How to pick a GEO vendor: Comparing Greater China AI search consultants, plus 8 questions to ask

Six months ago, a brand CMO brought us a quarterly report with solid numbers: stable search rankings, SEO scores over 90. That same quarter, when people asked ChatGPT or Perplexity "who should I recommend in Taiwan's XX industry," his brand never showed up. Not once. A competitor did, three times.

He was puzzled. Ranked first on Google, yet invisible to AI.

That's the gap between GEO, generative engine optimization, and traditional SEO. If you're hiring outside help to fill it, picking a GEO vendor takes more care than hiring an SEO firm. The market is too new. People selling keywords last month are advertising themselves as "AI search advisors" today, with the same underlying methods.

Three types of GEO vendors in Greater China

The teams we've worked with, and the ones clients see in pitch meetings, split into three categories. None is right or wrong, each solves different problems.

Most brands start with the first type. Pricing is clear, output is tangible. But there's a catch with GEO: being cited by AI isn't about volume. It's about whether your content can be understood by the model, trusted by it, and actually retrieved when it generates answers. These factors don't overlap nearly as much as people expect.

Bring these 8 questions to your pitch

Pitch meetings can be deceptive. Print out this list and work through each question, pressing vendors for specifics.

Question one: How do you measure success? If they're talking about rankings, traffic, and DA scores instead of "which AI engines cite your brand, on which queries, how many times," they're selling the same old wine in new bottles.

Question two: Which engines do you monitor? ChatGPT, Perplexity, Gemini, Google AI Overviews, plus Greater China models like Doubao, Kimi, and Ernie. Chinese and English models perform differently; so do domestic versus international ones. Watching just one means flying half-blind.

Question three: How do you handle structured data and crawlability? Schema, entity markup, clean parsing, that's the foundation. If they can't explain llms.txt, Article/FAQ schema, or entity disambiguation, they haven't done the technical work.

Question four: Is your content written for citation or for reading? AI favors sentences with clear definition, data, and stance. Ask them to show you an example of content actually designed for citation.

Question five: Who actually does the work? Is a senior consultant taking the engagement and outsourcing to writers, or do you get engineers and strategists in the room together? Quality differences usually show in this answer.

Question six: How do you attribute those citations back to your work? Attribution is messy in AI. Someone willing to say honestly, "we can't trace this 100 percent; here's how we get close", is more credible than someone guaranteeing results.

Question seven: How deep is your Greater China expertise? Simplified versus Traditional Chinese, Taiwan dialect and terminology, regulatory context, local authoritative sources. If your content is translated from English, models sense the non-native feel, and trust erodes.

Question eight: After the contract ends, who owns the capability? Almost nobody asks this, but it's essential. Do you want a system your team can operate going forward, or a black box that stops working the moment the relationship ends?

Why field deployment matters

I should be direct about Tenten's angle. The way we approach GEO is FDM, field deployment marketing, and it's different from the other two models. We don't send reports. We put people in your operation.

The reason is simple. GEO fails ninety percent of the time not because people don't know what to do. It fails because cross-functional execution gets stuck. You need engineering sprints to change schema. Legal needs to clear content changes. The business team owns industry data. A consultant sending emails can't push this forward. Someone embedded in your standups gets it done in two weeks.

It's easy to say in a demo that "you'll get cited by AI." The hard part is opening your analytics three months later and seeing your brand actually show up in AI-generated answers, with your team knowing how to maintain it afterward.

If you're evaluating vendors, finish all eight questions before you look at pricing. Notice whose answers are specific, who admits what they can't do. The vendor willing to discuss tradeoffs is usually the one who's actually done the work.