MarsBit argued that a wide range of sectors are now trying GEO, including gold, industrial equipment, marking machines, tourism, eyewear stores, education, hotels, cultural tourism, animal traceability, ERP software, villa gates, home building materials, and hats.
The article’s view is that roughly half of those industries are wasting money. Its reasoning is not that GEO itself does not work, but that many companies misunderstand what it is. In the piece, GEO is framed not as search optimization, but as brand building. That means a company’s brand profile, stage of development, and content capability matter a great deal. Not every business is a fit, and not every business needs to move now.
Three types of businesses the article says should hold off
Highly localized businesses
The first group is businesses with strong geographic limits. The article points to education and eyewear stores as examples. These are local businesses by nature, with customers often concentrated within a few kilometers and service areas limited to a city or even a district.
As the piece explains it, GEO works by getting AI systems to recommend a brand across the broader internet. But when a user asks which eyewear store is best in Beijing’s Chaoyang district, the likely source material for AI is local content from platforms such as Dianping, Meituan, and Xiaohongshu, not broad brand articles placed across the web.
The article uses Pangdonglai as another example. It says the retailer is strong, but remains tied to Xuchang as a regional brand. AI is unlikely to recommend that kind of regional retail brand to users nationwide. Even if some platforms have begun to support segmented recommendations by area, the overall applicability is still weak and the return on investment is not attractive.
Education and training faces another problem. The article says policy compliance in that sector already sits in a gray zone, and AI platforms are only likely to tighten content review. In that setting, publishing more content still does not make it easy to get cited.
Overly vertical niche categories
The second group is very narrow vertical categories, including marking machines, animal traceability, and villa gates. The article argues that the decision path in these industries does not run through AI chat in the first place.
Its example is villa gates. A buyer in that market is unlikely to ask Doubao which villa gate brand is best. The real decision scene, the article says, is more likely to be an offline building materials market, a designer recommendation, or an engineering channel. In that case, spending on GEO means paying for visibility in a channel where users were not searching to begin with.
The test offered in the article is simple: when target customers make a purchase decision, do they use AI conversation to compare and choose? If the answer is no, GEO is an ineffective spend.
Companies with no brand base that only want leads
The third group is companies with little brand foundation that want to use GEO as a direct lead-generation tool. The article says this is one of the most common misunderstandings in the market. Many small and mid-sized business owners hear that AI search is the next traffic entry point and immediately ask whether they can put money in the way they do on Douyin and pull out a batch of customers.
The article’s answer is no. It says GEO is not performance advertising. In the current model described in the piece, AI crawls information across the web and decides which brands are worth recommending, while recommendation slots are limited to three to five positions. If a company is not already present in user perception, AI will not create consensus on its behalf. It will cite consensus that already exists.
That leads to the article’s core challenge: if a company has no brand awareness, no accumulated content, and no industry reputation, why would AI recommend it?
Which businesses the article sees as better candidates
Legacy brands seeking a refresh
The first group the article sees as a good fit is older brands looking for renewal. It describes this as the clearest return case it has seen so far.
Its example is OSM. The brand has been around for more than 30 years, and the article says its product strength is not weak, with pearl whitening as its core consumer association. But in recent years it has been squeezed by newer consumer brands that are better at Douyin and short-video seeding, and those brands have taken away attention.
The article says AI recommends differently from short-video platforms. Instead of rewarding short-form distribution tactics, AI places weight on source authority, brand history, product strength, and user reputation taken together.
On that front, legacy brands hold an inherent advantage. The article notes that they often have three decades of media coverage, encyclopedia entries, industry reviews, and real user accumulation. The issue is that this information sits in scattered places and has not been structured for AI systems. GEO, in this framing, is about reorganizing, updating, and optimizing brand assets that already exist so AI can actually see the strength that was already there.
The piece says this kind of refresh tends to cost less and show results sooner because the company is not building a brand from zero. It is making an existing brand discoverable again.
New brands trying to build awareness
The second group is new brands that need to establish recognition. The article splits this into two cases.
One is a brand defining an entirely new category. It points to Lingyuzhou’s Xiaofangji, described as a new product form that ranks No. 1 on JD.com’s AI hardware sales list. In that situation, GEO helps a brand show up when users describe a need in natural language. The battle is not over recommendation slots in an existing category, but over the mental entry point for a new one.
The other case is a new brand with a real category-level breakthrough. The article uses Qingxian Intelligence in the ergonomic chair market as an example. There are already established players in that track, including SIHOO, but if a newcomer has obvious differentiation in materials, design, or target user positioning, GEO may help it carve out space in AI recommendations. The article adds a condition: the differentiation has to be real, not a simple white-label shell change.
Companies that can invest in content over time
The third group is companies with genuine long-term content capability. The article says GEO Index Future acknowledged in its own writing that AI-generated content does not survive beyond 15 days. Platforms refresh their source base regularly, and low-quality, repetitive, and obviously machine-generated material is quickly downgraded.
What keeps getting cited, in the article’s view, is content written by real people, with depth and distinct viewpoints. That is why it argues GEO is not a one-off budget line. It is a long-term content buildout. Only companies that already have content teams and sustained output are likely to stay in the game.
AI and data companies
The fourth group is AI and data companies. The article says its research found GEO results are generally strong in that segment.
The reason, it argues, is straightforward. These companies already produce technical documents that are naturally machine-readable. Markdown formatting, clear logical structure, and a high degree of organization make the material easy for AI to parse and easy to cite.
If a company is already publishing technical documentation, API documents, and developer tutorials, the article says it may already be doing GEO without realizing it.
Why the article still urges restraint now
Platform rules are changing fast
Even for companies that fit the profile, the article says spending should still be measured. The first reason is the speed of platform change.
It says Doubao adjusted source weighting just last month, and Qwen is still iterating on its citation logic. A tactic that works today may stop working after one platform update. A content matrix built over three months can be wiped out by a single algorithm change.
The article compares this to the SEO era, when every major Baidu update would knock out a batch of sites. In GEO, it says, the pace may be even faster because AI platforms iterate at a much higher speed than search engines.
Results are difficult to measure precisely
The second reason is measurement. The article describes GEO as still being in a brand-advertising phase: a company knows that half its budget is being wasted, but it does not know which half.
According to the piece, no GEO service provider can currently tell a client that a 100,000 yuan spend produced a certain number of precise conversions. What they can show is how many times a brand was mentioned by AI. The path from mention to transaction, though, remains broken. There is no CTR, no conversion rate, and no user traceability.
The article says GEO only becomes meaningfully iterative once it reaches a performance-advertising stage where the full path from AI recommendation to final purchase can be tracked precisely. Until then, GEO spending is still brand spending, suitable mainly for companies that have room in their brand budgets to experiment.
Industry disorder remains serious
The third reason is the state of the industry itself. The article notes that in July this year, Chaoyang district in Beijing issued what it described as the GEO industry’s first fine. The problems at the penalized company included falsified qualifications, falsified data, and falsified case studies.
The article says that is only the tip of the iceberg. Referring to an earlier piece, it argues that the certainty of service providers’ revenue is much higher than the certainty of clients’ outcomes. Providers can be sure they will collect payment. Clients cannot be sure what they will receive in return.
In a market where industry standards have not yet been established, the article says caution is not timidity. It is rational.
A real opportunity, but one with a threshold
The conclusion of the article is that GEO is a real opportunity, but not a universal one.
For companies without a brand base, without content capability, and without the willingness to invest over time, the article says there is no need to rush in. Its suggestion is to wait until platform rules are more stable, attribution systems are established, and the market has eliminated a share of fraudulent service providers. At that point, the cost may be lower and the certainty higher.
For legacy brands seeking a refresh, new brands looking to claim position, or businesses that already have a content gene, the article says it can make sense to start. But the premise matters: this should be treated as a brand budget, not a customer-acquisition budget. The metric is the change in brand mindshare over six months to one year, not the number of leads generated next month.
The article ends with a warning not to be carried away by the idea that AI search is the next traffic entry point. The traffic entry point may be real. That does not mean everyone should be standing in front of it.