Community content is about half of AI citations in every industry studied

Trendos, an AI search visibility vendor, said in a sponsored post published by Search Engine Journal on 27 August 2026 that it examined 107 million AI answers this year and found that community and user-generated content accounts for close to half of the leading citation sources in every one of the three industries it studied. In the same data set, brand-owned pages account for 2% of the top 10 cited domains in IT and solutions services.

One thing has to be stated before anything else: that Search Engine Journal article is a sponsored post, written by Trendos, the vendor whose own platform produced the figures. The numbers have not been independently audited by a third party. That is a material limitation and it stays in view for the rest of this piece. It is not a reason to throw the data out. It is a reason to treat the percentages as one vendor's reading of its own dataset rather than as an industry census, and to check the pattern against your own measurements before you move budget. The original is on Search Engine Journal.

What Trendos says it measured

Trendos says it looked at 107 million AI answers during 2026 and pulled the top 10 citation sources for three industries: IT and solutions services, consumer goods, and communication services. It did this across four engines, namely ChatGPT, Perplexity, Gemini and Google AI Overviews. Every cited domain was then sorted into one of three buckets.

• Community and user-generated content: Reddit, YouTube, Quora, forums and social platforms.
• Brand, retail and owned: first-party sites and marketplace listings such as Amazon, plus price comparison resources.
• Independent editorial and reference: press, Wikipedia and B2B directories.

The reported shares are the average share of each engine's top 10 citation sources, weighted equally across the four engines. Trendos notes the rows are rounded and may not total 100. That is the extent of the published method. There is no sampling description, no prompt list, and no per-engine table beyond the industry averages, which is worth holding on to when you decide how much weight to give any single percentage.

The shares, industry by industry

These are the three rows Trendos published in its sponsored post on Search Engine Journal on 27 August 2026, expressed as the share of each engine's top 10 citation sources, averaged with equal weight across ChatGPT, Perplexity, Gemini and Google AI Overviews.

Read the rows across rather than down. The community column barely moves. The other two columns swap places almost completely between IT services and consumer goods, which is the finding that actually changes what a marketing team does with its week.

Reading the 2% figure without overreading it

The 2% for brand, retail and owned sources in IT and solutions services is the sharpest number Trendos published, and it is the one most likely to be repeated out of context. It needs handling with care, because of what it is a share of.

It is the share of the top 10 cited domains per engine, averaged across four engines. That is a ranking of the most frequently cited domains, not a measure of whether a given vendor's own website ever gets cited at all. A domain can be cited thousands of times in AI answers about IT services and still sit outside the top 10 most-cited domains, because those top slots are occupied by a handful of very large aggregators that appear across an enormous number of queries. One company's site cannot realistically out-frequency Reddit or G2 across an industry's whole question space, and it does not need to in order to be cited in the specific answers where a buyer is asking about that company's category.

So the honest statement of the finding is this: in IT services, the domains that engines lean on most often are almost entirely third-party ones, and any individual vendor site is a small, dispersed part of the tail. That is not the same claim as "your website does not matter". Your own pages still carry the product detail, pricing logic and specification language that a directory listing or a review paraphrases, and they are what an engine reaches for when a prompt names your brand directly. Trendos does not publish branded-prompt data, so nothing in the study speaks to that case either way.

Where the 2% does bite is in planning. If a team's entire plan for getting recommended by AI engines is to publish more pages on its own domain, the data says most of the citation surface in IT services sits somewhere that team is not working. That argues for a split of effort, not for abandoning owned content. Practical work on owned assets belongs under AI SEO and content marketing, and the off-domain half is where generative engine optimisation earns its keep.

Community and UGC sits at about half in every industry

The second headline is quieter but applies more broadly. Community and user-generated content came in at 51%, 50% and 50% across the three industries. Trendos frames this as the constant in the data: whatever the category, public conversation between people makes up about half of the leading citation sources.

The uncomfortable implication for a brand is that the largest single block of citation surface is one it cannot publish into, edit, or take down. A Reddit thread from three years ago, a YouTube teardown by someone who bought the product, a forum reply that got upvoted, all of these are eligible sources and none of them are under brand control. Participation is possible. Ownership is not.

Trendos does not publish any figure for how often those community sources are positive, negative or neutral about a named brand, and it does not break the community bucket down by platform. So the correct reading is about volume of citation surface, not about sentiment or about which platform to prioritise. If you want to know whether Reddit or YouTube matters more for your category, that is a question you have to answer with your own prompt testing, because this study does not answer it.

What the study recommends for IT and software brands

Trendos says much of the 47% editorial share in IT services is B2B review directories, and names Clutch, G2, GoodFirms, Gartner, SourceForge and Slashdot. Its advice for that industry has four parts, and all four sit on other people's platforms.

Claim and complete the directory profiles

Claim listings on the two or three directories buyers in your specific category actually shortlist from, rather than defaulting to the largest name, and fill every field: categories, descriptions, features, integrations, pricing and screenshots. Trendos argues half-empty profiles get skipped by buyers and engines alike, and that category and feature wording should match the terms prospects use when they ask an engine a question.

Earn reviews steadily rather than in one batch

Ask recent, satisfied customers for reviews on the directories that actually surface in AI answers, and aim for recency and steady volume together. Trendos is explicit that a regular trickle of current reviews beats one large batch that then goes stale. This is the part most likely to be done badly, because a review drive is easy to run once and hard to run every month.

Reply to reviews, including the critical ones

Trendos frames replies as a content act rather than a customer service act: a response adds fresh, quotable text to a page that engines cite directly. Replying only to the positive reviews leaves the critical ones as the last word on the page, in text the engine can quote.

Pitch trade media something citable

The advice on earned coverage is to list the trade outlets that already rank for your category, work that short list, and pitch original data, a genuine product comparison or an expert take rather than a press release. Trendos also makes a small but useful point: the piece has to name the product clearly, because the exact product name string is what an engine repeats.

Communication services runs on Reddit and Wikipedia

For communication services, Trendos reports community and UGC at 50%, independent editorial and reference at 40%, and brand, retail and owned at 10%. Within that, it says Reddit is the top source in three of the four engines, and Wikipedia leads one engine's entire list.

The specific advice that follows is narrow and concrete: maintain an accurate, well-sourced Wikipedia entry, because several engines quote it directly. That is unusual as marketing advice because it is not a marketing channel, it is a factual record with its own editorial rules that a brand does not control and should not attempt to control. What a brand can do is make sure the sourcing that a Wikipedia entry depends on exists in the first place, which loops back to press coverage.

What the study does not say

This part matters as much as the percentages, because a vendor-authored study invites people to fill gaps with assumptions.

• Nothing about Thailand, and nothing about any Asian market. The three industry rows are global averages with no geographic split published.
• Nothing about which specific brands were cited or which won share. Domains are grouped into buckets, not ranked by brand.
• No methodology detail on how the 107 million answers were sampled, how prompts were selected, or over what exact window within 2026.
• No time series. There is no before-and-after, so the study cannot tell you whether any of these shares are rising or falling.
• No per-engine breakdown beyond the industry averages, other than the qualitative note that Reddit leads three of four engines in communication services.
• No sentiment data, no volume data, and no link between citation share and any commercial outcome such as traffic, pipeline or revenue.

What this means for Thai marketers

Since the study publishes no Thailand data, the useful move is to treat the three rows as a hypothesis about where citation surface concentrates in your category, then test it locally rather than adopt it.

Two structural differences are worth thinking about for a Thai brand. First, the community bucket in this study is described using Reddit, YouTube, Quora, forums and social platforms, and the balance between those platforms in Thai-language conversation is not the same as the English-language balance the global average is dominated by. Whether Thai-language answers in your category lean on the same platforms is a testable question and not one the study answers. Second, the B2B review directories named are all international ones, which means a Thai B2B seller working the Thai market has to check whether those directories carry meaningful listings and reviews for its category in this market at all before pouring effort into them.

There is no public benchmark for a "good" citation mix in Thailand, by industry or otherwise. Nobody publishes one, and any number quoted as a Thai benchmark should be treated as invented until the method behind it is shown. The workable substitute is your own baseline, measured on your own prompts, which is the same discipline that SEO in Thailand has always required for rankings. On the directory question specifically, the slower path is the more accountable one: earning reviews on a third-party profile takes months and cannot be faked at scale, whereas publishing another twenty pages on your own domain can be done in a fortnight and, if this data is directionally right, addresses a small share of where IT services citations actually come from.

How to measure your own citation mix

An industry average is a poor substitute for your own numbers, and building your own is not complicated. The point of the exercise is to produce a citation mix you can defend, in your language and your category.

1. Write a fixed list of 30 to 50 prompts a real buyer in your category would type, split between unbranded questions and questions that name your brand. Freeze the list, because the value comes from repeating it unchanged.
2. Run the list against the engines your buyers actually use, and record every cited domain for every answer, not just the ones you like.
3. Sort each domain into the same three buckets Trendos used, so your numbers are comparable to the published rows.
4. Record whether your own brand is named in the answer text, which is a separate question from whether your domain is cited. Both are worth tracking and they move independently.
5. Repeat monthly on the same prompt list. Your own trend line is more useful than any single industry average, because it reflects your category, your language and the engines your buyers use.

If ChatGPT is the engine that matters most to your buyers, the same measurement discipline applies to ChatGPT SEO work: fix the prompt list, record the citations, and let the data rather than the anecdote decide where the next quarter goes.

FAQ: questions this data raises

Does the 2% figure mean a software vendor's website is not worth investing in?

No, and the study does not claim that. The 2% is the share of the top 10 most-cited domains per engine in IT and solutions services, so it describes which domains dominate the citation lists, not whether an individual vendor site is ever cited. Owned pages still carry the product detail that third parties paraphrase, and Trendos publishes no data on branded prompts where a vendor's own site is the obvious source.

Can the figures be trusted given the article is sponsored?

They should be treated as one vendor's reading of its own unaudited data, which is weaker evidence than an independent study but stronger than nothing. Trendos both produced the dataset and sells tools in the category the data describes, and Search Engine Journal labelled the piece as sponsored. Use the rows as a hypothesis to test on your own prompts rather than as a settled fact.

Does the study say anything about Thailand or Southeast Asia?

No. The published figures are industry averages across four engines with no geographic breakdown at all, so there is nothing in the study about Thailand, Southeast Asia or any other region. Any Thailand-specific claim attributed to this data is an extrapolation, not a finding.

Which engines were covered?

ChatGPT, Perplexity, Gemini and Google AI Overviews, weighted equally in the averages. Trendos does not publish separate tables per engine, apart from noting that Reddit is the top source in three of the four engines for communication services and that Wikipedia leads one engine's entire list in that industry.

What should a B2B brand do first if it takes this seriously?

Claim and fully complete profiles on the two or three B2B review directories its buyers actually shortlist from, then build a monthly rhythm of earning and replying to reviews. Trendos names Clutch, G2, GoodFirms, Gartner, SourceForge and Slashdot as the directories that make up much of the editorial share in IT services, and its advice is to pick by category fit rather than by brand size.

If you want to know where AI answers in your category actually pull from, the answer starts with a fixed prompt list and a month of records, not with an industry average from a vendor deck. Relevant Audience can help set that measurement up and act on what it shows.