The client

A top-five NEV brand in China by volume—strong in classic search and social. In an early-2026 review, marketing found an awkward gap: when users asked AI “what NEV to buy on a 300k budget,” they almost never appeared.

Diagnosis

We ran 240 core purchase intents in DZOS. Three findings:

  • Mention rate 34%, first-recommend rate only 12%. When they were named, it was usually a supporting slot.
  • AI cited forums and creators, not official pages. The site was mostly images and video. Text density was too low for a citable fact.
  • Specs disagreed across channels. Range for the same model had three versions—site, dealer page, and press. The model lowered confidence in the brand.

The plan: three tracks at once

Track one: one fact base

With the product team we built a standard fact list for every model on sale—one official number per spec, with test method and conditions. That list then corrected the site, encyclopedia entries, dealer pages, and the press kit.

Track two: rebuild on-site structure

Visual model pages got a structured text layer: spec tables, FAQ, use-case notes, and a fair comparison with same-price rivals. Each heading answers one concrete question.

Track three: long-tail decision questions

From “AI cannot answer this” items in monitoring, we published 180 pieces on real purchase worries—“winter range discount,” “what battery warranty actually covers,” “what you need to install a charger.” Those are the questions people ask AI while deciding.

Results

At the month-four review:

  • On core purchase intents, first-recommend rate rose from 12% to 68%
  • Overall mention rate rose from 34% to 91%
  • Among AI citations, official sources rose from 6% to 54%
  • Organic site traffic from AI search entries grew 340% month on month

What transfers

The lesson worth keeping: the gap is often not “no marketing,” but “marketing AI cannot read.” This client already spent heavily on content—it was built for people to look at, not for a machine to parse. Once the fact layer was there, existing brand equity finally showed up in answers.