained source list; Google AI Overviews and AI Mode remain part of Organic Search — Google Analytics Help, Default Channel Group: https://support.google.com/analytics/answer/9756891
-
Only 16% of respondents in a survey of roughly 30 Fortune 500 consumer-brand CMOs said they systematically tracked AI search performance; in many cases, brand websites comprised only 5–10% of AI-search references — McKinsey, New Front Door to the Internet: https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/new-front-door-to-the-internet-winning-in-the-age-of-ai-search
-
Opollo's study of 312 B2B firms reported 14.2% conversion for AI referrals versus 2.8% for Google organic; Ahrefs reported 12.1% of signups from 0.5% of its visitors — AirOps, AI Referral Traffic vs Organic Search: https://www.airops.com/blog/ai-referral-traffic-conversion-rates
-
Semrush's 4.4x figure covered more than 500 marketing and SEO topics; published conversion studies have sample bias, attribution fragmentation, small denominators, and temporal instability, and transactional ecommerce can behave differently — Marshal, AI Search Traffic Value: https://www.runmarshal.com/field-notes/ai-search-traffic-is-4x-more-valuable-than-organic
-
Practitioner benchmark: first AI mentions within 1–2 weeks, consistent citation patterns at 8–12 weeks, and measurable business impact at 3–6 months — Goodie, How Long Does It Take to See Results From AEO: https://higoodie.com/blog/how-long-does-aeo-take/
-
Branded web mentions correlate 0.664 with AI Overview brand visibility versus 0.218 for backlinks, across 75,000 brands — Ahrefs, An Analysis of AI Overview Brand Visibility Factors: https://ahrefs.com/blog/ai-overview-brand-correlation/
-
GA4 supports custom AI-assistant channel groups for sources outside or beyond the maintained default, while
page_referreris the traffic-source input when no campaign parameters override it — Google Analytics Help and Developer Documentation: https://support.google.com/analytics/answer/13051316, https://developers.google.com/analytics/devguides/collection/ga4/reference/config#page_referrer -
Self-reported attribution and controlled UTM parameters can supplement incomplete referral data — Discovered Labs, AEO Performance Metrics: https://discoveredlabs.com/blog/aeo-performance-metrics-what-to-measure-and-how-to-track-ai-citations
