What AI models recommend when an IT buyer asks.

A monthly record of the products twelve AI models name when a buyer asks for a SIEM, a CI/CD platform, a managed database, an identity provider or any of the IT categories in the index. Each question is asked six ways on behalf of three company sizes, every answer is judged by one fixed method, and the standings are published with the raw record behind them.

StatusThe first edition is being collected. Fifteen categories have been through two collection runs; the edition publishes when the launch set is collected and the calibration repeat has set the noise floor. Nothing is ranked by hand and no placement is for sale.

models, one tier

Each lab's low-cost model, the ones a buyer meets first, asked the same questions on the same day.

framings per question

Direct, paraphrase, comparative, budget, scale and negative, so one ask is never the answer.

buyers

A 30 person company, a 500 person company and a 5,000 person enterprise with SSO, SOC 2 and procurement review.

The method is published

The index uses the method of the GTM AI Recommendation Index, from the same publisher: a judge reads every answer and writes one label per product named, a calibration repeat measures how much a standing moves with nothing changed, and only a move larger than that noise floor is reported as movement. The full method, the judge rubric and the second-judge check are on the method page.

What the index does not do

It does not recommend a product, it does not take payment for placement, and it does not claim the models can be steered. It reports what the models said, with the record to check it against.

When the first edition publishes

The release note goes to readers of the GTM index on the day. Ask for it there, and the IT edition arrives on the same list.

Get the release note