Skip to contentSubmitium

12 min read

How to rank in ChatGPT: what 200 AI answers cite

We asked ChatGPT, Perplexity, Gemini and Google's AI Overviews 50 software questions and sorted the 1,770 sources they cited. What gets a product named.

A question to an AI assistant, "What is the best uptime monitoring tool for a small SaaS?", and its answer with the kinds of pages it cites: comparison pages, product pages, videos and discussions, directories and review sites. 200 answers, 1,770 sources.
On this page

Ask ChatGPT "What is the best uptime monitoring tool for a small SaaS?" and it names three or four products, with links to the pages it read. Those pages decide which products it names. So which pages are they?

On October 30, 2026, we asked four AI assistants the same 50 questions about software: ChatGPT, Perplexity, Gemini and the AI Overview at the top of Google's results. We collected 200 answers and sorted the 1,770 sources they cited. This is what we found:

  • Comparison pages are the first source. "Best X" lists, "X vs Y" pages and "alternatives to X" pages are 44% of the 1,770 sources. Each assistant cited at least one in 41 to 50 of its 50 answers.
  • Software companies write many of them. In a random sample of 60 cited comparison pages, 21 were on the site of a company that sells a product in that category.
  • ChatGPT reads the products' own pages. Home pages, pricing pages and docs of the products it names are 38% of its sources, in 42 of its 50 answers.
  • Google's AI Overviews lean on YouTube and Reddit. Videos and discussions are 34% of their sources. YouTube alone was cited 84 times, in 37 of the 50 overviews.
  • Directories and review sites are 5% of the sources, with G2 first. Tool directories also wrote 9 of the 60 comparison pages we checked.
  • The first page of Google is not required. Half of the AI Overviews' sources came from sites on Google's first page for the same search. For ChatGPT, 17%. In all, 897 different sites were cited, and 647 of them in a single answer.
Bar chart of the sources cited by each assistant. ChatGPT: 38% product pages, 33% comparison pages, 5% directories and review sites, 1% videos and discussions, 22% other. Perplexity: 63% comparison pages, 8% product pages, 5% directories, 1% discussions, 24% other. Gemini: 70% comparison pages, 23% product pages, 5% directories, 2% discussions. Google AI Overviews: 34% videos and discussions, 28% comparison pages, 18% product pages, 2% directories, 18% other.
1,770 sources cited in 200 answers to 50 software questions, October 30, 2026.

How we tested

We wrote 50 questions of the form "What is the best X?", on the tools a startup or a small SaaS buys: a CRM, an email API, a status page, a Next.js boilerplate, a directory submission service. The full list is at the end of this article.

  • ChatGPT and Gemini answered through their website, asked from the United States.
  • Perplexity answered through its API (Sonar), which can cite other pages than its website.
  • Google showed an AI Overview for all 50 searches "best X", which we read with its sources.

A page cited twice in the same answer counts once. We sorted the sources into five kinds:

  • Comparison pages: pages whose address or title says best, top, vs, alternatives, compared or review.
  • Product pages: pages on the site of a product the answer names: its home page, pricing, features, docs.
  • Videos and discussions: YouTube, Reddit, Quora, LinkedIn, Facebook, Medium and DEV.
  • Directories and review sites: G2, Capterra, Gartner, Product Hunt, AlternativeTo, and the 199 directories of our catalog.
  • Other: other articles, news, documentation, and the logo images ChatGPT shows next to product names.

The sorting is done by these rules, and we checked samples by hand. Answers change from one day to the next, and from one person to another. Read the figures as orders of magnitude, not as exact shares.

What AI assistants cite

The four assistants do not read the same pages.

  • Perplexity and Gemini answer mostly from comparison pages: 63% and 70% of their sources. Perplexity cited one in all 50 of its answers.
  • ChatGPT reads comparison pages to find the products, then the products' own pages to describe them: 287 links to home pages, pricing pages and docs. It answered 7 of the 50 questions without any source, from what it learned before. A product that is often written about has a better chance there.
  • Google's AI Overviews cite YouTube videos (84 links) and Reddit threads (21 links) next to comparison pages.

No single site decides. These are the sites cited in the most answers:

SiteAnswers that cited itWhat was cited
YouTube37Videos, all in Google's AI Overviews
Zapier26"Best apps" lists from its blog
G220Reviews and "best X" guides
Reddit17Discussion threads
Dupple16"Best X" guides
Guideflow16Its comparison and product pages
HubSpot13Product pages and blog articles
DEV13Articles by developers
Toolradar10"Best X" guides from a tool directory
SoftwareCompare9Comparison pages from a software directory
Product Hunt6Product pages

Out of 200 answers, the most cited site appears in 37. Most of the 897 sites were cited once: a small site with the right page can be one of the sources.

Who writes the comparison pages AI cites

We took 60 cited comparison pages at random and looked at who publishes each one.

Bar chart of who wrote 60 comparison pages cited by AI assistants: 21 by a company that sells a product in that category, 11 by media and review sites, 9 by tool directories and comparison sites, 19 by other companies, agencies and personal blogs.
60 comparison pages taken at random among the 684 cited, October 30, 2026.
  • 21 are on the site of a company that sells a product in that category. A password manager's blog ranking password managers, a product tour tool listing product tour tools, an uptime monitoring service comparing uptime monitoring services.
  • 11 are from media and review sites, such as TechRepublic, Tom's Guide or Security.org.
  • 9 are from tool directories and comparison sites, which list many tools and publish "best X" pages on top of their listings.
  • 19 are from other companies, agencies and personal blogs.

The uptime monitoring question shows how it works. Perplexity cited 6 pages from Hyperping's blog, among them "best uptime monitoring software", and named Hyperping in its answer. Google's AI Overview cited the same page and also named Hyperping. ChatGPT cited UptimeRobot's own comparison of uptime monitoring tools and its pricing page, and named UptimeRobot.

We asked about our own category too: "What is the best directory submission service for a new startup?". None of the four assistants named Submitium, which went live on October 21. The service named by three of them had published its own comparison of directory submission services, and ChatGPT cited it.

Do you need to rank on Google first?

For each question, we compared the sources with the first page of Google for the same search, "best X".

Bar chart of the share of cited sources that came from a site on Google's first page for the same search: Google AI Overviews 50%, Gemini 32%, Perplexity 24%, ChatGPT 17%.
Sources from a site that was on Google's first page for "best X", October 30, 2026.

Google's AI Overviews stay close to Google's results: half of their sources come from sites on the first page. The other assistants go further: 17% for ChatGPT. Google explains part of it in its page on AI features (opens in a new tab): its AI answers use "query fan-out", several related searches across subtopics. ChatGPT and Perplexity run their own searches, with their own crawlers.

A page that answers a narrow search, like "best uptime monitoring for a small SaaS", can be cited without ranking for "uptime monitoring". That is good news for a new site, which cannot win the broad searches in its first months. Our guide to SEO for a new website explains how to choose those searches.

How to get your product named by ChatGPT

1. Let the AI crawlers read your site

OpenAI and Perplexity each use their own crawler for search. OpenAI says OAI-SearchBot (opens in a new tab) is "used to surface websites in search results in ChatGPT's search features", and that robots.txt changes take about 24 hours to apply. Its other crawler, GPTBot, collects pages to train models: you can block it and still appear in ChatGPT's search. Perplexity's search crawler is PerplexityBot (opens in a new tab). If your robots.txt allows every crawler, you have nothing to add. If it lists crawlers one by one, add these two:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

If your site runs behind Cloudflare or another firewall, check that its bot settings don't block them. Google's AI Overviews use Google's normal index: Google says "there are no additional requirements" beyond being indexed and eligible for a snippet.

2. Publish your own comparison page

The comparison pages AI cites are mostly ordinary: a short answer at the top, a table, a few lines on each tool, the price, who it fits, the date. 21 of the 60 we checked were written by a company that sells one of the tools. If you write one:

  • List the real alternatives, including the ones that beat you on some points, and say on which.
  • Give prices, limits and plans as text, not only in images.
  • Put the year and the date of the last check on the page, and update it.
  • Write one page per question buyers ask: "best X for small teams", "X vs Y", "alternatives to Y".

A page that only praises your own product is an ad, and readers see it. The comparison pages cited in our study compare.

3. Make your product pages easy to quote

ChatGPT used the products' own pages in 42 of its 50 answers, mostly to describe them. Say in plain text, on your home page and pricing page:

  • What the product does, in one sentence.
  • Who it is for, and who it is not for.
  • The price of each plan, and what is included.
  • The limits that matter: free plan, number of users, integrations.

4. Get into other people's comparison pages

The other 39 of our 60 comparison pages were written by media, tool directories, agencies and bloggers. Find the ones cited for your category: ask the assistants your buyers' questions and open the sources. Then:

  • Write to the author with facts: what your product does, its price, a screenshot. Many of these lists carry the year in their title and are updated.
  • Get listed on the tool directories and comparison sites of your category. 9 of the 60 comparison pages in our sample were published by such sites, next to their listings. Our articles on directory submission and Product Hunt alternatives list 199 of them, with what each one asks.

5. For Google's AI Overviews: videos and discussions

37 of the 50 AI Overviews cited YouTube. A short demo of your product, or a video comparing tools in your category, can be one of those sources. Reddit was cited in 16 overviews: answer the questions people ask about your category where they ask them, with your name and your role, and within each community's rules.

6. Skip the shortcuts

  • llms.txt and other AI files. Google says "You don't need to create new machine readable files, AI text files, or markup to appear in these features."
  • Pages written only for AI. The pages cited in our study are written for people choosing a tool. A page that answers that person is the one an assistant can use.

How to check if AI assistants name your product

  • Ask them. Each month, ask ChatGPT, Perplexity, Gemini and Google the five to ten questions your buyers ask. Note which products they name and which pages they cite. Answers change: one check tells you little, the trend over months tells you more.
  • Read your analytics. Visits from chatgpt.com, perplexity.ai and gemini.google.com show up as referrals. Links from ChatGPT often carry utm_source=chatgpt.com.
  • Open the cited pages. If a competitor's comparison page is cited for your category and you are not on it, you know which page to write and which authors to contact.

Where Submitium fits

Submitium does one part of this: listings on directories. Directories and review sites were 5% of the sources in this study, and tool directories wrote 9 of the 60 comparison pages we checked. A listing also gives your site a link, which helps it rank for the narrow searches AI assistants run.

We submit your product by hand to directories and launch platforms that give a free dofollow link, keep their badges in one badge wall on your site, and check the link on each live listing six times over 90 days. A directory costs $0.99, less from the 51st, and a first order starts at 30 directories. See the pricing and the full list of directories.

Questions

How do I rank in ChatGPT?

Let OAI-SearchBot read your site, publish a comparison page for your category with real alternatives and prices, make your product pages say plainly what the product does and costs, and get into the comparison pages others write. In our study, comparison pages were 33% of ChatGPT's sources and the products' own pages 38%.

How does ChatGPT choose which products to recommend?

It searches the web, reads the pages it finds, and names products from them. In our study, it cited at least one comparison page in 41 of 50 answers. For 7 questions, it answered without any source, from what it learned during training.

Does ChatGPT use Google results?

It runs its own searches, with its own crawler, OAI-SearchBot. In our study, 17% of ChatGPT's sources came from a site on Google's first page for the same search. For Google's own AI Overviews, it was 50%.

What is LLM SEO?

The work of getting a product named in the answers of AI assistants (large language models). It is mostly SEO aimed at the pages they read: comparison pages, product pages, and the sites that mention you.

Not in Google's AI features: Google says no AI text file or special markup is needed to appear in AI Overviews or AI Mode.

How long does it take to appear in ChatGPT's answers?

For answers with sources, once its crawler has read your pages and they answer the question: OpenAI says robots.txt changes take about 24 hours to apply to its search. For answers without sources, it depends on what the model learned before its release, which changes only with a new model.

Directly, a little: directories and review sites were 5% of the sources in our study, G2 first. Indirectly, more: tool directories publish many of the "best X" pages AI assistants cite, and their links help your pages rank for the searches assistants run.

The 50 questions

Each question was "What is the best …?", for: uptime monitoring tool for a small SaaS, free CRM for a small startup, email marketing tool for a SaaS startup, invoicing software for freelancers, project management tool for a small remote team, note-taking app for developers, screen recording tool for product demos, appointment scheduling tool for consultants, form builder for collecting leads, website builder for a startup landing page, privacy-friendly web analytics tool, customer feedback widget for a SaaS app, status page tool for a SaaS, password manager for small teams, AI writing assistant for blog posts, AI image generator for marketing visuals, transcription tool for meetings, newsletter platform for creators, help desk software for a small support team, live chat widget for a website, SEO tool for a small business, backlink checker, social media scheduling tool, link-in-bio tool, time tracking app for freelancers, e-signature tool for small businesses, online survey tool, product analytics tool for a startup, error tracking tool for a web app, feature flag tool, changelog tool for a SaaS, documentation tool for an API, waitlist tool for a product launch, affiliate program software for SaaS, directory submission service for a new startup, Product Hunt alternative to launch a startup, Next.js SaaS boilerplate, headless CMS for a Next.js site, authentication provider for a SaaS app, payment provider for a SaaS with subscriptions, transactional email API, background jobs tool for TypeScript, vector database for an AI app, AI agent framework, hosting platform for a Next.js app, managed Postgres database for a small SaaS, product tour tool for user onboarding, A/B testing tool for a website, proposal software for agencies, and AI chatbot for customer support on a website.

More from the blog

9 min read

llms.txt in 2026: who has one, and who reads it

We checked /llms.txt on 3,202 websites and read 30 days of logs of a site that has one. 36% of new websites have one. GPTBot never asked for it.