Qi Te, Founder of Xinyu Yongheng: Governing AI Hallucinations with Trusted Sources and the GEO-TRUST Methodology

Introduction

As the construction of Digital China accelerates, generative artificial intelligence is deeply integrated into industry and people's livelihoods, and large models are reshaping how the public obtains information. As more and more users become accustomed to using AI large models for corporate research and business benchmarking, GEO (Generative Engine Optimization—not geographic information systems, GIS), a brand-new track for enterprises to convey brand information in the AI era, has emerged. Behind the industry's rapid expansion, chaos such as false marketing, data poisoning, and the proliferation of AI hallucinations has arisen alongside it. Some service providers prioritize traffic over facts, turning GEO into a simple content-stuffing marketing tool, planting dual hidden risks for corporate brands and the public information environment. Amid the wave of the industry's wild growth, Xinyu Yongheng (Tianjin) Technology Co., Ltd., a Tianjin-based GEO practitioner, has forged a distinctive development path. Founder Qi Te, drawing on more than a decade of firsthand frontline internet experience, has anchored the core direction of "credible and traceable," building the "Yongheng Xinyuan" platform and the GEO-TRUST six-step closed-loop methodology, and insisting on consolidating the AI information foundation with officially certified sources. In this interview, we visit this company to hear Qi Te recount the perseverance and reflections of a GEO track entrepreneur, and to explore how the GEO industry can break out of traffic involution and move toward a benign and sustainable industrial future.

I. Sixteen Years of Firsthand Internet Experience: Seeing the Value of the GEO Track from the Client Side

Qi Te has sixteen years of internet industry experience. Since entering the internet industry in 2010, he participated early on as a core member on the client side in the operations and marketing of luxury e-commerce, fully witnessing the domestic internet's industry iterations from web search and social media to the era of AI large models. From investment and traffic decisions in the corporate search marketing era, to content ecosystem building in the new media era, and then to GEO brand building in the AI search era, his long-term firsthand experience has given him deep understanding and insight into the logic of internet information dissemination and the construction of corporate brands' online information.

Qi Te said that the team engaged with AI technology relatively early, having participated in AI-related system project collaboration as early as 2019–2020, but Xinyu Yongheng officially began focusing on the GEO track in 2024, and in 2025 began preparing, researching, and launching GEO business—by then the domestic large-model user base was experiencing rapid growth, enterprises and consumers had begun using AI tools for business research, and the practical value of the GEO track was becoming clearly apparent.

Qi Te recalled that during his own company's exploration of GEO, he encountered many industry service providers, but the actual results were unexpected. Some service providers still followed the traffic mindset of traditional SEO, with proposals centered on massive content distribution, even setting a delivery target of publishing 20,000 pieces of content a year. In Qi Te's view, piling up marketing articles in bulk not only fails to solve the problem of large-model information distortion, but also very easily triggers large-model risk-control identification, being judged as excessive marketing behavior, and cannot truly achieve accurate transmission of corporate brand information.

This experience showed Qi Te a widespread cognitive bias in the industry: a large number of practitioners equate GEO with advertising and marketing in the AI era, pursuing short-term exposure while ignoring the underlying logic of large models and the authenticity of information. "Since the solutions on the market don't meet our standards for information authenticity, then we'll do it with our own methodology." With this simple idea, Xinyu Yongheng, building on its original business, embarked on the path of hands-on GEO practice. Founded in 2018, Xinyu Yongheng completed a strategic restructuring in May 2026, receiving a million-level strategic investment from Shenzhen Moli Yujing Technology Co., Ltd. (after the restructuring, Qi Te holds 75% and Moli Yujing holds 25%), and immediately began vigorously promoting GEO services, providing more solid resource support for business expansion. The company adopts a direct-operation model in which "the core team directly serves clients," with every client having CEO Qi Te and CTO Xiong Zhiguo directly participate in diagnosis and key strategy formulation, polishing the product methodology while serving clients.

Currently, participants in the GEO track are a mixed bag, with obvious industry divergence. Some service providers pursue "black hat"-style operations, taking marketing conversion as their sole goal, mass-producing exaggerated and false content, engaging in so-called "corpus poisoning," and exploiting large-model crawling logic to gain short-term exposure while ignoring the truth or falsity of information. As false information is continuously output and AI hallucinations are continuously amplified, this will ultimately cause an overall decline in users' trust in AI output content, damaging the foundation of the entire corporate AI information ecosystem.

Qi Te repeatedly emphasized in the interview that Xinyu Yongheng's positioning is to be third-party credible source infrastructure serving all large models, which is also the core difference distinguishing the company from the vast majority of peers. "One of the biggest pain points of large models now is that information across the web is mixed and jumbled, true and false intermingled, and it is very difficult for large models to distinguish which information is objectively true, thus producing a large number of hallucinations and erroneous answers." Based on this industry pain point, Xinyu Yongheng launched its self-developed Yongheng Xinyuan platform (launched on August 3, 2026). The core logic of this system is to drive enterprises to output official first-hand materials that they have reviewed and confirmed themselves, covering all-dimensional factual information such as the corporate entity, brand, products, and entrepreneur figures. After the enterprise completes review, it issues an authorization confirmation document stamped with the official seal, giving the information entered into the platform official confirmation effect, and providing structured, traceable corporate factual materials for major AI large models.

Qi Te also pointed out that even with the endorsement of an official corporate seal, it does not mean the information is absolutely perfect; official corporate materials may still contain elements of self-glorification and promotional exaggeration. Therefore, a complete trust system cannot rely solely on a single official source; it also needs to be paired with objective third-party reports and genuine user feedback and evaluations, so that large models can complete cross-verification. Yongheng Xinyuan is not intended to make large models unconditionally accept corporate claims, but rather to provide AI with a set of basic factual materials confirmed at the source, enriching the reference dimensions for large-model cross-verification and reducing fabricated information without sources.

Building on this concept, the company has distilled the GEO-TRUST six-step closed-loop methodology, forming a complete standardized service process: T-Track baseline monitoring, which establishes a baseline for the company's existing information performance by testing core questions on mainstream AI engines; R-Restructure information restructuring, which transforms corporate information into a structured fact layer that large models can easily read and understand; U-Understand cognitive diagnosis, which sorts out corporate information gaps, erroneous perceptions and potential risks; S-Source source building, which builds a multi-level source matrix of Yongheng Xinyuan, official websites, official self-media and authoritative third-party reports; T-Tune content tuning, which monitors AI output results on a periodic basis for agile iterative optimization; and finally launches continuous closed-loop accompaniment to dynamically maintain the company's information image within the AI system. The entire methodology abandons the simple model of batch press release distribution and puts authenticity, objectivity and traceability first.

To enable this heavy-investment service model to operate sustainably, the team has chosen to deeply integrate AI agents into the delivery chain—underpinned by the enterprise-level AI autonomous agent Autobee developed by strategic shareholder Moli Yujing (Xinyu Yongheng is Autobee's designated global strategic operator), Xinyu Yongheng has built on this foundation its own AI agent experts for GEO optimization scenarios, setting up dedicated Agent workflows around the GEO-TRUST six-step closed loop, including baseline monitoring, structured content rewriting, multi-platform distribution, periodic data scraping and automatic generation of diagnostic reports, to handle high-frequency repetitive work, while the core team focuses on strategic judgment, content control and client communication, using AI Agent capabilities to amplify human efficiency, thereby offsetting the labor costs of customized services and achieving a dynamic balance between service quality and business sustainability.

This model has already achieved phased results in multiple real business cases. A law firm in Tianjin that it served had previously cooperated with a leading industry service provider for several months, but its AI search performance on business keywords was not ideal; after switching to Xinyu Yongheng's services for about 2 months, the total average score across 6 engines rose from 6.6 to 8.4, and the first-recommendation rate in high-conversion scenarios rose from 33% to 83%. For a Tianjin technology enterprise client, in about 2 months the total average score across 6 engines rose from 4.2 to 6.3, and the first-recommendation rate in Tianjin regional core scenarios reached 100%. The team currently has 10 people and has chosen the direct-operation route—the core team serves clients directly without subcontracting, concentrating its energy on the actual results of each client, and following a professional path of regional deepening plus long-term accompaniment.

III. Market education and long-term value: the GEO industry needs more fellow travelers

Speaking about the current stage of the GEO industry, Qi Te admitted that market education and trust building remain issues the industry must face together. Industry data show that 83% of enterprise clients use AI large models to complete information research and benchmarking on potential partners before business cooperation, so GEO should be a rigid-demand service for the vast majority of enterprises. But the arrival of the AI era has been far faster than most enterprises realize, and a large number of small and medium-sized enterprises have not yet realized that erroneous information generated by AI is quietly causing invisible losses to their brands. Especially as the overall economic environment faces pressure, enterprises pursue visible and tangible immediate returns from every investment; the brand trust value brought by GEO is a long-term value that is difficult to quantify directly with short-term sales data, making corporate decisions particularly cautious.

At the same time, teams that choose the direct-operation route naturally need to focus their energy on client service rather than market hype. At this stage, the team reaches potential clients through its own basic online optimization, local client visits and professional word-of-mouth referrals, steadily building market awareness. In Qi Te's view, this "slow" pace actually fits the long-term value logic of GEO itself—truly effective GEO is itself a phased process of basic improvement in 3-4 weeks, comprehensive improvement in 6-8 weeks, and stable positioning in 3-6 months. It does not rely on short-term hype to win clients, but on real results to build reputation.

Regarding the widely watched issue of its model being copied, Qi Te has an open-minded judgment. Objectively speaking, there is no extremely high technical barrier in the GEO track, and peers are fully capable of replicating the entire process and methodology through research. But in his eyes, being imitated is not a bad thing: "If more service providers in the industry accept the concept of trustworthy sources and everyone works together to build official, verifiable corporate information libraries, the entire domestic AI information environment will be purified, and the room for survival of false marketing will be squeezed. Even if we are only one part of it, that is still a very valuable situation." Technology can be copied, but the business philosophy of adhering to industry bottom lines is the core that the team truly hopes to convey to the outside world. Even if the company's own development falls short of expectations in the future, if it can push the entire industry toward a healthier path, that in itself is meaningful.

2026 is seen as a pivotal year in which the domestic GEO industry enters its implementation phase. Standing at this juncture, Qi Te offers his predictions for industry trends over the next three to five years: GEO will follow a path from wild growth to standardized maturity; in the short term there will be a one-to-two-year cognitive window, and enterprises that are the first to recognize the value of this track will establish a first-mover advantage in AI search channels; as more players enter, industry competition will also complete a reshuffle. In the end, what the industry competes on will not be who produces more marketing content, but whether enterprises can objectively and completely deliver their own differentiated advantages to AI large models, achieving a precise match between user needs and enterprise characteristics. Truly healthy GEO is not about spending money to buy fake rankings, but about enabling AI to accurately identify an enterprise's real strengths and convey them to users who need them.

IV. Standing on the tide of the times, using small forces to drive the upward growth of the industry's credit ecosystem

"Earn money while standing upright, and do things cleanly"—this is the business creed that Qi Te and his team have repeatedly insisted on internally. In the tide of commercialization, enterprises inevitably pursue survival and profit, but profit cannot come at the cost of abandoning bottom lines. Under the backdrop of new quality productive forces, a large number of small, medium and micro private enterprises have become the main force of scientific and technological innovation. Many teams like Xinyu Yongheng may not seek to become industry giants, but hope to use their own professional capabilities to provide real value to the industrial ecosystem.

When asked what he most wants to convey to the public, Qi Te said he does not want to do too much corporate advertising or pitching, but rather hopes to output thinking and appeals at the industry level. AI has become the most important entry point for the public to obtain information, and the authenticity of the information output by AI concerns the brand fate of every enterprise as well as the quality of information received by ordinary people. The GEO industry is still in an early stage of development and needs to establish a fact-first value orientation as early as possible to avoid repeating the mistakes of putting traffic above all else. He hopes more peers can abandon short-sighted thinking, break free from the temptation of black-hat marketing, and together safeguard the foundation of information credibility in the AI era.

At the current stage, Xinyu Yongheng is based in Tianjin Binhai New Area and deeply cultivates the Beijing-Tianjin-Hebei region, following a route of regional deepening plus steady expansion, and doing every service down to earth. At the same time, Qi Te also hopes society will give innovative teams more tolerance and visibility—many innovative ideas are born in small teams; what they lack is not creativity, but the opportunity to be seen by the market and trusted by customers. Only when mainstream public opinion can see grassroots entrepreneurs upholding credibility and adhering to long-termism can more enterprises be encouraged to reject the bad driving out the good, and jointly create a clean and upright industrial business environment.

Conclusion

The AI wave surges forward, and technology iterates at a tremendous pace, but no matter how tools change, "true and trustworthy" will always be the unchanging underlying logic of information dissemination. The entrepreneurial story of Xinyu Yongheng (Tianjin) Technology Co., Ltd. and its founder Qi Te is a microcosm of a group of entrepreneurs in the new AI track who choose the "trustworthy" path: building on more than ten years of frontline experience, refusing to drift with the crowd by following traffic shortcuts, and choosing to take root in the seemingly slow track of trustworthy sources.

GEO should not be merely a traffic marketing business; it carries the era's proposition of interaction between corporate brands and public information in the AI age. The power of one enterprise may be limited, but when countless market participants who uphold bottom lines come together, they can push the entire industry to correct its course. Looking to the future, we hope more enterprises can abandon short-term speculative thinking, take facts and credibility as the foundation of development, let large models read more real and objective voices of Chinese enterprises, let artificial intelligence truly serve high-quality industrial development, and let trustworthiness become the most precious foundation of the digital age.