Does adding 'for enterprise' change who AI recommends?

A qualifier is the condition an AI answer attaches to a recommendation, such as "budget-friendly pricing" or "gdpr and ccpa compliance requirements". We tagged every recommended-brand observation into five qualifier classes — enterprise, small business, budget, compliance, and beginner — and compared, within each market, the most-recommended brand in a qualifier class against the most-recommended brand in the unqualified pool. A market-qualifier cell is one market crossed with one qualifier class, and it counts only when both pools have at least 20 observations and an untied winner.

In 457 of 516 market-qualifier cells, or 88.6%, the qualifier winner was a different brand from the unqualified winner. That is about 9 in 10. The overall AI winner of a market usually does not win the market's qualified scenarios.

A qualifier pool is much smaller than the unqualified pool, so some winner changes could come from small-sample noise alone. We measured that directly: for every cell we drew a random subsample of unqualified observations of the same size as the qualifier pool and checked how often the subsample's winner differed from the full unqualified winner.

The control changed the winner in 163 of 408 control cells, or 40.0% (408 rather than 516 because a random subsample can produce a tied winner, and tied cells are dropped on both sides of the study). The real qualifier pools changed the winner 88.6% of the time — 2.2 times the noise baseline. Qualifiers move winners in a directional way that random sampling does not.

Every qualifier class changed the winner in at least three-quarters of its cells. Budget led at 91.4% (160 of 175 cells), then enterprise at 89.0% (129 of 145), compliance at 88.2% (105 of 119), and beginner at 87.5% (28 of 32). Small business was lowest at 77.8% (35 of 45).

A plausible reading of the gap: many tracked markets already skew toward small-business buyers, so the unqualified winner often is the small-business pick, while the overall winner is rarely also the cheapest option — which would leave the budget cell open for a different brand.

In CRM software platforms, the unqualified winner is Pipedrive; under enterprise qualifiers the winner is Salesforce. The same pattern repeats across software markets: the overall winner is a mid-weight product, and the enterprise winner is the established enterprise platform.

The counts show how the evidence splits. In travel and expense management software, SAP Concur takes 57 of 116 enterprise-qualified recommendations in a market that Zoho Expense wins overall. Enterprise qualifiers in the evidence carry phrases such as "microsoft ecosystem integration" and "commission plan design for enterprise reps".

The small-business qualifier reverses the direction. In payroll and HRIS software, Rippling wins the market overall, but Gusto wins the small-business cell with 79 of 137 qualified recommendations. In CRM and sales pipeline software, the overall winner Pipedrive gives way to Less Annoying CRM.

Several of these pairs mirror the enterprise table: the same market has one overall winner, one enterprise winner, and one small-business winner. The class folds in startup and freelancer terms — which is why the venture-investor market appears in the table — and its observed phrasing includes "ease of use and affordable plans".

Budget cells do not hand the market to a cheaper plan of the overall winner; they hand it to a different brand positioned on price. Warby Parker wins online eyewear overall, but Zenni Optical wins the budget cell with 75 of 403 budget-qualified recommendations. TurboTax loses its budget cell to FreeTaxUSA, NordVPN to Surfshark, and HelloFresh to EveryPlate.

The observed budget phrasing is consistent: "budget-friendly pricing" and "free tier availability" are among the most frequent qualifier phrases in the class.

When the observed condition is regulatory — the most frequent phrase in the class is "gdpr and ccpa compliance requirements" — the winner moves to the brand positioned on compliance. Rippling wins payroll and HRIS software overall, but Deel takes the compliance cell with 76 of 428 qualified recommendations. PandaDoc loses the electronic-signature compliance cell to Docusign, and Polymarket loses the prediction-market compliance cell to Kalshi.

These are not obscure markets: the compliance class has 119 measurable cells, third most of the five classes.

The brand that wins a qualifier cell is rarely a stranger to the market. In 433 of 457 flipped cells, or 94.7%, the qualified winner already appears somewhere on the market's unqualified list; only 24 winners are absent from it entirely. But it sits low: the median qualified winner ranks 8th on the unqualified list, and 270 of 457, or 59.1%, sit below the unqualified top 5.

The displacement runs the other way too. In 218 of 457 flipped cells, or 47.7%, the overall winner does not appear in the qualified pool at all — the engine simply never recommends it for that scenario. Only 90 of 457 overall winners, or 19.7%, stay in the qualified top 3.

The change goes deeper than the #1 spot. Comparing each cell's top 5 brands with the market's unqualified top 5, the two lists share on average 1.6 of 5 brands. In 251 of 516 cells, or 48.6%, they share at most one brand, and in 89 cells, or 17.2%, they share none. Only 4 of 516 cells share all five.

A qualifier does not reorder the market's shortlist. It substitutes most of it.

Some brands win the same qualifier class in multiple markets. Deel wins the compliance cell in 4 markets, and Uniqlo wins the budget cell in 4. Zoho CRM, OneTrust, SAP Concur, Qualtrics, and Gusto each win a class in 3 markets.

That is qualifier ownership: the engine has attached a positioning to the brand that travels across markets. It is the evidence-level version of a segment leader.

Of 390,119 recommended-brand observations, 44,508, or 11.4%, carried a qualifier in one of the five classes. Budget was the largest class at 14,438 observations (3.7% of all recommendations), then compliance at 11,913 (3.0%), enterprise at 11,178 (2.9%), small business at 4,184 (1.1%), and beginner at 2,795 (0.72%).

Both engines attach these conditions at nearly the same rate: 11.6% of ChatGPT recommendations (10,844 of 93,455) and 11.3% of Google AI Overviews recommendations (33,664 of 296,664). The conditioning is not specific to either engine in this window.

The study window ends April 25, 2026 and covers ChatGPT and Google AI Overviews only, because qualifier-rich evidence extraction on the newer engine pair is too thin in our data to measure; restricting to one engine pair keeps the qualified and unqualified pools on the same engines and period. Including the thinner May-July 2026 evidence from ChatGPT Search and Google AI Mode, the winner-change rate is 80.8% (442 of 547 cells) — the finding is not an artifact of the window.

Qualifier classes are keyword rules over the extracted qualifier and need text; the small-business class includes startup and freelancer terms. Cells need at least 20 observations and an untied winner in both pools, and the 40.0% random-subsample control bounds how much of the change small samples alone could produce. Winner means most-recommended brand in the pool. The study is observational: it describes which brands engines recommended under which stated conditions, not why.

Your market's overall AI winner is not your competitor in segment-qualified questions. 88.6% of measurable qualifier cells have a different winner, the qualified winner typically ranks 8th overall, and half the time the overall winner is not recommended for the scenario at all.

Treat each qualifier cell you sell into as its own market. Check who AI recommends for your category with an enterprise, small-business, budget, or compliance condition attached, and make the segment positioning explicit on your own pages — the engines are already routing recommendations by these conditions, using the positioning language they can find. External segment evidence points the same way: review platforms publish separate leaderboards by company size because segment leaders differ, and AI answers reproduce that structure.