A marketing lead asks ChatGPT which tools are best for their category. Three competitors come back, each described in detail. Under the answer sit the sources: a review platform comparison, a forum thread from someone who switched providers, and a "best of" roundup from a tech publication. The brand's own website, with its carefully crafted positioning, appears nowhere. Every source that shaped the answer was written by someone else.
That is the uncomfortable centre of AI search visibility. AI tools don't take your word for it; they repeat what others say about you. Your own pages still matter for facts, specifications and pricing, but whether you are recommended at all depends heavily on how independent sources describe you. That makes a large part of AI search optimization an earned-media discipline, and it rewards only the honest version of it.
This post covers earning mentions on independent sources. It builds on brand entity optimization, which covers keeping your own profiles and listings consistent, and on AI share of voice, which shows how to see which sources are shaping answers in your category.
What the data says, and why it varies
Plenty of studies now analyse which websites AI tools cite, and their headline numbers often disagree. That is not because some are wrong; it is because they measure different engines, different questions and different time periods. The share of citations going to brands' own websites, for example, ranges from roughly a quarter to more than two-thirds depending on the study and the engine. So treat any single figure with care, and look instead at the patterns that keep repeating.
Three patterns hold up well. First, engines differ sharply. One audit found only about 11% overlap between the sources ChatGPT and Perplexity cited, and 2026 tracking shows different engines leaning on different platforms, with Reddit, YouTube and Wikipedia trading places at the top depending on the tool. Second, "best of" lists carry enormous weight. Research from Ahrefs reported that 43.8% of AI chatbot citations came from "best X" style listicles, and independent lists were cited far more often than brands' self-published ones. Third, review platforms and community discussion show up consistently, especially for software and consumer products, where buyers want evidence from people who have actually used the thing.
The five kinds of source that matter
Most third-party citations come from a handful of source types. Each earns trust for a different reason, and each has an honest way in.
airSlate
Send, sign, and manage documents online with legally binding e-signatures, templates, and workflows.
Best for: Teams sending contracts for e-signature
| Source type | Why AI trusts it | The honest way in |
|---|---|---|
| Editorial "best of" lists | Independent comparison, clearly structured | Be worth including; make evaluation easy |
| Review platforms | Volume of real customer experience | Ask every customer; respond to all reviews |
| Forums and communities | Candid, first-hand discussion | Participate openly; earn voluntary mentions |
| Video | Demonstration you can see | Creator reviews, tutorials, real demos |
| Reference and industry sources | Established, verifiable facts | Accurate listings, data, expert commentary |
Start with your own citation map
The most common mistake is chasing mentions everywhere at once. Instead, find out which sources actually shape answers for the questions that matter to you, and aim there.
1. Take 20 priority buyer questions from your research, the ones where you most want to be recommended.
2. Run each across the AI tools your buyers use, several times each, and record every source cited.
3. Tally the sources. You will usually find a short list of domains doing most of the work: a few review platforms, two or three publications, a couple of forums, some videos.
4. Mark where you appear and where you don't.
Example result: across 20 questions, one review platform is cited 14 times, a single publication's roundup 9 times, and a forum community 7 times. You have a thin, outdated profile on the review platform, no presence in the roundup and no mentions in the forum. That is your work list, in priority order.
The point: effort spent where AI tools are already looking beats effort spread across every possible site.
Finding the right questions in the first place is covered in question research for AI search. Because engines cite such different sources, run the audit per platform rather than pooling everything together.
The reality AI tools don't take your word for it. They repeat what others say about you. Your own pages still matter for facts, but whether you are recommended at all depends on how independent sources describe you.
Earning each kind of mention
Best-of lists. Look at the lists your audit surfaced, study what the included products have in common, and be honest about whether you compete on those criteria. When you approach the writer, offer genuine help: trial access, a clear fact sheet, a candid account of who your product suits and who it doesn't, and results they can verify. Original data is especially persuasive, because it gives the writer something new. Publishing your own list with yourself at the top is far less valuable, since self-published lists are cited much less often and readers discount them. Honest comparison content still has a place on your own site, built on the principles in comparison pages that win the buyer search.
Review platforms. Volume and recency matter, and brands present on several review platforms tend to be cited more. Ask every customer for a review at a moment of genuine satisfaction, spread requests across the platforms your audit identified, keep your profiles complete and current, and reply to reviews, including critical ones. Customer photos and videos in reviews add weight, handled with the permissions covered in getting customer content and the rights to use it.
Forums and communities. These are heavily cited, which makes them tempting targets for manipulation, and that is exactly why the honest route matters. Have real people from your company take part openly, clearly identified, answering questions helpfully even when your product isn't the answer, and following each community's rules on self-promotion. The strongest community presence comes from customers who mention you unprompted because the product and support earned it.
Video. Several engines cite video heavily. Tutorials, honest demos and independent creator reviews all count, and careful selection of creators matters, as covered in how to find and vet influencers. Reference and industry sources round it out: accurate directory listings, industry association profiles, expert commentary for journalists, and original research others will cite. Getting that work seen is a distribution job, the discipline behind a content distribution system.
What not to do
The shortcuts are well known and they backfire. Fake reviews are banned outright under the FTC's rule on consumer reviews and testimonials, finalised in 2024, which also prohibits incentives that depend on a review being positive. Undisclosed promotion, such as employees posting as ordinary users, paid mentions presented as independent opinions, or sponsored list placements without disclosure, is deceptive endorsement, covered in our guide to FTC disclosure rules. Fake community accounts break platform rules, get banned and tend to be exposed publicly, often by the very communities you were trying to win over.
There is a practical argument as well as an ethical one. AI systems are designed to weigh independent, consistent evidence, and the sources they rely on actively fight manipulation. Tactics that fake independence are working against the direction everything is moving. The durable approach is slower and more reliable: be genuinely good, make it easy for people to say so, and make it easy for writers to verify it.
Measure mentions, not just citations
Track two things separately over time: whether you are mentioned in answers, and whether your own pages are cited as sources. Third-party work often shows up first as more mentions, and as better descriptions of what you do, before it changes citations to your own site. Use a fixed question set, run it regularly, and read the trend by platform rather than any single result, using the share-of-voice method referenced above. Keep your own pages strong in parallel, since the sources that cite you often link back for details, following the principles in answer engine optimization and the GEO playbook. For the bigger picture of how these disciplines fit together, see our plain-English guide to AI search optimization.
The short version
AI tools don't take your word for it; they repeat what others say about you. Your own website still matters for facts, but whether you are recommended depends heavily on independent sources. Studies of AI citations disagree on the exact numbers because they measure different engines and questions, but three patterns hold: engines differ sharply in which sources they cite, "best of" lists carry huge weight, with Ahrefs research putting 43.8% of AI chatbot citations on such listicles, and review platforms and communities appear consistently. Focus on five source types: editorial lists, review platforms, forums, video and reference sources. Start with a citation audit across your priority questions, find the handful of domains doing most of the work, and aim there first. Earn list placements by being worth including and easy to evaluate; build reviews by asking every customer and responding to all of them; join communities openly; support honest video. Avoid fake reviews, undisclosed promotion and fake accounts, which break the law and platform rules. Measure mentions and citations separately, per platform, over time.
Want AI tools to recommend you, not your competitors?
We help brands earn the independent mentions that shape AI answers.
Explore SEO →Frequently asked questions
Why do AI tools cite third-party sites instead of brand websites?
Because a recommendation is more trustworthy when it comes from someone other than the seller. When a buyer asks an AI assistant which product is best for a particular need, the system looks for evidence that a brand is genuinely well regarded, and independent sources such as editorial reviews, comparison lists, review platforms, forum discussions and videos provide that evidence in a way a brand's own marketing cannot. Brand websites still matter: they are frequently cited for factual details such as pricing, features and specifications, and some studies find they make up a large share of citations on certain engines. But whether a brand appears in a recommendation at all often depends on how it is described elsewhere. The practical implication is that AI search optimization is partly an earned-media discipline. Your own pages need to be clear and accurate, but being named consistently and positively across independent sources is what makes you a credible answer.
Which third-party sources do AI tools cite most?
It varies considerably by engine and by category, and studies measuring it often disagree because they use different queries and methods. That said, several patterns are consistent across 2026 research. Community platforms, especially Reddit, and video, especially YouTube, sit near the top of citation rankings for several major engines, alongside Wikipedia. Editorial 'best of' and comparison lists are heavily cited; Ahrefs research reported that 43.8% of AI chatbot citations came from 'best X' style listicles, and independent lists were cited far more often than brands' own self-promotional lists. Review platforms such as G2, Capterra, Trustpilot and Consumer Reports are also prominent, particularly for software and consumer products. Crucially, different engines favour different sources, with one audit finding only around 11% overlap in cited sources between ChatGPT and Perplexity, so the right mix of sources depends on which AI tools your buyers use and what you sell.
How do you get your brand into best-of lists?
Earn it by being genuinely worth including and making it easy for editors to evaluate you. Start by identifying the lists that AI tools actually cite for your priority questions, since these are the ones worth effort. Then look at what the included products have in common and whether you genuinely compete on those criteria. When you approach the writer or publication, offer something useful: trial access, a clear fact sheet, honest details about who your product suits and who it does not, and original data or customer results they can verify. Keep your review platform profiles current, since editors often check them. Be careful with paid placements: if money changes hands, the relationship generally needs to be disclosed, and undisclosed pay-for-play lists erode trust and can create legal risk. Publishing your own list with yourself at number one is far less valuable, because self-promotional lists are cited much less often than independent ones and readers discount them.
Is it acceptable to post about your brand on Reddit to get AI citations?
Only transparently and as a genuine participant. Reddit and similar communities are heavily cited by several AI engines, which makes them tempting targets for manipulation, but fake accounts, undisclosed employee posts and coordinated promotion break community rules, get removed and banned, and can damage your reputation when discovered. They can also fall foul of advertising law, since undisclosed promotion by people connected to a brand is a form of deceptive endorsement. The legitimate approach is to have real people from your company participate openly, clearly identified as working for you, answering questions helpfully even when your product is not the answer, and following each community's specific rules on self-promotion. Better still, focus on giving customers reasons to mention you voluntarily: a strong product, excellent support and easy-to-share results. Genuine community goodwill is slow to build, but it is the only kind that lasts and the only kind worth having.
