How to Get Recommended by ChatGPT (What 40% of Its Citations Reveal)

You get recommended by ChatGPT by getting recommended in the places ChatGPT reads. That's the entire mechanism. Semrush analyzed 150,000+ LLM citations across 5,000 keywords and found that 40.1% point to Reddit, ahead of Wikipedia (26.3%) and YouTube (23.5%).

So the real question is not "how do I persuade the model". It's "which sources does the model pull for my buyers' questions, and how do I show up in them". That work has a name: generative engine optimization.

This guide is industry-agnostic. It works for ecommerce brands, local services, coaches, agencies and apps. If you sell software, read the SaaS-specific version next, it goes deeper on prompt mapping.

How does ChatGPT decide what to recommend?

Two layers produce every recommendation. Training data sets the model's background knowledge of your category. Retrieval fills in the live answer: for commercial questions like "best running shoes for flat feet", assistants search the web, pull a handful of sources, and build a shortlist from what those sources agree on.

Reddit dominates that source pile, and not by accident. OpenAI signed a data partnership with Reddit in May 2024, giving ChatGPT real-time access to threads. Google pays about $60M per year to train on the same corpus. The models lean on Reddit because their makers paid for exactly that.

The detail that should shape your whole strategy comes from Profound's citation research:

99% of ChatGPT's Reddit citations point to individual discussion threads, not subreddit homepages or corporate content, and the median cited thread is about a year old (Profound).

One specific thread where your brand earns a genuine recommendation can feed AI answers for a year or more. And the audience keeps compounding: ChatGPT passed 900 million weekly users in early 2026. The full data set lives in our Reddit AI search statistics roundup.

The 4 levers, ranked by impact

Everything that moves ChatGPT recommendations fits into four buckets. Here they are, ordered by impact per hour invested.

1. Reddit threads (highest leverage)

Reddit's share of citations in commercial categories grew another 73% between October 2025 and January 2026 (Tinuiti × Profound). This is where recommendations are actually manufactured.

The work: find the threads your buyers' prompts surface, then contribute real answers inside them. Value first, your brand named only where it honestly fits, always within each subreddit's self-promotion rules, and from an account with some history behind it (karma guide here). One useful comment in a thread that gets cited beats fifty drive-by mentions. And don't automate the posting: Reddit's 2026 automation crackdown is wiping out exactly that behavior.

2. Third-party listicles and review pages

"Best X for Y" roundups get cited constantly. Pull the sources cited for your prompts, list the recurring articles, and pitch their authors a short, specific addition they can paste as-is. Same logic for review platforms: G2 for software, Tripadvisor for travel, Houzz for home services. Whatever your category's trusted aggregator is, assistants retrieve it.

3. Structured data

Models quote what they can parse. Add FAQ and product schema, use headings that match real questions, lead every page with a one-sentence answer, and publish an llms.txt index of your key pages. Then check robots.txt: plenty of sites still block GPTBot, then wonder why they're invisible in AI answers.

4. Your own site's content

Vendor pages get cited less than community and editorial sources, but they're the only citation you fully control. Publish honest comparison pages, keep pricing public and current (assistants love quoting prices, including stale ones), and write answer-first: the conclusion in the first two sentences, the context after.

Your first 30 days

A four-week sequence that works in any industry:

• Week 1: map. List 20-30 prompts your buyers actually type. Run them through ChatGPT, Perplexity and Google AI Overviews. Record who gets named and which sources are cited. That citation list is your target list.
• Week 2: groundwork. Identify the 5-10 communities where your buyers hang out (how to find them), read their rules, start contributing without any pitch. The free Reddit Leads Finder surfaces live threads where people are already asking for what you sell.
• Week 3: place. Earn a real position in 3-5 recurring threads, following the promotion playbook, and pitch 3-5 of the listicles from your week 1 list.
• Week 4: fix your site. Schema, answer-first rewrites, llms.txt. Then re-run the week 1 prompt set and compare.

This loop is what Readyt runs as a product: it finds the Reddit threads AI engines cite in your niche, helps you show up in them cleanly in about 15 minutes a day, then tracks which AI answers mention you, from $49/month. We built it after running the same playbook on our own app, Speechly. Fair caveat: it acts on Reddit specifically, it's not a monitor-everything enterprise suite.

How do you know it's working?

Never trust a one-off check. AI answers are violently volatile: Semrush's three-month tracking saw ChatGPT cite Reddit in roughly 60% of answers in early August 2025, then about 10% by mid-September.

So measure weekly, on three numbers:

1. Visibility rate: the % of your prompts where you're named at all
2. Average position: where you sit in the shortlist
3. Share of model voice: your mentions versus competitors'

You can start free with the AI Visibility Checker. And the effort is worth measuring properly: a visitor coming from AI search is worth 4.4x a traditional organic visitor (Semrush, 2025). Fewer clicks, but pre-sold ones.

FAQ

Paul-Marie Hamon

Founder @ Readyt

Paul-Marie is the founder of Readyt, the Reddit growth platform for SaaS. He has generated 16K€+ in pre-sales in 2 months using nothing but Reddit, and now helps founders turn Reddit threads into their #1 acquisition channel.