Citebase
Guides Pay $197

Other people’s pages

Reddit, reviews, and forums: where AI assistants pull recommendations

When an AI answer recommends a product, it is often repeating what people said on forums, review sites, and Reddit threads it found. Google says its AI features can reflect discussions across the web, including forums. Google and OpenAI both have data agreements with Reddit. None of them publishes how much weight these sources get, so the useful work is finding the threads that matter for your category and making sure what is there is true.

What the companies document

  • Google. Its optimization guide for generative AI features says those features can show what is said about products across the web, including in blogs, videos, and forum discussions. It adds that seeking inauthentic mentions is less helpful than it might seem, because ranking and spam systems still apply.
  • Google and Reddit. In a February 2024 post, Google announced access to Reddit’s Data API, which it described as real-time, structured, unique content. Google said the deal would allow more content-forward displays of Reddit information across its products, and that it does not change how Google uses publicly available, crawlable Reddit content.
  • OpenAI and Reddit. In a May 2024 announcement, OpenAI said it would get access to Reddit’s Data API and bring Reddit content into ChatGPT and new products, especially on recent topics.
  • OpenAI on reviews. The shopping help article says review summaries in ChatGPT’s product results come from public websites, and that reviews and ratings are not verified by OpenAI.
  • Perplexity. Its help center says answers combine several sources found on the web and include citations. It does not say which kinds of sources it prefers.

That is all of it. You will see percentages online for how often Reddit or review sites get cited. None of the companies above publishes such a figure, and this guide does not repeat any.

Why threads and reviews get quoted

This is reasoning, not vendor documentation. A buyer’s question often looks like a forum post title: “what do you use for scheduling podcast guests?” A thread with that title, where several people name tools and say why, is a close match for both the question and the answer the buyer wants. A review page that compares three tools has the same shape. A product homepage usually does not.

Threads also carry first-hand detail that marketing pages leave out: what broke, what it cost after the trial, what support was like. Google’s guide asks for content with a first-hand perspective. Whether or not a model weights that directly, it is the kind of detail that makes a thread useful to quote.

Find the threads that matter for you

  1. Run the questions

    Use five buyer questions with the manual check. Perplexity and Google AI Overviews show their sources openly. ChatGPT shows them when it searches.

  2. Open every forum and review link

    Note the site, the date, whether you are mentioned, and what is said. A three-year-old thread that calls you “abandoned” matters more than a hundred kind tweets if it is the one being cited.

  3. Search the communities directly

    Search the forums and review sites in your category for your product name and for your top competitor’s. Read what people say when you are not in the room.

The mentions guide covers why a mention on a cited page can matter more than a link elsewhere.

Where to look first

The cited pages from your own buyer questions are the best guide. If you have not run those yet, these are the usual places to check, roughly in order of how often they come up in product categories:

  • Subreddits and other forums for your category and for the job your buyer is doing, not only for your product type.
  • Review and comparison sites that list your category.
  • Q&A sites and community help forums where people ask for tool suggestions.
  • Video reviews and their transcripts. Google’s guide names videos alongside blogs and forums.
  • Integration partners’ community forums, where people ask what works with what.

That order is a starting point from general practice, not data. Your own citations will tell you which matter for your category.

Taking part without astroturfing

The honest version is simple, and it holds up over time.

  • Say who you are. If you reply about your own product, say you work on it, in the reply itself. The FTC’s rule on consumer reviews and testimonials requires officers, managers, and other insiders to clearly disclose their relationship when they write reviews or testimonials. Even outside that rule, communities notice an undisclosed founder fast.
  • Answer the question that was asked. If a competitor fits the person’s need better, say so. A reply that only pitches gets ignored, removed, or worse.
  • Read the community’s rules first. Most forums and subreddits set their own rules on self-promotion. Follow them, and ask a moderator when unsure.
  • Correct facts, not opinions. “We added CSV export in March, here is the changelog” is a fair reply. Arguing with someone’s bad experience is not.
  • Never post as a fake customer. Sock-puppet accounts and paid posts presented as organic are what Google means by inauthentic mentions, and fake reviews are prohibited under the FTC rule.

Reviews: what to do and not do

Review sites are a common source in product answers, and the rules are clearer than for forums.

  • Ask every customer, not a filtered list. The FTC rule bars offering compensation or incentives conditioned on a review being positive or negative.
  • Do not suppress bad reviews. The rule prohibits using unfounded legal threats, intimidation, or false accusations to get reviews removed, and misrepresenting a set of displayed reviews as all or most of them when negative ones were suppressed.
  • Do not run a fake independent review site. The rule prohibits presenting a site you control as independent reviews of a category that includes your products.
  • Reply to negative reviews in public, calmly. A reply that fixes the problem is part of what a reader, and possibly an answer, sees next to the complaint.

The FTC announced the rule on 14 August 2024 and said it allows civil penalties against knowing violators. It also covers buying fake followers or views to misrepresent influence for commercial purposes.

On your own site, Google’s review snippet guidelines say a business that controls the reviews about itself is not eligible for star ratings on pages using LocalBusiness or Organization markup, and that ratings must come directly from users. Testimonials on your homepage are fine to show. They will not earn stars in Google.

When a thread gets you wrong

  1. Make sure your own pages state the fact clearly, so anyone who checks finds the right answer.
  2. Reply in the thread with the correction, your affiliation, and a link to the source on your site. Keep it short.
  3. If the thread is old and locked, you cannot change it. Make sure newer, accurate discussion exists, by being helpful in current threads, not by planting one.
  4. Rerun your buyer questions later and see whether the cited sources changed.

Expect this to be slow. A cited thread can stay cited for a long time. The pages you control are the fastest thing to fix.

A short checklist

  1. A list of the forum threads and review pages cited for your buyer questions, with dates.
  2. A note on what each says about you, and whether it is accurate.
  3. Accurate, disclosed replies where a correction is useful and allowed.
  4. A review request that goes to every recent customer, with no condition on the rating.
  5. No insider reviews without disclosure, and no suppression of negative ones.

Related: why an assistant recommends a competitor, and getting cited by Perplexity.

Sources