Your e-commerce store has well-crafted product pages, reviews, a site that works — but when someone asks ChatGPT where to buy a product in your sector, you never come up. The problem isn't SEO: it's that AI has learned to answer by reading thousands of real questions on forums and Quora, and your brand never appeared in any of those discussions. Whoever was in those threads gets recommended; whoever wasn't disappears, regardless of product quality. Building that presence is done systematically, starting from the questions your customers already ask every day.
Industry forums hold conversations that AI reads as “what real people say”. If your brand is absent there, you’re invisible to a key channel.
Let me give you a concrete example. Imagine an e-commerce store selling specialty food from Molise in Campobasso: caciocavallo from Agnone, white truffle from the Matese, extra virgin olive oil from the hills of Larino. Real, niche products with a story. The site is well cared for, the product pages are well written, there are Trustpilot reviews. And yet, when I ask ChatGPT or Perplexity “where to buy caciocavallo from Agnone online” or “best e-commerce stores for Molise truffle”, the brand doesn’t show up. Ever.
The reason isn’t the site. It’s that AIs, when answering niche queries, draw from threads where people have already discussed the same question. And those threads — on Quora, on food forums, on vertical subreddits — don’t talk about that brand.
Why AI treats Quora and forums as the “voice of real demand”
When an AI model has to answer a specific question (“which is the best maker of X?”), it needs signals about what real users think, not just what the brand says on its own site. A site is a self-interested statement; a forum thread is a third-party statement.
In the field of research on generative model behavior, question-and-answer structured forums have a characteristic that makes them particularly useful for training: every thread is an explicit pattern of the type “user’s question → reasoned answer”. It’s exactly the form in which AI produces output. In previous articles on backlinks as citation proxy and implicit reference weight I explained how AI weighs unlinked mentions and indirect citations. Quora and industry forums are the terrain where those mentions accumulate in the most useful possible form for the model.
It follows that an expert answer on Quora with a natural mention of your brand isn’t “a post on Quora”. It’s a question-answer pattern that the model can replicate when someone asks it the same question three months later.
The reverse engineering I did on specialty food forums
Because I wanted to understand where AI was pulling its answers from in niche food, I ran a reverse-engineering exercise on 12 queries like “where to buy aged caciocavallo from Agnone”, “best artisanal Molise olive oil”, “white truffle from the Matese online”, “producers of ventricina from Molise”, and so on, on ChatGPT and Perplexity.
I looked at the cited sources when available, and then searched for those same brand names in the threads of three specialty food forums in Italian and English, plus Quora (in Italian and English) and a few food subreddits.
The pattern, on a small but clear sample: the brands that AI named spontaneously appeared in at least one reasoned thread on a forum or Quora, where a user (often an enthusiast, not the brand itself) had recommended them inside a long answer. The brands absent from AI answers were also absent from those threads, even though they had well-crafted sites.
It’s an indicative test, not a study. The sample is small, the sector specific. But the pattern matches what I’ve been seeing for months across other vertical sectors: where there’s third-party conversation, there’s AI memory. Where there isn’t, the brand is an orphan entity.
One detail worth attention: in Perplexity’s answers, when the forum thread was linked as a source, the model didn’t simply cite the brand. It often picked up the angle with which the brand had been described in the thread (“small producer, long aging, direct sales”). This tells me that being named isn’t all that counts: what counts is the semantic context in which you’re named. I talked about this too in the article on E-E-A-T for AI: the model is learning not only “who”, but “how the who is described”.
A Quora account signed “XYZ Company” that answers “what’s the best oil from the South” with “our oil is the best because…
The test you can run yourself in 20 minutes
Want to see whether your brand is “heard” by the forums AI reads? Do this exercise.
Open ChatGPT and Perplexity. Run 5 real queries from your sector — the kind of things a curious customer would ask. For our Molise producer they would be: “best Italian caciocavallo to give as a gift”, “where to buy white truffle online”, “lesser-known regional extra virgin olive oils”, “typical cured meats from Molise”.
Look at which brands get named. Take those 3-5 names and search for them on Quora (site:quora.com brand name) and in the most active forums in your sector. Almost always they’re there.
Then search for your own brand with the same logic. If it doesn’t show up in any reasoned thread, you’ve found one of the reasons you’re left out of AI answers.
It’s an entry-level check. Real analysis requires professional tools that measure citation frequency and the diversity of the sources AI uses in its generated answers. But to understand “does the problem exist or not”, twenty minutes is enough.
Answer 2-3 questions a week, with long answers (300-500 words), well-reasoned, with the brand cited inside a concrete example (“when I select caciocavallo from Agnone for XYZ, I always look for…
The mistakes I see most often
Answering as an official brand. A Quora account signed “XYZ Company” that answers “what’s the best oil from the South” with “our oil is the best because…” doesn’t work. AI, like Google before it, can recognize self-promotion. The answer must be signed by a person — the owner, the master cheesemaker, the buyer — with real expertise and a mention of the brand inside a line of reasoning, not as a verdict.
Answering 2 questions and disappearing. The pattern useful to the model is systematic presence: 20-30 answers over time, not 3 in a week. The model weighs the consistency and recurrence of the voice.
Choosing only popular questions. Questions with 500k views are saturated with commercial answers; AI treats them with skepticism. Questions with 2-10k views but specific queries are gold: less competition, higher chance of being the “canonical answer” the model will use.
Ignoring the real vertical forums. Quora is useful, but for a specialty food brand from Molise, dedicated Italian and international gastronomy forums and food communities weigh more. Every sector has its places: architecture firms have other places, automotive has others still. The work is mapping the 3-5 places where your ideal customer really asks questions.
What can you actually do?
The operational actions, in order:
- Make a list of the 20-30 most frequent questions in your sector. Look at Google’s “People Also Ask”, the Quora questions with the most followers, the most-commented threads on vertical forums.
- Choose a real author inside the company. For the Molise e-commerce store it would be the owner or the head of purchasing — someone who knows the producers personally.
- Answer 2-3 questions a week, with long answers (300-500 words), well-reasoned, with the brand cited inside a concrete example (“when I select caciocavallo from Agnone for XYZ, I always look for…”). Never as a commercial conclusion.
- Track in a spreadsheet which questions you answered, with which signature, with which thread URL. After 3 months, run the ChatGPT/Perplexity test again and see if anything changes.
The connection with the work on brand entity I described in named entity recognition is direct: the more your brand appears in reasoned textual contexts, the more the system recognizes it as a stable entity, with a clear profile of what it does and where it operates.
Where this fits within the work on AI visibility
Quora and industry forums aren’t a magic channel. They don’t replace a well-built site, they don’t replace digital PR on industry media, they don’t replace a presence on Wikipedia when it’s relevant. But they’re the connective tissue that tells AI “there’s a real conversation around this brand”. Without that tissue, visibility in AI answers stays fragile: the model finds you on your site, but has no external confirmation of what you’re worth.
In the upcoming articles in the series on content distribution I’ll show you other channels where AI “listens”: the presence on Reddit as a recommendation engine, the use of LinkedIn as an AI source for B2B, and the podcast guest appearance as a citation signal that LLMs are learning to weigh.
The thread stays always the same: showing up in AI answers isn’t a matter of one channel, it’s a matter of consistency of voice across the places where real users ask real questions.
