How to get your brand mentioned in ChatGPT and other AI answers

How AI assistants decide which brands to mention

When someone asks ChatGPT, Gemini, Claude, Perplexity or Google's AI Overviews a buying question, such as "what is the best tool for X", the answer comes from two places. One is what the model learned in training. The other, increasingly, is a live search: the assistant looks the question up, reads the pages it finds, and writes an answer from them. Most assistants show the pages they used as citations.

That second path is the one a brand can influence in months rather than years. If the pages the assistant retrieves for your question describe your product as a fit for that need, you are likely to be named. If they never mention you, you are unlikely to appear, however good the product is.

Two things follow. First, visibility is per question. A brand can be named for one question and missing from a nearly identical one, because a different set of pages was retrieved. Second, the sources matter as much as your own site. Assistants lean heavily on third-party pages: review sites, comparison lists, forums, video and Reddit. Data published by Profound, working with Reddit, looked at more than four billion AI citations and found Reddit the most cited domain across the answer engines it tracked, so the places people discuss products carry weight in what gets named.

None of this is a secret formula. It is closer to ordinary reputation: be clearly described, accurately, in the places people and machines look when they ask. The rest of this guide is the practical version.

Start by measuring where you stand

Before changing anything, find out what the engines say today. It takes an afternoon and it changes what you do next.

1. Write 20 to 50 buying questions in the words a customer would use. Include the plain category question ("best tool for faceless YouTube videos"), comparisons ("X vs Y"), problem questions ("how do I stop doing Z by hand") and price questions ("how much does X cost").
2. Ask each question in each engine you care about, in a fresh session without your history, and record three things: were you named, which brands were named instead, and which sources were cited.
3. Repeat it on a schedule. Answers change from run to run and from week to week, so one reading is a snapshot, not a score. Weekly or monthly is enough to see a trend.

Two readings tend to surprise people. The first is who is named instead of you: it is often a competitor you had not considered, or a category you did not think you were in. The second is the cited sources. If the same Reddit threads, list articles and review pages appear across questions, those are the pages that decide the answer, and they are where your effort should go.

Doing this by hand is fine at the start. At 50 questions across several engines it becomes tedious, which is the reason tools exist. The tracking guide covers how to set up the questions and read the results.

Make your own pages easy to quote

Your site cannot be cited if it does not answer the question. The pages that get used tend to share a few habits.

Answer first. Put the direct answer in the first two sentences of the page, then explain. An assistant building an answer looks for a passage it can lift, and a page that opens with a long story gives it nothing to lift.

Use the question as the heading. A heading such as "How much does it cost?" matches the way people ask and the way assistants search. Under it, give the number, the plan names and what is included, not "contact us".

Be specific and checkable. Prices, limits, supported platforms, dates, and named alternatives are all things an assistant can state with confidence. Vague claims ("the leading platform") give it nothing to say, and invented ones, when spotted, make the whole page less trustworthy.

Compare honestly. Pages that set out how you differ from named alternatives, including where a competitor is the better choice, are useful to readers and are the kind of page assistants cite for "X vs Y" questions. A comparison that only praises you reads as an advertisement and is less likely to be reused.

Use tables and lists for comparisons, and add structured data such as FAQ markup where it fits. Structure does not guarantee a citation, but it makes a page easier to parse.

Say when it was updated. A visible date and a habit of keeping prices and features current help both readers and engines that prefer fresh pages.

Earn mentions on the sites the engines already cite

Because assistants lean on third-party sources, your visibility depends on being described well off your own site. Look at the sources cited for your questions and sort them:

• Lists and comparisons ("best X tools"). Many are written by affiliates or agencies. You can contact the author with accurate information, a trial, or a correction. Don't offer money for placement without saying so, and prefer being useful to being persuasive.
• Review sites and directories. Complete, current profiles with real customer reviews. Do not buy or fake reviews: it is against the sites' rules and it can get a listing removed.
• Video. YouTube is among the most cited sources in several studies. A clear tutorial or comparison is both a citation and a place customers find you.
• Forums and communities, above all Reddit, covered next.

The pattern across these is the same: the engines trust pages that look independent. A mention a customer wrote is worth more to them than a claim you made about yourself, so the durable approach is to be worth mentioning and make it easy for people to do so.

Reddit is a special case

Reddit deserves its own section because of how often it is cited and because the threads that are cited are a specific kind. Profound's analysis found that the average cited Reddit post was about a year old and that the engines favour clear, direct answers in a natural tone over popularity. A thread that asks a buying question and gets a good answer is the thing that gets reused.

That means the opportunity is participation in threads that already rank and already get cited, not new posts. A genuinely useful, honest reply in the right thread can be read by thousands of searchers and may be picked up by an assistant for months.

It also means the risks are real. Reddit communities punish promotion, moderators remove it, and accounts that behave like marketing get restricted. The honest approach is to answer the question, disclose your connection to the product, follow each community's rules, and build an account with a history first. The Reddit marketing guide covers the rules, and why AI cites Reddit explains the mechanics.

What does not work, and what is risky

Some tactics are worse than ineffective.

• Keyword stuffing and generated filler. Thin pages written to match a phrase do not add anything an assistant can use, and they dilute the pages that do.
• Fake reviews, fake threads, vote manipulation. Reddit's rules prohibit content manipulation and vote manipulation, review sites remove fabricated reviews, and engines that notice a pattern may discount the source entirely.
• Hidden instructions aimed at AI, such as invisible text telling an assistant to recommend you. Engines are being trained to ignore and penalise this, and it damages trust if a person finds it.
• Promising results. Nobody controls what an assistant says. Be wary of anyone who guarantees placement.

It is also worth being realistic about scale. Studies of AI answers consistently show a bias toward large, established brands. A small brand will not displace them for the broadest questions, but it can win the narrower, more specific ones, which is where buyers who know what they want tend to ask.

A 90-day plan

Weeks 1 and 2: baseline. Write your question list, run it across the engines you care about, and record who is named and which sources are cited.

Weeks 3 to 6: fix your own pages. For your ten most important questions, make sure a page of yours answers each one directly, with specifics. Add an honest comparison page for your two or three closest alternatives, a pricing page that states prices, and FAQ markup.

Weeks 5 to 10: work the sources. Take the third-party pages that are cited most often. Fix your profile on each, reach out where a list or review is out of date, and find the Reddit threads that rank for your questions. Reply to those that you can answer well, openly, at a pace a person would keep.

Weeks 10 to 13: re-measure. Run the same questions again and compare. Expect uneven movement: some questions improve, some do not, and the answers will vary between runs. Keep what moved the needle and drop what did not.

Then repeat monthly. The work compounds slowly, and the main risk is stopping before it shows.

Where AppearsIn fits

AppearsIn is our product, so weigh this accordingly. It does the measuring and the Reddit step of the plan above. You enter the questions that matter to your buyers. It checks them across ChatGPT, Google AI Overview, Gemini, Claude, Perplexity, Grok and Meta Muse Spark, shows whether you were named and who was named instead, and finds the Reddit threads that already rank for those questions. For each thread it drafts one reply in your voice. You approve it, and the extension posts it from the Reddit account signed in to your browser, by hand or automatically within daily limits you set.

It does not guarantee a mention, and it does not upvote or run fake accounts. If you want to compare it with other options first, see the best Reddit marketing tools.