You can't buy, request or guarantee a citation from ChatGPT or Perplexity. When they search the web, they cite pages their own search found and judged relevant to the question, and OpenAI says plainly that "Placement is not guaranteed." What you control is whether your pages can be found, whether they answer the questions buyers actually ask, and whether other sites name your product.
- Be crawlable: allow OAI-Searchbot and keep product, pricing and docs pages public.
- Match real questions: one page per buyer question, with a title and opening paragraph that answer it.
- State facts plainly: pricing, features, who it's for and its limits, nothing you can't back up.
- Get named elsewhere: reviews, launches, directories and articles give AI search third-party pages to find.
- Stay realistic: directories, launch platforms and backlinks help people and crawlers find you. They are not a citation ticket.
Disclosure: DirPilot sells directory submissions, so we have an interest in you getting listed, and we say below where listings don't help. How ChatGPT and Perplexity cite comes from their own help pages and the research linked below, all of which can change; directory data comes from DirPilot's directory pages as of our October 2026 checks.
Cited vs mentioned
"ChatGPT recommended us" can mean two things:
- A citation: the answer links to your page, as an in-line citation in ChatGPT or a numbered source in Perplexity.
- A mention: the answer names your brand, with or without a link.
A 2026 University of Toronto study, What Drives AI Citations?, followed 158,099 answers with 468,037 citations to commercial questions over two 90-day periods. It found that "Every page ChatGPT cites is one its own search retrieved", while ChatGPT named brands without any supporting page in 46% of its mentions. In the authors' words: "Pages are thus cited when they are searched; brands are also named when they are remembered."
A new startup can't count on being remembered: in the study's market of small software vendors, they "depend on being found by search and on third-party pages that name them." The study is observational, covers English-language US questions and used the engines' APIs, which "need not behave like their consumer applications". Evidence, not a rule, but for an unknown brand, being found by search is the job.
How ChatGPT finds and cites sources
From OpenAI's help article Searching the web with ChatGPT:
- When it searches: ChatGPT "may search the web automatically when your question would benefit from current information", or when the user chooses Search.
- How it cites: "Responses that use web search may include citations." A Sources button, when available, shows "cited sources and other relevant links."
- How it queries: with other search providers, it "typically rewrites your query into one or more targeted queries". Your page has to match those searches, not just the user's words.
- Reliability: "Search results and citations can be incomplete, outdated, or incorrect."
- Eligibility: "ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information. Placement is not guaranteed." To be eligible, "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses."
OpenAI's transparency page adds that language models evaluate content "based on meaning, intent, and relevance", and automated systems pick results "considering factors like user intent, relevance, and recency." Its API docs say responses cite "URLs found in the web search results." OpenAI publishes no weights, so nobody outside it can say exactly how a source gets picked.
How Perplexity finds and cites sources
Perplexity says it uses AI "to search the internet in real-time" and that "Each answer includes numbered citations linking to the original sources" (How does Perplexity work?).
The number of cited sources depends on the search. A basic search "gives you a fast answer and is best for quick facts." Pro Search goes wider: "Instead of a quick answer drawn from one or two sources, it synthesizes information from a wide set of high-quality sources and links directly to each one" (Getting started; see also What is Pro Search?). In a quick answer, competition for those few slots is tight.
Source labels, and what they don't mean
Some citations carry a Government, Academic or Trusted label (Understanding source labels). Perplexity "rates the whole website a source comes from, not each individual page," asking questions such as "Does the site correct its mistakes?" and "Does it say who wrote each piece?"
Your site probably won't have a label, and that's fine: "No label does not mean a site is low quality, and it is not a negative judgment." Nor can you buy one: "Perplexity's partnerships, payments, and other business arrangements do not affect a site's label." Still, name your authors and correct errors visibly.
The help pages we read describe no site-owner eligibility step like OpenAI's. Perplexity's developer documentation has a Perplexity Crawlers page; check it before editing robots.txt.
ChatGPT vs Perplexity at a glance
| ChatGPT | Perplexity | |
|---|---|---|
| Searches when | Current information helps, or the user picks Search | You ask a question |
| Sources shown as | In-line citations and a Sources panel | Numbered citations |
| Sources per answer | Not stated | One or two (basic), many more (Pro Search) |
| What decides | Meaning, intent, relevance, recency | Not published in its help pages |
| Site owner step | Allow OAI-Searchbot and OpenAI's searchbot IPs | None documented; labels via source review |
How this differs from Google SEO
Classic SEO ranks a page for a keyword, with links as one input (see how many backlinks you need to rank and what a good Domain Rating is). An AI assistant rewrites the question into its own searches, retrieves pages and cites a handful.
In the Toronto study, what that search retrieved "explains far more of the variation than page properties (32% against 2.8%)", and the rewritten queries "rarely coincide with Google's or Bing's results". The authors' conclusion: "a good rank in classic search is not a proxy".
The basics overlap: a page must be crawlable, indexed and useful either way. Google's spam policies even cover "attempting to manipulate generative AI responses in Google Search." That rule is for Google's own products, not OpenAI's or Perplexity's, but it's a sensible bar.
What raises your odds of being cited
1. Let the crawlers in
Follow OpenAI's one documented requirement: allow OAI-Searchbot in robots.txt, and make sure your host, CDN or bot protection doesn't block OpenAI's published searchbot IPs. Keep pages you want cited public, without a login wall or noindex, with the main text in the HTML. In the Toronto study, pages needing JavaScript for their main text got fewer Gemini citations; for ChatGPT and Perplexity the gap wasn't statistically significant.
2. Answer the questions buyers ask
Collect real questions from sales calls, support tickets and community threads, and give each important one a page or section whose title and first paragraph answer it in the buyer's words. For example, a scheduling tool might answer "Does it work with Google Calendar?"
That's the content finding that held up best in the Toronto study: "pages whose title and paragraphs match the questions people ask are cited more, whereas FAQ blocks, structured data, summary boxes and page speed appear effective only across websites." It also found that "at equal rank, concise pages from the original source win over roundups and forums."
Not every study agrees on details. The observational GEO-16 study, built on 1,702 citations from Brave Summary, Google AI Overviews and Perplexity, found that "Metadata and Freshness, Semantic HTML, and Structured Data showed the strongest associations with citation" in its English-language B2B SaaS pages. Neither study proves cause. Clean HTML, accurate titles and update dates are cheap, but won't replace a page that answers the question. Google's helpful content guidance sets a fair standard either way: content "created primarily for people, and not to manipulate search engine rankings."
3. Make product facts easy to quote
Put price, features, limits and fit in plain text: a pricing page with real numbers, features and integrations in words rather than only screenshots, who the product is and isn't for, dated comparisons you can support, and an About page that says who's behind it.
The original GEO paper (Aggarwal et al., KDD 2024) tested presentation changes in a benchmark: adding citations, quotations and statistics "achieved a relative improvement of 30-40%" on one of its two visibility measures, results varied by domain, and keyword stuffing did 10% worse than the baseline on Perplexity. These are lab results. The engine was handed fixed sources (on Perplexity, as file uploads), so the test measured how much an answer used a page, not whether it got found. And use only real statistics and quotes.
4. Get named on other sites
In the Toronto study, a niche vendor was named mainly when a cited third-party page named it. So build those pages: review profiles, launch pages, directory listings, integration partner pages, podcast show notes and write-ups by real users. Our first backlinks playbook covers where these come from.
5. Stay discoverable the classic way
Relevant links and listings help people and crawlers find you; our post on whether directory submissions work covers what they do for Google. We know of no public evidence that listings cause ChatGPT or Perplexity citations, and the help pages above don't mention them.
Where directories and launch platforms fit
A listing or launch page is a third-party page that names your product, puts it in a category and describes it, so an assistant searching your category could retrieve it. Could, not will: we have no data showing listings lead to AI citations. What listings do reliably is more modest: visitors, more paths for crawlers, sometimes an early followed link, and pages that appear when people search your name.
Some examples, as of when DirPilot last checked (October 2026). DR is Ahrefs' Domain Rating; "unchecked" means we haven't confirmed the link type on a live listing.
| Directory | Type | DR | Link type | Price | Approval |
|---|---|---|---|---|---|
| Product Hunt | Launch platform | 91 | Nofollow (ugc) | Free | Not listed |
| G2 | Review site | 91 | Unchecked | Free, paid plans | ~5 days |
| Crunchbase | Company database | 91 | Unchecked | Free | ~0 days |
| Startup Fame | Startup directory | 84 | Dofollow (free plan needs badge) | Free, or from $19 a month | ~7 days |
| BetaList | Pre-launch platform | 77 | Dofollow | Paid | Not listed |
| Peerlist | Weekly launches | 77 | Unchecked | Free | ~0 days |
The three highest DRs are nofollow or unchecked. That's typical: on our high-DR directories list, about half of the links checked on live listings are nofollow, "Product Hunt among them." For being found and named, the page matters more than its link type, or as our launch platforms page says: "Treat a link as a side effect of a good launch, not the reason for one." Confirm prices and terms on each directory's own page.
To pick yours, see our best startup directories (30 ranked by DR), Product Hunt alternatives (13 launch platforms), free dofollow directories or every directory we track.
A practical checklist for startups
- Public pages: product, pricing, docs, FAQ and About open without a login.
- Crawler access: OAI-Searchbot and OpenAI's searchbot IPs allowed, no noindex on key pages, sitemap submitted.
- A question list: 10 to 20 buyer questions, each answered by a page's title and first paragraph.
- Plain facts: pricing, features, audience and limits in text, with update dates.
- Honest comparisons: only claims you can back up, with sources linked.
- Listings written per site: same name and one-liner, but its own description on each directory. Our guide to submitting your startup to directories covers the prep.
- Launches you can show up for: as our launch platforms page advises, "pick only as many launches as you can show up for."
- A short directory list: our free directory shortlist asks five questions, then ranks directories with price, link type and approval time.
- Verified listings: after a few weeks, confirm each listing is live. The free dofollow link checker tells you "whether your link is followed and whether anything on the page cancels it"; our guide to checking dofollow backlinks covers manual checks.
- Monthly spot checks: ask ChatGPT (search on) and Perplexity your question list, and log whether you're cited, mentioned or missing, and who's cited instead. Expect movement: in the Toronto study, "the cited set changes over weeks".
What not to do
- Invent facts. An assistant may repeat them, and a buyer may check them. Google's generative AI content guidance says "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing."
- Mass-produce thin "GEO" pages. Google's spam policies count "Using generative AI tools or other similar tools to generate many pages without adding value for users" as scaled content abuse.
- Buy links or reviews. "Exchanging money for links, or posts that contain links" is link spam under the same policies.
- Trust citation guarantees. Anyone promising "we'll get you cited by ChatGPT", directory services included, is promising what OpenAI says isn't guaranteed.
FAQ
Can I pay to get cited by ChatGPT?
No. OpenAI's help page describes no paid route. The only step for site owners is allowing OAI-Searchbot and OpenAI's searchbot IPs, and even then "Placement is not guaranteed."
Does a Product Hunt launch get me cited by Perplexity?
Not directly. A launch page is a public page about you that a search could find, but nothing in Perplexity's help pages gives launch pages special weight. Product Hunt's links are ugc, a nofollow type; launch there for the audience.
Do directory backlinks help AI citations?
No direct effect is documented. Relevant listings help people and crawlers find you and add pages that name you. Choose them for fit, not as an AI tactic.
How do I check if ChatGPT cites my site?
Ask your buyers' questions with search on and open Sources; in Perplexity, read the numbered citations. Repeat monthly.
Should I write content only for AI assistants?
No. A clear page that answers a buyer's question serves buyers, Google and AI search at once. Google calls content "primarily made to attract visits from search engines" a warning sign.
Is GEO different from SEO?
Generative engine optimization, a term introduced by the 2023 GEO paper, means improving visibility in AI-generated answers. It overlaps with SEO, since crawlable, useful pages win in both, but the target differs: being one of a few sources picked for the assistant's own searches, which in the Toronto study didn't track Google rankings.
Rather hand it over?
Getting listed where your product fits is one part of being findable, and it's the part DirPilot does. We don't promise AI citations; nobody can. We send you the list of directories before we start, you can swap any of them out, we submit by hand, then check that each listing is live, links to you and is dofollow, and report which pages Google has indexed at day 30.
- Economy, $99: manual submission to 30+ directories, medium-DR free directories only.
- Business, $149: 60+ high-DR free and paid directories, a unique description for each, listing fees included.
- First, $199: 100+ high-DR free and paid directories, plus help on launch day.
Each plan is a one-time payment with no subscription, and work starts within one business day. If a directory rejects your product, we replace it with one of similar authority at no extra cost. Compare the plans, or work through the checklist yourself.