Quick picks: where are you starting?
- You don't know what buyers ask ChatGPT: start with the AI visibility prompt generator, then run step 3 below.
- You want to know what LLM SEO is before you start: read what LLM SEO is and what moves it.
- You already know Reddit threads get cited in your category: follow how to use Reddit to improve AI search visibility.
- You want the subreddits AI answers cite most: see Reddit marketing for AI search.
- You want someone else to do the off-site part: compare Reddit marketing services for AI search visibility, with prices.
What does ranking in ChatGPT mean?
Ranking in ChatGPT means two things: your brand is named in the answer, and your page is one of the few sources it links. There is no list of ten results to climb. The slots are also scarce. In our November 2, 2026 scan of 24 category questions per engine, ChatGPT cited 80 sources in all, about 3.3 per answer. Claude cited 326 for the same questions, Perplexity 240 and Gemini 158. A page that ChatGPT reads but doesn't pick earns nothing, so every step below is about being one of the three or four it does pick.
What our own test showed, November 2, 2026
ChatGPT named us more often than the other engines did, and gave our own pages a bigger share of its links, but it still left us out of 17 of 24 category answers. We ask four AI engines the same 8 buyer questions about our category three times each, plus 4 questions that name us, and count who gets named and which pages get linked. This is the run from November 2, 2026.
| Engine | Named us in category answers | Sources cited per answer | Our pages among the cited links |
|---|---|---|---|
| ChatGPT | 7 of 24 | 3.3 (80 links) | 7 of 80 (8.8%) |
| Claude | 5 of 24 | 13.6 (326 links) | 4 of 326 (1.2%) |
| Perplexity | 3 of 24 | 10.0 (240 links) | 12 of 240 (5.0%) |
| Gemini | 2 of 24 | 6.6 (158 links) | 7 of 158 (4.4%) |
How does ChatGPT find the pages it cites?
ChatGPT turns the question into a few web searches, reads what comes back and links a handful of the results. In our November 2 scan it ran 70 searches for 24 category answers, about 2.9 per answer, worded differently from the question. It named us all three times on two questions, and on both its searches used the words our pages use: "daily commenting service" and "most cited subreddits". Where its searches turned to Reddit ads and agency lists, it left us out.
| Question we asked ChatGPT | Searches it ran (from our scan log) | Named us? |
|---|---|---|
| Is there a service that posts comments about my company on Reddit every day? | "Reddit marketing agency daily commenting service brand Reddit community engagement"; "Reddit brand monitoring engagement service human replies Reddit agency" | Yes, 3 of 3 |
| Which subreddits do AI search engines cite most for software recommendations? | "AI search engines most cited subreddits software recommendations Reddit citations study"; "subreddits most cited ChatGPT Perplexity software recommendations study" | Yes, 3 of 3 |
| How much does Reddit marketing cost? | "Reddit ads minimum daily budget cost per click official Reddit ads 2026"; "Reddit ads pricing minimum spend CPM CPC Reddit business"; "Reddit marketing agency monthly pricing Reddit community management cost" | No, 0 of 3 |
| What are the best Reddit marketing services for B2B SaaS companies? | "B2B SaaS Reddit marketing agency services Reddit organic community marketing"; "Reddit marketing agency B2B SaaS case studies"; "Reddit ads agency B2B SaaS Reddit marketing services" | No, 0 of 3 |
How to rank in ChatGPT, step by step
Do the steps in order. Each one says what done looks like, so you know when to move on.
| Step | What to do | What done looks like |
|---|---|---|
| 1. Let OpenAI's search crawler in | Allow OAI-SearchBot in robots.txt and in your firewall or CDN bot rules. OpenAI's crawler page says sites that opt out of it "will not be shown in ChatGPT search answers, though can still appear as navigational links", and that robots.txt changes take about 24 hours to register. Blocking GPTBot (training) is a separate setting | Your server logs show OAI-SearchBot fetching your pages with 200 responses |
| 2. Get indexed in Google and Bing | Submit your sitemap to Google Search Console and Bing Webmaster Tools, and turn on IndexNow so new pages are announced the day they go live. OpenAI's crawler page doesn't name a search partner, so cover both | Your key pages show as indexed in both tools |
| 3. Write the questions and record a baseline | List 20 to 30 questions a buyer would ask, some naming you and most not. Ask each in ChatGPT with search on, in a fresh chat, and log who is named and which links appear | A dated sheet: named or not, and cited links, per question |
| 4. Publish an answer-first page for each search | For each question ChatGPT gets wrong or answers without you, publish a page whose first 40 to 60 words answer it in the buyer's words. Put your price and who it's for near the top | Every question on your list has one page of yours that answers it in the first lines |
| 5. Get onto the pages it already cites | Open the links in your baseline. Ask to be added to the lists and directories, claim your review profiles, and correct pages that describe you wrongly | You appear on the cited pages that are open to you |
| 6. Be part of the forum threads | Answer real questions in the Reddit threads and forums that show up in your baseline, from accounts with history. Follow each subreddit's rules | Useful answers in the threads that get cited for your questions |
| 7. Keep your facts the same everywhere, then re-check monthly | Use one wording for what you sell, your price and who it's for on your site, profiles and listings. Re-run the same questions every 30 days | A month-on-month count of answers that name you |
Which sources does ChatGPT cite most?
It depends on the category, so check yours before you pick a lever. Across our study of 41,592 buyer questions from January 9 to July 14, 2026, ChatGPT cited a Reddit thread in 29.6% of the answers where it cited anything, and Wikipedia in 18.6%. In our narrower November 2 scan about Reddit marketing services, ChatGPT linked a Reddit thread in only 1 of 24 answers. There it leaned on companies' own pages and lists: our site in 7 answers, growreddit.com and verticality.co in 5 each, axios.com in 4, and Reddit's and OpenAI's help centers in 3 each. Your own baseline tells you which of those two pictures your category looks like.
How do you know it's working?
Count two things: answers that name you, and visitors ChatGPT sends. On our site, DataFast counted 31 visitors from ChatGPT from October 15 to November 11, 2026, against 15 in the 28 days before. In our scan, ChatGPT added ?utm_source=openai to links it cited, which is one way those visits show up in analytics. ChatGPT also got our facts mostly right once it named us: 80 of the 83 facts it stated about us across 43 answers were correct. Read the same question list every 30 days. Daily checks mostly show noise.
Doing it yourself or having it done
Steps 1 to 4 and 7 are work on your own site and profiles. Steps 5 and 6 are slower, because they depend on other sites and communities. CommunityMentions, which is ours, does the forum part for you: it posts comments in the threads AI answers cite, for $1,000 a month after a free sample and a $1 setup charge. To weigh it against other services and prices, see Reddit marketing services for AI search visibility.
Is one scan enough to go on?
No single chat proves anything, which is why we ask each question three times: on the ReddiReach alternatives question ChatGPT named us in 1 of 3 runs, and only the run that searched our domain by name did. The scan covers 8 questions in one category, Reddit marketing services, so the baseline from step 3 is the number that counts for you. Our 41,592-query study measured what engines linked, not what people read or bought. OpenAI publishes no ranking factors, so the steps come from its crawler page and from what we measured.
Sources
- Overview of OpenAI Crawlers, OpenAI, read November 13, 2026: OAI-SearchBot, GPTBot, OAI-AdsBot and the 24-hour robots.txt note.
- CommunityMentions AI-answer scan of November 2, 2026: ChatGPT, Claude, Gemini and Perplexity, 12 buyer and brand questions run three times on each engine plus follow-up capability questions, 226 answers read; the searches ChatGPT ran are from the scan's own log.
- Where AI answers get their sources, CommunityMentions study of 41,592 queries, January 9 to July 14, 2026, published July 20, 2026.
- CommunityMentions server logs (Vercel), November 6 to 12, 2026: 1,073 requests from OAI-SearchBot, 189 from ChatGPT-User, 14 from GPTBot and 296 from Googlebot.
- DataFast analytics for communitymentions.com, AI referrals October 15 to November 11, 2026, against September 17 to October 14, 2026.
- Our price and terms: pricing section and company page, read November 13, 2026.