ChatGPT recommends the products it finds on the pages it reads right before it answers: "best X" listicles, review platforms, Reddit threads and, more and more, the brands' own product and pricing pages. To get recommended, get your product onto the pages ChatGPT already reads for your category, and make your own pages easy to quote. Writing a pile of your own content and hoping the model notices works too, but it is the slow way.
Below: a two-minute test that shows you exactly which pages decide your category, the four kinds of source and how to get onto each, and what our data on 200 sites says about who gets cited at all.
Key takeaways
- ChatGPT search rewrites a prompt into one or more targeted web searches, then recommends from the pages it finds. If you are not on those pages, you are not in the answer.
- In Ahrefs' study of 750 "best" prompts, "best X" blog lists made up 43.8% of the page types ChatGPT cited, and brands placed higher on those lists were more likely to be recommended.
- Since August 2026, ChatGPT's default model reaches for brands' own product and pricing pages far more often, so those pages need plain answers on price, features and fit.
- In Outrank's study of 200 sites, sites with Domain Rating 30+ got 29× the AI citations of sites under DR 10, and Google impressions were the strongest single predictor.
- Nobody can guarantee a ChatGPT recommendation, and ads in ChatGPT do not change its answers. You can only make yourself hard to leave out.
The two-minute test
Open ChatGPT and ask the question your buyer would ask: "What are the best [your category] tools for [your use case]?"
You get a list. Your competitors are probably on it. You might not be.
Now click Sources under the answer (or ask "Which sources did you use?"). If ChatGPT searched the web, you will see the pages it pulled the answer from: a listicle from this year, a G2 category page, a Reddit thread, a comparison post on a blog you have never heard of.
OpenAI's own help page explains what happened: ChatGPT "typically rewrites your query into one or more targeted queries" and sends them to search providers, then may send follow-up queries after reading the first results. The recommendation is a summary of what those searches returned.
Run the same prompt three or four times, reword it the way different buyers would ("for a small team", "cheapest", "for agencies"), and repeat it in Perplexity and Google AI Mode. Write down every source URL. That list is the real battlefield for your category. Our free tool can run a question on ChatGPT, Perplexity, Gemini and Claude at once and label each source as a listicle, comparison, product page, forum and so on.
Where ChatGPT's recommendations come from
Almost every source on your list falls into one of four buckets. Each one has a different way in.
| Source | What it looks like | How you get onto it |
|---|---|---|
| "Best X" blog lists | "The 9 best meeting schedulers in 2026" | Pitch the author a reason to add you (see how to get into the listicles AI quotes) |
| Review platforms | G2, Capterra, Trustpilot category and profile pages | Claim your profile and ask real customers for reviews (see what ChatGPT reads about your brand) |
| Community threads | Reddit "what do you use for X?" posts, LinkedIn articles | Answer as yourself, with disclosure (see which social platforms AI cites) |
| Your own pages | Pricing, feature, comparison and docs pages | Make them crawlable and specific (see why ChatGPT skips your site) |
Get onto the pages it already reads
Most founders react to the test by writing more of their own content. It helps eventually. The faster move is to get onto the pages that are already in the answer.
Look at the listicles and comparison posts on your source list. Email the person who wrote each one. Offer something that makes their article better:
- a free account so they can actually test your product,
- a fact their article is missing (a new pricing tier, a feature competitors lack),
- a fresh screenshot for a section that has gone stale,
- a correction, if they describe you wrongly.
Writers of these lists update them often. In Ahrefs' study of lists ChatGPT used as sources, 79.1% had been updated in 2025 and 26% in the previous two months. A writer who updates every quarter is happy to have a reason to do it.
Position matters too. Ahrefs found that brands placed higher on a cited list were more likely to be recommended. Being number 9 of 12 is a start, not a win.
Make your own pages easy to quote
Third-party lists get you into the conversation. Your own pages increasingly close it.
According to the same August 2026 analysis, ChatGPT's new default model often uses the site: operator to search inside brand domains, with queries like "site:yourbrand.com pricing". In that analysis, listicles fell from 15.47% to 8.19% of retrieved pages while product pages rose to 16.39%. In our own deep dive into ChatGPT's searches across 48 questions, 85% were site: searches on one brand's own website. When the model looks you up directly, it needs to find answers, not slogans.
- Pricing page: prices in plain text, what each plan includes, who it is for.
- Feature pages: one page per job your product does, named the way buyers say it.
- Comparison pages: "[You] vs [competitor]" and "[competitor] alternatives", honest about where the other tool wins.
- Crawlable: server-rendered HTML that OpenAI's search crawler is allowed to fetch. Our ChatGPT visibility checklist covers the technical side.
Selling physical products? OpenAI's shopping help page says product results are chosen from structured product data (price, descriptions), reviews and other third-party content, and that they are not ads. Shopify merchants are already connected through Shopify Catalog. Other merchants can apply to send OpenAI a direct product feed.
Authority decides who makes the shortlist
Outreach gets you onto pages. Authority decides whether ChatGPT, Perplexity and Google's AI features find your own site at all.
In Outrank's study of 200 customer sites, AI citations climbed steeply with Domain Rating and with verified backlinks. Google impressions were the strongest predictor of citations (Spearman 0.80). That fits how these assistants work: they search first, then summarize what ranks.
Median AI citations per site
Hover or tap the chart for exact values.
One more number from that study worth knowing: ChatGPT produced only 4.8% of the citations, against 41.6% for Perplexity. ChatGPT gets the headlines, but the same work gets you into every assistant.
This is the slow, compounding part, and it is what Outrank automates: an SEO article a day on the questions your buyers ask, published to your site, plus backlinks from relevant sites to build the authority that gets those pages ranked and quoted.
Track it like a ranking
Re-run your prompts every month, or use an AI visibility tracker such as OtterlyAI or Peec AI, which run your prompts on a schedule. Answers shift as models change, so a single screenshot proves little.
Measure three things: how often you are mentioned for your category prompts, which sources mention you, and which sources mention competitors but not you. The third list is your outreach plan for next month.
FAQ
Can you pay ChatGPT to recommend your product?
No. ChatGPT started testing ads in the US in February 2026 for Free and Go users, but OpenAI states that ads are labeled, shown separately and do not influence answers. Product results in shopping answers are not ads either. The only way into the answer itself is through the pages ChatGPT reads.
Why does ChatGPT recommend my competitor and not me?
Usually because your competitor appears on the listicles, review pages and threads ChatGPT found for that prompt, and you do not. Run the two-minute test above, list the sources, and check which ones mention them but not you. Those pages are your gap.
Does ChatGPT recommend products from training data or from the web?
Both. Without a web search, it answers from training data, which can be more than a year old. With search, it rewrites the prompt into queries and builds the answer from current pages, with citations. Buying questions ("best X for Y") often trigger a search, which is why getting onto current pages matters more than hoping the model memorized you.
How long does it take to get recommended by ChatGPT?
A mention added to a page ChatGPT already cites can count as soon as ChatGPT reads the updated version, which depends on how often that page gets recrawled. Building your own site's authority is slower: in Outrank's study of 2,119 sites, the median site's Google clicks grew slowly for about three months before compounding. Expect weeks for outreach wins and months for authority.
How do I know if ChatGPT recommends my product?
Ask it your buyers' prompts, several times and in several phrasings, and check the sources. For anything beyond a spot check, use an AI visibility tracker that runs a fixed set of prompts on a schedule across ChatGPT, Perplexity and Google's AI features.
