Chapter 2

AI Demand and Substitution

Generative AI is the two-sided term in Gartner's case. Management calls it the highest-demand topic it sells advice on, and the pipeline reflects that pull. The same technology appears in the company's own 10-K as a substitution risk: large language models that could answer client questions without a Gartner subscription. This chapter weighs both sides against what the numbers actually show. The honest read: through 2025 the demand side is visible in the book and the substitution side is disclosed but not yet measurable in renewals — a lag that keeps the structural question open rather than closed.

The threat, in Gartner's own words

Gartner's Insights subscriptions produced about 78% of revenue in 2025 [1]. What that revenue buys a client is access to written research, Magic Quadrants, and analyst inquiry. A general-purpose LLM answers questions in the same shape. The company says so directly in its risk factors, and the language has escalated in step with the technology.

Source: Form 10-K risk factors, FY2022 [2], FY2023 [3], FY2024 [4], and FY2025 [5].

The FY2022 filing treated AI as background disruption. The FY2023 filing — the first written after ChatGPT's launch — introduced the specific mechanism: AI chatbots that "may provide substantive content … in query responses to users which could reduce the need to enter our websites" [6]. By FY2025 the disclosure carries two sharper edges: third parties building AI that "could reduce demand for our products," and clients loading Gartner's proprietary insights into an LLM, which "could reduce the value of our offerings" [7]. This is boilerplate in the sense that every research firm now carries it — but it is Gartner naming, in its most-scrutinized document, the exact way the franchise could erode.

The demand side, visible in the book

Against that disclosure sits a demand signal that is larger and easier to measure. Across 2024–2026 management has described AI as the single biggest topic clients ask about, and the supporting activity is concrete rather than rhetorical.

Source: Q4 FY2025 earnings call, CEO remarks [8].

In Q2 2025 the CEO called AI "the single largest demand area across all the topics we cover for virtually every role" [9]. By Q4, AI accounted for more than 6,000 documents in the library and roughly 200,000 in-depth client conversations during the year [10]. In Q1 2026, with contract-value growth still soft, AI remained "one of the most requested topics across all the roles we serve" [11]. The bull framing is straightforward: enterprises facing an unfamiliar, fast-moving technology buy independent guidance on how to adopt it, and Gartner sells the map.

The counter-argument is equally clean, and it is the crux of the structural thesis: demand for advice about AI is not the same as demand for Gartner. If a buyer can extract a serviceable answer from a general-purpose model, the willingness to pay for a syndicated subscription weakens even as interest in the subject rises. The demand figures above confirm engagement with the topic; they do not, on their own, prove the subscription is the thing clients still need.

The direct test: what would show erosion, and whether it has

The useful discipline here is to separate what management asserts from the metric that would contradict it. Gartner tracks the reason for every lost renewal and every lost new-business opportunity, and it asks its salesforce to log any instance of a client naming AI as a substitute [12]. On that internal evidence the CEO's read is specific: clients raising AI as a replacement for Gartner is "one that we do not hear frequently," and in Q4 2025 it was "less of an issue or less confirmed than even before" [13]. Asked in Q1 2026 whether the consulting slowdown reflected "something structurally worse… given AI," he attributed it to deferred client decisions rather than substitution [14].

That is management's best evidence, and it is real evidence — a tracked loss-reason series, not a slogan. The limitation is timing. Gartner's contracts run twelve months or longer, so a client who is quietly substituting does not register as a loss until the renewal comes due; the loss-reason log is a lagging read on a book that reprices slowly. The metric that would show early erosion — wallet retention, the dollars a retained client base spends year over year — did roll below 100% in 2025, with GTS wallet retention falling to 96% [15]. Management attributes that break to the US federal pullback, tariffs, and deferred budgets rather than AI — a case the report's first chapter (The Subscription Engine) documents in detail. The reading that fits the evidence: the metric that could reveal AI substitution turned down in 2025, but the identifiable drivers behind the turn are cyclical, not the model on a client's desk.

Sources: Q2 FY2025 [16] and Q4 FY2025 earnings calls, renewal uplift [17] and loss-reason tracking [18]; FY2025 10-K, wallet retention [19] and risk factors [20].

AskGartner and the moat

Gartner's answer to the substitution risk is to put an LLM in front of its own proprietary content rather than let a public one stand in for it. AskGartner, launched to licensed users in August 2025 and fully rolled out by October, is a generative-AI interface whose responses are grounded in Gartner's research rather than the open internet — the company's distinction is that its answers carry "direct references to our distinctive insights" drawn from data behind its firewall [21]. The early tell that matters most for the thesis: licensed users who used AskGartner "had substantially higher renewal rates than those who did not, even with the same levels of engagement" [22].

That is a genuine defensive signal — the same technology framed as the threat is being used to raise retention — but three cautions belong next to it. It is a first-cycle, self-reported cross-tab, not yet a retention rate proven across a full renewal cohort. Heavier AskGartner users may be the more-committed clients who were always going to renew. And the tool is a cost as well as a shield: the 10-K notes AI has "required, and will continue to require, additional investment and increased costs" [23]. AskGartner's value depends on the proprietary corpus underneath it — the 500,000 annual analyst conversations and 27,000 vendor briefings a public model cannot see — which is where the durable part of the moat, if there is one, actually sits.

Pricing: the pressure point to watch

If AI erodes the franchise, the first place it shows up may be price, not volume — a client who can get 70% of the answer free negotiates harder on the subscription. Asked directly in Q1 2026 whether AI competition was forcing a pricing rethink, the CEO said clients "feel our pricing is appropriate," and that when price is cited as a reason not to buy it is usually broader cost-cutting rather than Gartner's price relative to a substitute [24]. The CFO's structural defense is that Gartner sells to the top of the org chart — the CIO, CFO, and their teams — "where there tends to be much less price sensitivity" [25]. Plausible, and consistent with the model — but it is management's characterization, not a disclosed price-realization series, and it is the claim most worth testing as the AI tools improve.

What would change the read

On the evidence through Q1 2026, the substitution threat is disclosed and mechanically credible but not yet visible in the numbers, while the demand pull is visible and large; the cyclical explanation for the 2025 stall fits the identifiable drivers better than the structural one does. That read is provisional by design, because the metric that would falsify it lags the contracts. Three checkable items would move it:

Wallet retention by channel, outside the US federal book: if GTS and GBS wallet retention stay below 100% after the federal comparison laps in 2026, the cyclical explanation weakens and the structural one gains. The figures appear each quarter in the earnings supplement and annually in the 10-K.

The loss-reason log: management asserts AI-as-substitute is rarely cited and falling. A rising share of losses attributed to AI, disclosed on a call, would be the clearest single signal the thesis has turned.

AskGartner retention uplift across a full renewal cycle: a durable, repeated gap between user and non-user renewal rates would confirm the tool as a defensive moat rather than a cross-tab artifact.

Until at least the first of these resolves, the structural half of the case stays genuinely open — which is why the report treats the 2025 stall as an unfinished question rather than a settled one.