When I talk to business owners about ChatGPT, most of them assume it quotes the same big news sites they see on Google, or that it simply copies whatever ranks first. So they are surprised when I show them the real list. The site ChatGPT cites most is Reddit, the second is Wikipedia, and a little-known engineering explainer site sits in the top five, ahead of famous publishers.
This matters because a citation in ChatGPT is a recommendation in front of a very large audience. If you know which kinds of sites ChatGPT leans on, you know where your brand needs to be talked about. You stop guessing and start building a target list.
In this article I walk through the latest Ahrefs ranking of the most-cited websites in ChatGPT (the September 2026 update), explain what the numbers actually measure, group the sites by type, and show you which ones a normal business can influence. This is a source analysis, not a how-to. If you want the on-site steps, I cover those in my guide on how to get your brand mentioned in ChatGPT Search.
How Ahrefs built the ChatGPT most-cited list
The data comes from Ahrefs Brand Radar, a tool that sends large sets of search-style questions to AI assistants and records which websites each answer cites. For this list, Ahrefs looked at ChatGPT answers to a broad set of US queries across all topics, and counted how often each domain appeared as a cited source.
The key number is mention share. Ahrefs defines it like this:
“Mention share = a domain's citations as a percentage of the summed citations of the top sources.”
— Ahrefs, The 50 Most-Cited Websites in ChatGPT (September 2026)
Let me explain why that definition matters. The percentage is not a share of every citation ChatGPT made. It is a share of the citations that went to the top sources in the list. So when you read that Reddit has 16.8%, it means Reddit got 16.8% of the citations shared among the top domains, not 16.8% of all ChatGPT citations on the internet.
Here is a simple example with made-up numbers. Imagine the top 50 domains received 1,000,000 citations between them in a month. A 16.8% share would mean about 168,000 of those went to Reddit. Thousands of smaller sites outside the top 50 also got citations, but they are not in the denominator at all.
| What the list tells you | What it doesn't tell you |
|---|---|
| The relative weight of each top domain in ChatGPT citations (mention share) | The total number of prompts or citations behind the snapshot |
| How many distinct pages on each domain were cited | Results for any country other than the US |
| Each domain's Ahrefs Domain Rating (DR), a 0–100 link strength score | Splits by industry, topic or buyer intent |
| Rank movement compared with the previous month | The exact start and end dates of the data window |
One more thing you should know. Ahrefs says the page is refreshed every month by an AI agent:
“This article is generated and automatically kept up to date every month by Agent A, our AI agent.”
— Ahrefs, The 50 Most-Cited Websites in ChatGPT (September 2026)
That is not a reason to ignore the data. It is a reason to treat it as a monthly reading from a measuring tool, not as a finished research paper. I treat it as direction, not hard science.
The domains ChatGPT leaned on most in September 2026
I won't reproduce the whole top 50 here; you can see it on the Ahrefs page. Instead, here are the first twelve, with a label I added for what kind of source each one is. The labels are mine, not Ahrefs'.
| Rank | Domain | Mention share | Pages cited | Source type (my label) |
|---|---|---|---|---|
| 1 | reddit.com | 16.8% | 810,887 | Community / user-generated |
| 2 | en.wikipedia.org | 7.0% | 129,667 | Encyclopedia |
| 3 | consumerreports.org | 3.7% | 36,674 | Independent product testing |
| 4 | forbes.com | 3.1% | 36,505 | Publisher |
| 5 | engineerfix.com | 2.8% | 93,427 | Niche explainer site |
| 6 | legalclarity.org | 2.8% | 72,741 | Niche explainer site |
| 7 | healthline.com | 2.5% | 16,502 | Health publisher |
| 8 | youtube.com | 2.5% | 114,596 | Video / user-generated |
| 9 | pmc.ncbi.nlm.nih.gov | 2.4% | 69,801 | Government research archive |
| 10 | walmart.com | 2.3% | 144,045 | Retailer |
| 11 | tomsguide.com | 2.1% | 17,028 | Tech reviews |
| 12 | merriam-webster.com | 2.1% | 51,009 | Dictionary |
Source: Ahrefs, September 2026 snapshot, US queries. Source-type labels are my own.
Three things jump out at me:
- The top is concentrated, the tail is flat. Reddit alone has more share than the next two domains combined. After Wikipedia, no site goes above 3.7%.
- "Pages cited" and "share" tell different stories. Walmart had 144,045 distinct pages cited but only 2.3% share. Consumer Reports had about a quarter of that page count and a higher share. A few strong pages can beat a huge catalogue.
- Domain Rating is not destiny. engineerfix.com has a DR of 59 and legalclarity.org a DR of 76, yet both sit above Healthline (DR 92) and YouTube (DR 99).
Why Reddit sits on top of ChatGPT's sources
Reddit's lead is not an accident. In May 2024, OpenAI and Reddit announced a partnership. Fortune reported on May 16, 2024 that the agreement lets OpenAI show Reddit content in ChatGPT and use it to train its models:
“The deal allows OpenAI to display Reddit's content and train AI systems on its partner's data.”
— Fortune, Report on the OpenAI and Reddit partnership (May 16, 2024)
There is also a simpler reason. When people ask ChatGPT questions like "Is this brand any good?" or "What do people actually use for X?", the most direct evidence is other people's experience. Reddit threads are full of exactly that. They are long, specific and written in normal language.
For you, the takeaway is not "spam Reddit." Moderators and users remove promotional posts quickly, and a removed post helps nobody. The takeaway is that real customer conversation about your brand is now a visibility asset. I explain how forum discussions feed AI answers in my article on Reddit and AI search.
ChatGPT's surprising taste for explainers, dictionaries and research
This is where ChatGPT looks different from Google's AI features. A big slice of its top sources are sites that define things: what a word means, how a part works, what a legal term covers, what a study found.
- Explainer sites: engineerfix.com (engineering and mechanical topics) and legalclarity.org (plain-language legal explanations) together hold 5.6% share.
- Dictionaries: Merriam-Webster is in the top 12, even after falling nine places in the month.
- Medical research and health: the NIH's PMC archive and Healthline both sit in the top ten.
Why would this happen? My reading, and it is only a reading, is that ChatGPT gets a lot of "what is" and "how does" questions inside longer chats. A short, self-contained explanation on a page dedicated to one term is easy to quote. Sites with tens of thousands of these narrow pages give ChatGPT a matching page for almost any question.
A separate Ahrefs study of ChatGPT's top 1,000 most-cited pages points the same way. It found Wikipedia pages made up 29.7% of those pages, and that 28.3% of cited pages had zero organic keywords in Ahrefs' data. In other words, many of the pages ChatGPT likes are barely visible in Google.
“Nearly one-third of ChatGPT's citations point to pages with no traditional search visibility.”
— Ahrefs, 67% of ChatGPT’s Top 1,000 Citations Are Off-Limits to Marketers
For a business, this is good news. It means a clear explainer page on your own site, or a well-written definition you contributed to an industry glossary, can be cited even if it never reaches page one on Google.
Does ChatGPT just copy Google's top results?
No, and this is one of the most useful facts in the whole topic. Ahrefs compared ChatGPT and Perplexity citations against Google's results for about 3,311 short head terms. In that September 2025 study, only about 10% of ChatGPT's cited URLs appeared in Google's top 10, and about 21% appeared anywhere in the top 100. For Perplexity, about 65% of cited URLs were in Google's top 10.
“Across all three studies, the story is consistent: ChatGPT doesn’t follow Google’s sources, Perplexity does.”
— Ahrefs, ChatGPT Doesn’t Follow Google’s Sources, Perplexity Does
OpenAI's own help page explains part of the reason. ChatGPT rewrites a question into one or more shorter searches, sends them to search partners, and then applies its own ranking:
“ChatGPT ranks search results using multiple factors intended to help users find relevant, reliable information.”
— OpenAI Help Center, Searching the web with ChatGPT
The same page says some searches go to outside search providers, and that sites become eligible by allowing OpenAI's OAI-SearchBot crawler and not blocking its published IP addresses. When OpenAI launched ChatGPT search in October 2024, it also named content partners such as the Associated Press, Reuters, Hearst, Condé Nast and News Corp. If you want the details of which index sits behind ChatGPT, read my piece on which search index ChatGPT uses.
In brief: ranking well on Google helps, but it does not carry over to ChatGPT the way it does to Perplexity. You need to show up in the places ChatGPT checks on its own.
Which of ChatGPT's favorite sources can you actually influence?
Not every site on the list is a realistic target. You won't get your brand into a Merriam-Webster definition, and you shouldn't try to edit Wikipedia about yourself. But several source types are open to a normal business. Here is how I sort them.
| Source type | Examples from the ChatGPT list | Can you influence it? | Realistic action |
|---|---|---|---|
| Community and forums | Reddit, YouTube comments | Partly | Join relevant threads as yourself, answer questions honestly, encourage customers to share real experiences |
| Encyclopedias and dictionaries | Wikipedia, Merriam-Webster | Rarely | Only if you are genuinely notable; fix factual errors through the proper talk-page process |
| Independent testing and reviews | Consumer Reports, Tom's Guide | Indirectly | Make your product available for testing; you cannot buy a verdict |
| Publishers | Forbes, Healthline | Yes, with effort | Digital PR, expert commentary, original data journalists can quote |
| Niche explainer sites | engineerfix.com, legalclarity.org | Sometimes | Offer expert review or data; publish your own better explainers |
| Retailers and marketplaces | Walmart, Home Depot | Yes, if you sell there | Complete product listings, honest reviews, clear specifications |
| Government and research | PMC (NIH) | No | Cite this research accurately on your own pages instead |
The Ahrefs analysis of ChatGPT's top 1,000 pages classed only about 32% of citations as influenceable through outreach. That sounds low, but it still leaves a large, reachable slice. I go deeper into this triage in my guide to AI citation sources you can actually influence, and into the review side in how review data shapes AI recommendations.
How ChatGPT's sources compare with Google's AI features, Perplexity and Gemini
Ahrefs publishes the same style of list for other assistants, using the same method and the same September 2026 update. Putting them side by side shows how different ChatGPT's taste is.
| Assistant | No. 1 domain | Reddit share | YouTube share | Wikipedia share |
|---|---|---|---|---|
| ChatGPT | 16.8% | 2.5% | 7.0% | |
| Google AI Overviews | YouTube | 18.5% | 22.9% | 4.0% |
| Google AI Mode | 17.9% | 17.8% | 4.0% | |
| Perplexity | 21.6% | 20.8% | 6.3% | |
| Gemini | 28.5% | 14.6% | 8.8% |
Sources: Ahrefs lists for ChatGPT, AI Overviews, AI Mode, Perplexity and Gemini, September 2026, US queries.
The biggest difference is video. YouTube holds roughly a fifth of top-source citations in AI Overviews and Perplexity, but only 2.5% in ChatGPT. Social platforms like Facebook and Instagram are also big in Google's AI features and much smaller in ChatGPT. ChatGPT, on the other hand, gives more room to publishers, testing labs and explainer sites.
So if your plan for "AI visibility" is mostly YouTube videos, that may work well for Google's AI features and Perplexity but do much less for ChatGPT. You can read the full picture for each platform in my breakdowns of the sites Google AI Overviews cite most, the sites Google AI Mode cites most and the sites Perplexity cites most.
Niche differences hiding inside the ChatGPT average
The Ahrefs list is an all-topics average. That hides big differences by industry. Even within the top 50 you can see clusters:
- Cars: Edmunds sits in the top 15, alongside other automotive research sites further down the list.
- Health: Healthline and the NIH's PMC archive are both in the top ten.
- Home and DIY: Home Depot rose six places into the top 15 in the September update.
- Tech products: Tom's Guide and TechRadar both sit in the top 13.
If you run a dental clinic, a law firm or a B2B software company, your real list will look very different. A legal brand might find legalclarity.org and state bar sites matter more than Forbes. A SaaS brand might find software review platforms and developer forums matter more than Consumer Reports. The average tells you the types of sources ChatGPT trusts. Your own prompt research tells you the specific sites.
Turning the ChatGPT list into a mention target list
Here is how I would turn this data into a practical plan. I'll use a hypothetical example: a mid-sized US company that sells cordless power tools online and through big-box retailers.
- Write 30 to 50 real buyer questions. For example: "best cordless drill for home use," "is brand X better than brand Y," "what torque do I need to drive deck screws." Use the words customers use in emails and reviews.
- Run them in ChatGPT with search on, and log every cited domain. A spreadsheet with columns for prompt, cited domains and whether your brand was named is enough.
- Count the domains. In our example, imagine 40 prompts produce 220 citations. Suppose Reddit appears 38 times, Home Depot 21, an engineering explainer site 14, Tom's Guide 11 and your own site 3. That is your niche's version of the Ahrefs list.
- Tag each domain using the influence table above. Home Depot product pages and reviews (yes, you sell there), Reddit (partly), explainer site (sometimes), Tom's Guide (indirectly, via review units).
- Pick three to five targets for the quarter. In this example: improve your Home Depot listings and review volume, offer review units to two tool reviewers, and publish a clear torque explainer on your own site.
- Re-run the same prompts every month. Keep the prompts fixed so the comparison is fair. Track how often you are named and cited.
If your share of citations in this example went from 3 of 220 (about 1.4%) to 9 of 220 (about 4.1%) over two quarters, that is real progress you can report. I explain how to calculate this properly in my guide to share of voice in AI answers, and how to earn the publisher coverage side in digital PR for AI visibility.
What this ChatGPT citation data can't tell you
I like this data, but I want to be honest about its limits before you build a strategy on it.
- It is US only. If you sell in India, the UK or Australia, local publishers and marketplaces may matter more.
- Share is relative to the top sources. It overstates each domain's share of all citations.
- Sample size isn't published. We don't know how many prompts sit behind each monthly snapshot.
- The page has small inconsistencies. Its meta description names Amazon and Business Insider as top cited, but neither appears in the top 50 table. I trust the table over the summary.
- It moves monthly. Merriam-Webster dropped nine places in one update. A single month is a photo, not a film.
- Citations are not the same as recommendations. ChatGPT can name your brand without linking to you, and link to a page without naming you.
Questions marketers ask about ChatGPT's most-cited sites
Which website does ChatGPT cite the most?
In Ahrefs' September 2026 snapshot of US queries, Reddit was the most-cited domain in ChatGPT with 16.8% mention share, followed by Wikipedia at 7.0%. No other domain went above 4%.
Can I pay to be cited in ChatGPT?
Not through any program OpenAI describes. Its help page explains how ChatGPT ranks search results and says placement is not guaranteed; it describes no paid route into citations. What you can do is allow OAI-SearchBot to crawl your site and earn coverage on the kinds of third-party sites ChatGPT already trusts.
Why does ChatGPT cite small sites like engineerfix.com and legalclarity.org?
These sites answer narrow, definition-style questions in clear, self-contained explanations, and they have tens of thousands of such pages. ChatGPT seems to value that fit more than brand fame. Both have much lower Domain Rating than the big publishers around them in the list.
Is the ChatGPT most-cited list the same in every country?
We don't know. The Ahrefs data covers US queries only. Local publishers, marketplaces and review sites are likely to matter more in other markets, so run your own prompt checks in the countries you sell to.
How often does the list change?
Ahrefs refreshes it every month. In the September 2026 update, Merriam-Webster fell nine places while YouTube rose seven and the NIH's PMC rose eight, so a single month is a snapshot, not a permanent ranking.
Where I would start with ChatGPT citations this month
If you only do three things after reading this, make them these. First, check that OAI-SearchBot can crawl your site. Second, run 30 real buyer questions through ChatGPT and write down every site it cites. Third, pick the two or three influenceable sources that keep appearing and make a plan for each.
The Ahrefs list tells you ChatGPT trusts community discussion, reference content, testing labs, publishers and clear explainers. Your job is to make sure your brand appears truthfully in those places, and that your own site answers questions just as clearly.
If you would like help building your own citation target list, our team does this as part of our SEO services. Get in touch for a quote and we will start with the prompts your buyers actually ask.
Further reading on websites ChatGPT cites most
- The Websites Google AI Overviews Cite Most
- The Websites Google AI Mode Cites Most
- The Websites Perplexity Cites Most
Sources for this ChatGPT citation analysis
- The 50 Most-Cited Websites in ChatGPT (September 2026) (Ahrefs)
- ChatGPT Doesn’t Follow Google’s Sources, Perplexity Does (Ahrefs)
- 67% of ChatGPT’s Top 1,000 Citations Are Off-Limits to Marketers (Ahrefs)
- Searching the web with ChatGPT (OpenAI Help Center)
- Introducing ChatGPT search (OpenAI)
- OpenAI and Reddit partnership coverage (Fortune, May 16, 2024)
- The 50 Most-Cited Websites in Google AI Overviews (Ahrefs)
- The 50 Most-Cited Websites in Google AI Mode (Ahrefs)
- The 50 Most-Cited Websites in Perplexity (Ahrefs)
- The 50 Most-Cited Websites in Gemini (Ahrefs)
