In a popular Chinese response to Perplexity, news media, Wikipedia, and branders are almost half-packaged, while branding their own network is often the smallest of these three. This result is counterintuitive to most marketing teams: you place the budget on the official web content without running a path that is quoted in the news and encyclopedia, and the visibility of the AI response is systematically underestimated.
How does this investigation work?
In the first half of 2026, we sent to Perplexity, mainly on the basis of questions frequently asked by Taiwan's B2B and SaaS readers, a series of complex Chinese-language queries, with four types of intentions: product comparison, terminology interpretation, choice advice and industry trends. Each of these answers is based on its actual reference links, and each domain is included in seven barrels of the "News Media" "Wiki/encyclopaedia" "brand website" "The Forum and Society" "Government and Science" "The Electrician and Prices" "The Content Farm and Aggregation Station" to calculate the proportion of each category quoted. This is Tenten's own sample view, not a web-based survey, and the sample will shift as the subject is asked. Please consider it a sense of direction rather than a precise city.
Three main pieces: news, wiki, brand office network.
Spread seven buckets and quote highly concentrated in the top three. In terms of the average of our samples, it's roughly this way.
- The press accounts for about 38%: it includes financial, technological and industrial media, and it is the single largest source.
- Wikipedia/encyclopaedia accounts for about 17%: Wikipedia dominates, followed by specialized dictionaries and encyclopedia sites.
- The brander network accounts for about 14%: the company's own product pages, blogs and documentation.
- The remaining 31%: The forum community, government and academia, the electricity market, and the whole stock of content farms are not more than 10%.
In other words, the first three pieces add up close to 70%. You want to be quoted by Perplexity, the battleground is really concentrated, not "all over the place" but "news, encyclopedias, official networks," which are all lines of communication.
Why did the press eat the biggest piece?
There are three practical reasons why news is the highest. First, the media has a high frequency and a high time-effectiveness, and when responding to a question about trends and recent events, the natural preference is a new, cross-checkable source. Second, the mainstream media has a high level of power and clean structure, and machines can easily extract paragraphs and numbers. Thirdly, the same thing is often reported in multiple media outlets, creating cross-references, and the probability of a model citing one or two of them is naturally amplified. For brands, this means one thing: a product that has been reported by three financial or technological media outlets, the potential for exposure in AI's responses, may be better than 10 unconnected blogs on official websites.
Invisibility of Wikipedia
Wiki ratio is often underestimated. It's not like the news is on the rise, but once the problem involves definition, history, rules or character background, Wikipedia is almost a default quote. The inspiration for the B2B brand is not "go make your own wiki" , most companies are unable to reach the register, and hard copy is being deleted. The real story is: to make sure that the wikis that are relevant to your type are accurate and that the external links cited point to reliable first-hand data. When a model is used to read Wikis and then follow the references, your official web or white paper will be brought into the citation chain.
Brand office network: Not high, but it's the only thing you can control.
The Internet only accounts for about 14%, which sounds leaking, but in a different sense: journalism depends on media relations, Wikipedia depends on community consensus, and official networks are the only place where you can decide 100% on content, structure and updating the rhythm. Our observation is whether the official network can be quoted, not in quantitative terms, but in "extractability". The same product code, which is difficult to refer to in a sales essay, is written in clear answers, tables or defined sentences with clear numbers, and is clearly more frequently cited.
When actually implemented, these are the things that deserve to be addressed first: putting the core conclusions of each page ahead of the page, using the real questions asked by the reader as a title, adding verifiable numbers and dates to the key facts, and making the FAQPage structure of data often questionable. These adjustments don't change your brand tone, but they decide whether the machine can clean up your sentence and put it in the answer.
The network of officials is small and does not matter; it represents you in this competition, the only one that can be perfected without asking anyone else.— Tenten GEO content team
How do we translate this data into action?
If you only do one thing, find out what's missing in each of these three pieces of brands: The news piece, haven't you been mentioned by third-party media for six months? This wiki, are you in the source of your class quote? This part of the official web, is your focus page still in the pronunciation? Each of the three questions corresponds to a path of repair, and the resources should go to the one with the largest shortfall.
I want to know how many times I've been quoted in the answer, where the gap is, where we can track the actual quotes with Brand Radar, or make a direct 30-minute GEO diagnosis, and we'll get you. I'm going to run it, show you the news, Wikipedia, and the online news, and talk about whether to do it.
