Everyone knows that when pharma companies, or the life sciences industry as a whole, do GEO, they are obviously different from other industries such as e-commerce, education, technology, and FMCG because of regulatory requirements for online drug promotion, content attributes, and audience characteristics.
As a special industry with strong compliance requirements, high professionalism, and low tolerance for error, pharma has a clearly different GEO deployment logic from general industries. General-industry GEO focuses on traffic conversion and brand exposure, while pharma GEO focuses on authoritative transmission and precise reach under compliance requirements.
For pharma GEO, the core goal is not merely being mentioned. It is being mentioned accurately, cited traceably, and summarized compliantly. Under compliance requirements, pharma companies need to transmit authoritative information and build professional trust, so AI preferentially cites compliant pharma content when answering drug-related questions such as indications, dosage and administration, and adverse reactions, as well as disease diagnosis and treatment questions. This helps avoid chaotic and misleading online expressions, ensures target audiences such as physicians, pharmacists, and patients receive accurate information, and meets industry regulatory requirements. Error tolerance is extremely low.
This Is a Real Pharma GEO Project We Worked On
On the surface, because pharma regulation is strict and everyone cannot promote aggressively, it may seem fine if neither you nor I show up much.
However, when MeDomino recently worked on a GEO project for a pharma client, we found that the competitor brand was mentioned by AI far more often than our client's brand. That felt wrong. Wasn't everyone supposed to be constrained in promotion?
So of course we dug into the reasons behind the answers.
We started with the reference materials, looking at which content links AI used when forming its answer and why those links led AI to make such judgments. The result surprised us: the competitor had placed a large amount of content across a broad range of coverage. As a result, whenever we asked questions in AI, a high proportion of the cited materials came from content published by that competitor. No wonder AI trusted the competitor more.
We reached one conclusion: it is not that pharma companies cannot say anything; it depends on how they say it. The competitor had published a lot of content, but on closer inspection, it was all educational content written based on drug instructions and clinical research data, so it was not off-limits.
Then we analyzed citation channels and found that the competitor's content was visible across many channels, while our client's brand information appeared almost only on its own website, WeChat Official Account, and some public information from government agencies such as health commissions and drug regulators. It looked authoritative, but coverage was too narrow, and the brand's online share of voice was heavily squeezed by the competitor.
Optimization Thinking for Several Mainstream AI Large Models
After mastering this information, we compared answers and citation patterns for the brand's target questions across AI platforms and formed rough optimization thinking for several mainstream AI models:
1. DeepSeek
Statistics showed that nearly all top 10 cited materials in DeepSeek were authoritative platforms, such as PubMed, CNKI, MSD Manual, FDA, and Chinese medical information query platforms. In other words, these are not channels that companies can influence simply through online promotion. Perhaps for this reason, our client's brand appeared far more often than other competitors in DeepSeek answers and citations.
For DeepSeek optimization under this current situation, we believe the best approach is to respond to change with stability: no need to optimize it aggressively, just continue monitoring, detect changes in time, and act as needed.
2. Tencent Yuanbao
When people talk about Tencent Yuanbao, they often say it mainly cites Tencent ecosystem sources such as WeChat Official Accounts. But in our actual operation, that was not what we saw at all. The so-called platform ecosystem barrier had already been broken. For target questions about this client's brand in Tencent Yuanbao, the top 10 cited materials contained no WeChat Official Accounts or Tencent News platforms. Instead, they were mostly public health education websites.
These vertical health websites are not ordinary self-media. They have strict requirements for information publishers, such as medical background qualifications. These are channels pharma companies should prioritize for cooperation and expansion.
3. Doubao
Doubao is quite special, arguably the most distinctive AI platform. It is currently the only mainstream AI that can directly cite Douyin videos as reference sources as of February 2026. But it could not directly capture Douyin videos from the beginning. According to Doubao's own description:
Search enhancement phase, March 28, 2025
The new Doubao deep-thinking function, using a think-while-search mode, officially launched. It can call tools and search information multiple times based on reasoning, providing more comprehensive results. This function significantly improved Doubao's ability to retrieve and cite Douyin video content, enabling it to cite Douyin videos more precisely as references in answers.
Explicit video recommendation, August 2025
Doubao began automatically attaching short-video links from Douyin below answers. Users can click and directly enter a Douyin-like feed playback interface.
What does this mean? It means mainstream AI large models' information retrieval and citation capabilities are getting stronger every day, becoming more tolerant of content forms, and allowing information to flow more openly across channels.
It also means companies, especially pharma marketing, brand, and digital teams, must pay attention to multimodal content deployment + all-channel publication. Content needs to be deployed before AI model technology updates. You cannot wait until AI can crawl Douyin before registering a Douyin account and thinking about KOL cooperation. By then it is too late. Your competitor may already have hundreds of existing content assets. How will you compete?
4. Kimi, Qianwen, and others
Beyond the AI platforms singled out above, for other platforms such as Kimi, Qianwen, Baidu AI, and others, at least for this drug brand, we did not find obvious citation-channel characteristics. Overall, official media and authoritative institutions, health vertical platforms, and comprehensive media platforms such as Sohu, Sina, NetEase, and Toutiao each occupied a share.
This optimization logic is relatively simple: publish. Just produce content normally, distribute across channels, and cover as many platforms as possible. If there is budget to cooperate with paid media channels, even better.
Back to the Title: Why Doesn't AI Recommend Your Product?
At this point, the question of why AI does not recommend your product is more or less clear. To summarize, it is usually because:
- You have too little total content
- Your content covers too little key business information
- Your content formats are not rich enough
- Your content distribution channels are too few
- Your content distribution frequency is too low
- ...
Here comes the important point, friends. Do you think I said all this just to tell you DeepSeek does not need optimization, Douyin videos must be made, then cooperate with a few health vertical websites and publish some educational articles on Sohu and Sina, and GEO will be done?
Absolutely not.
I kept saying "this client's brand" above for a reason. After we worked with this one brand on GEO for a period of time, the client expanded the cooperation scope with us and added GEO monitoring for six brands.
The statistical results were surprising: without doing it, you would never know that the answer citation patterns of these six brands were completely different. Even for the same AI, the high-frequency top 10 citation channels were almost different for every brand.
Earlier I said DeepSeek's top 10 were all authoritative platforms, right? Under another brand, DeepSeek's high-frequency citations became Baidu Wenku, Sohu, and a hospital website.
So what I want to say is: do not over-trust so-called GEO tutorials online, including what I said above, which does not constitute tutorial advice. Some people say one AI platform prefers type A content and channel A; others say another platform prefers type B content and channel B. Let alone whether they have actually done GEO, industries, products, and business fields differ so much. There are not that many standard answers.
MeDomino can provide GEO services for clients because we first had practical experience, then distilled GEO optimization logic from that experience. When implementing pharma GEO projects, MeDomino does not start by imposing a so-called optimization standard. We first conduct data monitoring based on the specific needs of the brand team. Only after we have actual data do we provide targeted optimization suggestions, and even content production and distribution.
Pharma product lines are extremely rich. Each product covers different disease areas, patient groups, and physician departments, and promotion goals differ by product stage. Whether what you want AI to say is feasible at this stage must be considered comprehensively. Again, there is no standard answer. MeDomino provides personalized GEO solutions based on your business needs, with comprehensive continuous monitoring and analysis, timely discovery of AI platform rule changes, and full-speed optimization.
I sincerely suggest that pharma friends interested in GEO contact us for a conversation. Finding new opportunities in the AI era is rarely a bad thing.
------
We are MeDomino, a reliable service provider specializing in AI + digital intelligence precision marketing for life sciences.
