Where do AI assistants get their answers, and how do you get cited?
Short answer
For recommendation questions, AI assistants with web search rely heavily on third-party pages: review sites, "best of" lists, comparison articles, marketplaces, media and forums, plus the brands' own sites. If competitors appear in those sources and you don't, they get named. A source gap analysis lists the exact sites to target.
Why does AI recommend my competitors and not me?
Because when an assistant answers "which store should I use" or "which agency is best for X", it doesn't judge companies itself. It summarizes what it finds. If ten pages it trusts list your competitors and none lists you, the answer will name your competitors. Your own site can say you are the best; independent sources decide whether AI repeats it.
Which sources do AI assistants cite most?
The mix depends on the assistant, the country and the topic, but for commercial questions the same types of sources come up again and again:
| Source type | Examples | Why AI uses it |
|---|---|---|
| Rankings and "best of" lists | "Top 10 online electronics stores in 2026" | Direct answers to recommendation questions |
| Review platforms | Trustpilot, Google reviews, industry review sites | Independent opinions and ratings |
| Comparison and price sites | Product comparison portals | Prices, availability, store lists |
| Marketplaces | Amazon, Allegro, eMAG and others | Products, sellers, ratings |
| Industry media and blogs | Trade publications, expert blogs | Expert framing of the market |
| Forums and communities | Reddit, local forums, Q&A sites | Real user experience |
| Encyclopedic sources | Wikipedia, Wikidata | Basic facts about entities |
| Brand websites | Your homepage, About, product pages | Facts: prices, terms, range |
Industry studies of AI citations in 2025 consistently found that user-generated platforms such as Reddit and encyclopedic sources such as Wikipedia are among the most cited domains overall, with large differences between assistants. In your niche, the list will be specific, which is why you have to measure it rather than guess.
How do you find the sources that matter in your market?
Use an assistant that shows its sources (Perplexity, ChatGPT search, Gemini) and run your customers' questions through it. For each answer, record which domains are cited. After 15–20 questions, group the domains and count them. Then split them into two lists:
- Sources cited in answers that mention you. These already work for you.
- Sources cited in answers where you are missing. This is your source gap.
Automated monitoring does the same at scale: Rekomo collects the sources from answers with Google Search grounding every week and shows the domains that appear when you are not named. See how to measure AI share of voice.
How do you close the source gap?
Work through the list from the most cited domain down:
- Rankings and lists: contact the author or publisher, offer facts, a sample, a discount for their readers or an expert comment. Many lists are updated yearly.
- Review platforms: claim your profile, make sure the details are correct, ask real customers for reviews after purchase. Never buy or invent reviews.
- Comparison sites and marketplaces: list your products with full specifications and correct prices.
- Media: offer data, market commentary or a case study rather than a press release about yourself.
- Forums and communities: participate as yourself, answer questions helpfully, and don't astroturf; communities notice, and so do models.
- Wikidata: if your company meets the notability criteria, make sure an item exists with your official website.
Does your own website still matter?
Yes, in two ways. First, assistants cite brand sites for facts: prices, delivery terms, specifications, addresses. If those facts are clear and readable, AI quotes them correctly. Second, third-party authors check your site before writing about you. Make sure AI crawlers can access it (robots.txt for AI bots) and that the text is in the HTML (JavaScript and AI crawlers).
A useful metric is how often AI cites your own domain in answers with web search. If it never does, AI is learning about you only from others.
How long until new mentions change AI answers?
For assistants with live search, a new page can be used as soon as it is indexed and ranks for the relevant query, which can take days to weeks. For what models know without searching, mentions have to be in the web snapshot used for the next training run, which takes months. Track your share of voice weekly to see when it moves.
What shouldn't you do?
Don't create fake "independent" review sites, don't publish rankings on your own blog that put you first and present them as neutral, and don't post disguised promotional comments. Besides breaking platform rules, these tactics are increasingly detected and can make your brand look less trustworthy in exactly the sources AI reads.
The AI-Ready Pack includes a source gap report for your market. Start with the free audit to make sure AI can read your own site first.