The business case for GEO, in numbers that survive checking

Buyers moved first#

In March 2026, G2 surveyed 1,076 B2B software buyers. 51% said they now start product research with an AI chatbot more often than with Google. Twelve months earlier that figure was 29%.[1] In the same survey, 69% said chatbot guidance led them to choose a different vendor than they originally planned, and one in three bought from a vendor they had never heard of before the chatbot named it.[1]

This is not a B2B quirk. Pew Research found 49% of US adults now use AI chatbots, up from 23% in 2023, and 60% read the AI summary at the top of search results.[2] Semrush surveyed 643 US B2B professionals and 92% said AI shaped their vendor shortlist.[3]Forrester puts AI use somewhere in the most recent purchase process at 94% of B2B buyers.[4] Four independent studies, four methodologies, one direction.

The traffic converts better, which is the part that matters#

Visits from AI answers are still a minority of total traffic, and skeptics lean on that. The skeptics are measuring volume when they should be measuring intent. Adobe Analytics, watching over a trillion visits to US retail sites, found AI-referred traffic up 138% year over year by May 2026, and, more importantly, found that AI-referred visitors now convert 54% higher than other visitors.[5] A year earlier the same feed showed AI traffic converting worse than average. The quality flipped as the engines got better at sending people to the right place.

Similarweb clickstream data shows the same pattern from a different angle: referral traffic from ChatGPT converts at 7.1%, second only to paid search at 7.8%, and ahead of organic, social, email, and direct.[6] When ChatGPT started rendering clickable brand links in May 2026, referrals to tracked sites jumped 157.7% in a week.[6] And on Google itself, Seer Interactive measured 5.5 million queries and found that being cited inside an AI Overview roughly doubles your clicks per impression, while ranking organically under an AI Overview you are absent from cuts your clickthrough by about a third.[7]

The engine does the shortlisting before the click happens. What arrives on your site is a visitor who has already been told you are the answer. That is why this traffic converts like paid search you did not pay for, and why absence from the answer costs more than the raw visit counts suggest.

The budget already moved, quietly#

Conductor surveyed 250+ leaders at companies with 500+ employees: on average they put 12% of their digital budget into answer engine optimization in 2025, and 94% plan to increase it in 2026.[8] That is survey talk, so here is the harder evidence: payroll. Citizens Bank is hiring an Answer Engine Optimization Manager at $131K to $171K base. AWS, Home Depot, JPMorgan Chase, Capital One, Vanguard, Palo Alto Networks, and Texas Instruments all have GEO or AEO titled roles open.[9] Companies do not create titled headcount for trends they consider hype.

The supplier side confirms it. Dentsu launched a formal GEO consulting service.[10]Publicis shipped an agentic GEO offering through Performics and Sapient.[11] On the software side, Profound raised $96M at a $1B valuation in February 2026 with Fortune 500 logos on its customer page,[12] Sitecore paid a reported $225M for the GEO startup Scrunch,[13] and Adobe bought Semrush and shipped AI brand visibility as a native enterprise feature.[14]You do not have to believe any vendor's pitch to read acquisition prices.

What doing it in house actually takes#

Here is the part most coverage gets wrong. The hard part of GEO is not secret knowledge. The playbook is increasingly public, and the talent gap is a time lag, not a mystery: it is mostly senior SEOs who spent the last 18 months reskilling. If you have one of those people, keep them. The durable problem is operational, and it has three layers.

First, one measurement is worthless. AI engines answer the same question differently every time you ask. Peer-reviewed work on measuring AI visibility finds you need repeated runs per query before per-brand estimates stabilize, roughly seven or more, with confidence intervals, per engine.[15]Checking ChatGPT once a week by hand tells you approximately nothing.

Second, there are five to nine engines that matterand they cite almost entirely different sources, so each needs its own tracking. Google's free Search Console data now covers its own AI surfaces,[16]which is genuinely useful and covers exactly one vendor's engines.

Third, the ground moves without notice. When ChatGPT swapped models on May 23, 2026, SISTRIX measured 47% of citation share in German-language answers changing hands within 48 hours. Indeed lost 47% of its citations, Expedia lost 60%.[17] No changelog, no announcement, discovered only by vendors running continuous measurement. Classical SEO had named, dated Google updates. GEO has unannounced model swaps that reshuffle half the deck in a weekend.

This is why the in house question splits the way it does. Product truth, brand voice, and final approval cannot leave your building, and any vendor who claims otherwise is selling the content-farm version of GEO that platforms are already learning to punish. The measurement infrastructure, the cross-engine sampling, the reprioritization every time a model swap reshuffles the deck: that is the part where a shared platform beats a part-time analyst, because statistics do not care how smart your analyst is.

Does optimization actually work, and what happens at scale#

The foundational academic result says yes: the Princeton, Georgia Tech, and IIT Delhi team behind the original GEO paper measured up to 40% visibility gains from content changes as unglamorous as adding concrete statistics, quotable sentences, and real citations, across a 10,000 query benchmark.[18]In the field, Profound's published Ramp case study shows AI visibility on one page family going from 3.2% to 22.2% in a month.[19]The levers are real.

Now the honest part, which vendors skip. An AI answer has roughly ten citation slots. Visibility is share of a fixed pie, and the original research measured gains that came partly by displacing whoever held the slot before.[18] If every brand in your category optimized equally well tomorrow, the gains would cancel. So why is this still worth money? Because that is exactly what SEO has been for twenty five years, a zero sum contest that sustains a market north of $80B a year, for two reasons that transfer directly: execution quality never equalizes, and once you hold a slot, the spend flips from offense to defense. You are no longer paying to gain visibility. You are paying not to be the brand that quietly vanished when the model swapped in May.

The window argument follows from the same math. Consideration sets in AI answers are sticky: a stable few brands keep their slots across runs while everyone else rotates or never appears. Claiming a slot while your category sleeps is cheap. Displacing an incumbent after the category wakes up is not.

The numbers we refused to use#

A business case is only as good as its worst citation, and GEO marketing is full of numbers that do not survive a checking pass. Three examples we threw out of this article, so you know what the filter caught.

“Search volume will drop 25% by 2026.”Gartner's February 2024 prediction is probably the most quoted statistic in GEO sales decks.[20] It did not happen. Google still holds over 90% of search as of mid 2026, and the prediction drew methodological criticism the week it was published.[21] The real story, buyers moving high-intent research into chatbots while raw search volume holds up, is better supported and more interesting.

“Enterprise GEO contracts average $185,000 a year.” This one circulates with a precise-looking service breakdown, attributed to vendor blogs. We fetched the pages it supposedly comes from. The figure appears on neither. As far as we can tell it is a hallucinated statistic that AI summaries keep laundering into new articles, which, given the subject matter, is almost poetic.

McKinsey's trillion-dollar projections. Widely quoted figures on AI search revenue attributed to McKinsey trace only to secondary blog aggregations. We could not locate a primary McKinsey report containing them, so they are not in this article. Worth noting in the same breath: McKinsey's actual 2026 CMO survey ranks AI seventeenth on the priority list,[22] a useful reminder that GEO budgets today are moving bottom up through SEO and digital teams, not down from the boardroom.

We publish our own measurement methodology for the same reason we checked everyone else's: this category's biggest credibility problem is unverifiable numbers, from vendors whose tools disagree with each other on identical prompts. If a number cannot be traced, it should not move budget. Yours or anyone's.

Sources

1. 51% of B2B buyers start research with AI more often than Google (was 29% in 2025); 69% switched vendors on chatbot guidance; 1 in 3 bought from a previously unknown vendor (n=1,076, fielded March 2026): G2, “The Answer Economy”, April 2026.
2. 49% of US adults use AI chatbots (23% in 2023); 60% read AI summaries in search (n=5,119, fielded February 2026): Pew Research Center, June 2026.
3. 92% of B2B professionals say AI shaped their vendor shortlist; 83% say it influenced the final decision (n=643): Semrush, 2026.
4. 94% of B2B buyers used generative AI during their most recent purchase: Forrester, The State of Business Buying 2026, January 2026.
5. AI-referred traffic to US retail +138% YoY; AI-referred visitors convert 54% higher (Adobe Analytics, 1T+ visits): Digital Commerce 360 on Adobe Analytics data, June 2026.
6. ChatGPT referral conversion 7.1% vs paid search 7.8%; +157.7% weekly referral jump after clickable brand links: Similarweb, Gen AI Stats, May 2026.
7. Citation inside an AI Overview roughly doubles clicks per impression; organic CTR drops when an AIO is present and you are absent (53 brands, 5.47M queries): Seer Interactive, April 2026.
8. 12% of digital budget on AEO in 2025; 94% increasing in 2026 (250+ leaders, 500+ employee orgs; vendor-run survey, read with that in mind): Conductor, State of AEO/GEO, April 2026.
9. Fortune 500 GEO/AEO job requisitions incl. Citizens Bank AEO Manager at $131K to $171K: Everything PR compilation, July 2026; direct posting example at J.Jill careers.
10. Dentsu launches GEO consulting service: Dentsu newsroom, May 2026.
11. Publicis “Agentic Ready” GEO offering (Performics + Publicis Sapient): Publicis Groupe, 2026.
12. Profound raises $96M Series C at $1B valuation: Fortune, February 2026.
13. Sitecore acquires Scrunch AI for a reported $225M (price per Bloomberg reporting, not confirmed by Sitecore): Adweek, June 2026.
14. Adobe completes Semrush acquisition and ships Adobe Brand Visibility: Adobe newsroom, April to June 2026.
15. Repeated-run sampling with confidence intervals as the minimum for stable AI visibility measurement: Schulte, Bleeker and Kaufmann, “Don't Measure Once: Measuring Visibility in AI Search (GEO)”, arXiv:2604.07585, 2026.
16. Google Search Console generative AI performance reports (free, Google surfaces only): Google Search Central Blog, June 2026.
17. First documented ChatGPT core update (May 23, 2026): 47% of German-language citation share redistributed within 48 hours; Indeed −47%, Expedia −60%: SISTRIX, June 2026.
18. Content changes (statistics, quotations, citations) lift AI answer visibility by up to 40% on a 10,000 query benchmark: Aggarwal et al., “GEO: Generative Engine Optimization”, arXiv:2311.09735 (KDD 2024).
19. Ramp AI visibility 3.2% to 22.2% in one month (vendor-published case study, no contract terms disclosed): Profound, 2026.
20. The original prediction: 25% search volume decline by 2026: Gartner press release, February 2024.
21. Contemporaneous methodological critique of that prediction: Search Engine Journal, 2024. Google's 90%+ share as of mid 2026 per StatCounter-based retrospectives.
22. AI ranked 17th on CMO priorities; 94% report no meaningful progress integrating AI into marketing operations: McKinsey State of Marketing 2026, as reported by Entrepreneur, April 2026.

Every figure above was traced to the linked source during a July 2026 research pass; vendor-run surveys and vendor case studies are labeled as such and should be weighted accordingly. Three widely circulated statistics were excluded because they failed verification, and we say which ones. Noetio's own measurement methodology, including per-engine sampling counts and confidence intervals, is published at noetio.com/methodology.