Why AI assistants cite some brands and not others

Ask an AI assistant to recommend a product in your category and it will usually name a few brands with confidence. It can feel arbitrary, especially when a smaller competitor shows up and you do not. It is not arbitrary, but it is also not something you can buy or switch on. Understanding where answers come from shows you which parts you can influence.

Two places an answer comes from

Most AI answers draw on two kinds of knowledge.

What the model learned during training. Large language models are trained on a huge snapshot of public text. If your brand was discussed often, consistently and in the right context in that text, the model "knows" you and may name you even without searching. If you were barely mentioned, or only on your own site, it may not.

What the engine retrieves at answer time. Perplexity, Google's AI Overviews and AI Mode, and ChatGPT and Gemini when they search the web, fetch pages while composing the answer. They then summarise those pages and often cite them. Here, what matters is whether the pages the engine finds for that question mention you.

Training data changes slowly and you cannot edit it directly. Retrieval changes every time the engine searches, and it is shaped by pages you can create or improve. That is why the most practical work focuses on retrieval, while also building the kind of broad presence that ends up in future training data.

What tends to get a brand named

There is no published ranking formula, but the answers themselves show clear patterns when you read enough of them.

Third-party pages that list you

When an engine answers "best X for Y", it often leans on pages that already compare options: review platforms, directories, "best of" articles and comparison posts. If those pages list your competitors and not you, the answer usually follows them.

This is why the sources cited in an answer are so useful. They are a direct view of which pages the engine trusted for that question.

Consistent facts across the web

Engines are more confident naming a brand when independent sources agree about what it does, who it is for and how it compares. If your own site says one thing, your directory profiles say another, and an old article describes a product you no longer sell, the model has a muddier picture of you.

Pages that answer the question directly

A page titled "How to choose payroll software for a 20-person company" that answers the question clearly, with specifics, is easier to retrieve and quote than a generic product page. Engines retrieve passages that match the question; a page built around the question has more of them.

Real discussion in communities

Reddit threads, Quora answers and niche forums show up frequently as cited sources, because they contain people comparing options in plain language. A thread where several users recommend your competitors and nobody mentions you is a gap the engine will faithfully repeat.

What a brand can control

You cannot control the answer, but you can control a lot of the material it is built from.

1. Directory and review profiles

Find the review and directory sites that are cited for your questions. Depending on your market these might be G2, Capterra, Clutch, Trustpilot, Yelp or industry-specific directories. For each one:

• Claim or create your profile.
• Use the same company description, category and key facts everywhere.
• Fill in the fields engines tend to quote: who you serve, pricing model, location, main features.
• Encourage genuine customers to leave reviews. Do not buy or fabricate them; platforms remove fake reviews, and an engine summarising them can repeat any problems they reveal.

2. Comparison pages

Buyers ask comparison questions constantly: "A vs B", "alternatives to C". If the only comparison pages that exist are written by your competitors, those are what the engine will find. Publish honest comparison pages on your own site:

• Name the alternatives plainly.
• Say who each option suits, including when a competitor is the better fit.
• Use a clear table for features and pricing, and keep it up to date.

Honesty is not just ethics here. A one-sided page is less likely to be trusted or quoted than a fair one.

3. FAQ pages built from real questions

Take the questions you track and answer them on your site, one clear answer per question. Use the buyer's own wording in the headings. Add FAQ structured data (FAQPage schema) so the question-and-answer structure is explicit to crawlers.

4. Community presence, done properly

On Reddit and Quora, the goal is to be a useful participant, not to post adverts.

• Answer questions in your area where you have real expertise, and disclose that you work for the company when you mention it.
• Follow each community's rules; many ban self-promotion outright.
• Where an existing, frequently cited thread is missing an option, a helpful, transparent reply can add you to the conversation.

5. llms.txt

An llms.txt file is a short Markdown file at the root of your site that summarises who you are and links to your most useful pages. It is a proposed convention rather than a standard that every AI company has committed to follow, so treat it as a low-cost addition, not a fix on its own. It takes an hour to write and makes your key facts easy for any tool to read. We cover it in detail in how to write an llms.txt file.

6. Your own site's basics

None of the above helps if engines cannot read your site.

• Do not block AI and search crawlers you want to be found by in robots.txt.
• Make sure key content is in the HTML, not only rendered by JavaScript.
• State plainly what you do, for whom, and where, on the home page and about page.
• Keep pricing and product pages current; outdated facts get repeated.

What you cannot control

Be wary of anyone who promises guaranteed placement in AI answers. You cannot pay to be included in ChatGPT's or Gemini's organic answers, you cannot edit a model's training data, and engines change how they retrieve and rank sources without notice. What you can do is increase the number of trustworthy pages that describe you accurately for the questions that matter, then measure whether answers change over the following weeks.

Putting it together

1. Collect answers for your buyer questions and list the sources they cite.
2. For each frequently cited source, check whether it mentions you.
3. Fix the gaps in order of how often the source is cited: profiles first, then comparison and FAQ pages, then community threads.
4. Re-measure weekly and watch the trend in visibility share and citation share.

The brands that show up in AI answers are rarely there by accident. They are the ones that are easy to find, described consistently, and present in the places engines already look.