AI search15 min read

How to Get Cited by ChatGPT: What the 2025 and 2026 Studies Show

How to get cited by ChatGPT and rank in ChatGPT search: what OpenAI documents, what studies of 15,000 to 680 million citations measured, and what B2B teams can change.

Get a free audit

Chapters
  1. 01The short version
  2. 02What "cited" and "named" mean in a ChatGPT answer
  3. 03What OpenAI documents about how sources are picked
  4. 04What independent studies measured
  5. 05The role of Reddit, review sites and listicles
  6. 06What a company can actually change
  7. 07What does not work, or is unproven
  8. 08What this site's own ChatGPT lists show
  9. 09Methodology
  10. 10Sources

Short answer

To get cited by ChatGPT, let OpenAI's OAI-SearchBot crawl your site, then give ChatGPT a page worth retrieving: a direct answer, original data and a presence on the third-party pages it also cites. No study reviewed here proves any tactic causes a citation, and OpenAI promises no placement.

Key facts

  • llms.txt.: SE Ranking analysed nearly 300,000 domains in a study dated 15 October 2026. It found no correlation between llms.txt and AI citation frequency, and its model got more accurate without the variable. 10.13% of domains had one. Correlation across domains cannot rule out an effect for a given site, but there is no evidence of benefit. (SE Ranking, fetched 21 Nov 2026, Nov 2026)
  • In Ahrefs' study of 15,000 prompts, only 8.0% of ChatGPT in-text citations were in Google's top 10 for the original prompt, so Google rank is a weak guide. (Sources and method in this article, 21 Nov 2026)
  • They can differ: in Ahrefs' 2026 experiment of 34 self-promotional pages, ChatGPT cited conference pages but never named the author's conference for the target query. (Sources and method in this article, 2026)
  • Profound found Reddit was 1.8% of ChatGPT citations across 680 million citations (August 2024 to June 2025). (Sources and method in this article, August 2024)

By Shruti Kaushik, reviewed 21 November 2026. Markdown version

To get cited by ChatGPT, let OpenAI's OAI-SearchBot crawl your site, then give ChatGPT a page worth retrieving: a direct answer, original data and a presence on the third-party pages it also cites. No study reviewed here proves any tactic causes a citation, and OpenAI promises no placement.

I read OpenAI's own pages and eleven independent studies published between June 2025 and October 2026, then compared them with 17 ChatGPT ranked lists I recorded on 20 November 2026. The studies disagree more than most advice admits.

To track whether any of this changes your visibility, see 12 AI visibility tools compared.

Looking for brand mentions rather than links? Read how to get your brand named in ChatGPT answers, the companion guide on being named by ChatGPT, Perplexity and Google AI.

#The short version

The short version
Question What the evidence says Confidence
Can ChatGPT cite my site if I block OAI-SearchBot? OpenAI: no, opted-out sites are not shown in search answers High (OpenAI's own page)
Does ranking in Google or Bing get me cited? Weakly. 8.0% of ChatGPT in-text citations were in Google's top 10 (Ahrefs, 15,000 prompts) Medium
Which sites does ChatGPT cite most? Wikipedia led at 7.8% of citations (Profound, 680M citations). Reddit share swung from about 60% to 10% of responses in six weeks (Semrush) Medium, unstable
Do vendor pages or listicles get cited? Studies conflict: 57% vendor pages (Onely, 210 citations) against 3.6% self-promotional listicles (Peec, 232,000 citations). They measure different things Low
Does branding beat backlinks? Branded web mentions correlated at 0.664 and Domain Rating at 0.266 (Ahrefs, 75,000 brands). Correlation only Medium
Does fresher content get cited more? Yes, on average: ChatGPT-cited URLs averaged 958 days since publication against 1,432 for organic results (Ahrefs, 16.975M URLs) Medium
Does llms.txt help? No correlation found (SE Ranking, nearly 300,000 domains) Medium
Can I guarantee a citation? No. Nobody has published a method that does High

#What "cited" and "named" mean in a ChatGPT answer

OpenAI's ChatGPT search help page states that responses that use search "will contain inline citations," and that clicking the Sources button at the end of a response shows "all the cited sources, as well as other relevant links" (OpenAI Help Center, fetched 21 Nov 2026). OpenAI's launch post says the same in other words: chats include links to sources such as news articles and blog posts, opened from a Sources button (OpenAI, 31 Oct 2024).

That gives two separate outcomes:

  • Cited: a page of yours appears as a link, inline or in the Sources list. The reader can click through.
  • Named: your brand appears in the answer text, for example as item 4 in a ranked list. The reader may never see a link.

OpenAI does not define "named", so this split is mine. It matters because the studies below do not all measure the same thing: a page can be cited without your brand being recommended, and a brand can be named with no link to its site.

Citations only exist where ChatGPT chooses to search, so buyer questions with "best", "2026" or "vs" in them are the ones worth tracking.

#What OpenAI documents about how sources are picked

OpenAI is brief. This is what it says that bears on citations, and what it leaves out.

It says:

It does not say (on the pages I fetched):

  • Which search providers supply the index, beyond naming Bing as an example of a provider that may receive queries.
  • How retrieved pages are ranked or chosen for citation. An earlier guide on this site quoted a line about ranking "using multiple factors" and placement not being guaranteed. That line did not come back when I fetched the help page on 21 November 2026, so I do not rely on it here. The absence of any ranking documentation is itself the finding.
  • Absolute dates: the help page showed only "Updated over a month ago" and the FAQ "Updated: 23 days ago."
  • Any guarantee of inclusion, ranking or traffic.

OpenAI documents the gate (crawler access) and not the ranking. Everything after the gate comes from outside studies, which observe outputs, not OpenAI's system.

#What independent studies measured

Study 1: Google rank is a weak guide

Ahrefs ran 15,000 long-tail prompts, checked Google and Bing results for each, then asked AI assistants the same questions and recorded where cited URLs ranked. ChatGPT data was collected in early July 2025. For ChatGPT in-text citations, 8.0% were in Google's top 10 and 8.1% in Bing's top 10. Across assistants the average Google top 10 overlap was about 12%, and roughly 80% of AI citations did not rank anywhere in Google for the original query. Perplexity was higher at 28.6%. (Ahrefs, 11 Aug 2025, fetched 21 Nov 2026).

Limit: the prompts were long-tail, and short commercial queries were not tested.

Study 2: Which domains get cited

Two large studies disagree on size and on the metric.

  • Profound analysed 680 million citations across ChatGPT, Google AI Overviews and Perplexity from August 2024 to June 2025. For ChatGPT, Wikipedia was 7.8% of citations, Reddit 1.8%, Forbes 1.1% and G2 1.1% (Profound, 5 Jun 2025, fetched 21 Nov 2026).
  • Semrush tracked about 230,000 prompts weekly for 13 weeks (14 July to 12 October 2025), over 100 million citations, across ChatGPT search, Google AI Mode and Perplexity. In ChatGPT, Reddit and Wikipedia were far ahead of other domains before September. Reddit fell from close to 60% of responses in early August to about 10% by mid-September, and Wikipedia from roughly 55% to under 20%. Both stayed the two most-cited domains in October (Semrush, 10 Nov 2025, fetched 21 Nov 2026).

Where they conflict: Reddit is 1.8% of ChatGPT citations in one and "close to 60% of responses" in the other. The first counts share of all citations, the second the share of responses that include the domain, so neither means "Reddit is cited X% of the time." Semrush's own caveats: it shows how quickly patterns shift, not why, some figures are approximate, and Semrush sells AI visibility tools (Semrush, 10 Nov 2025).

Study 3: Brand mentions beat backlinks, as a correlation

Ahrefs studied 75,000 brands and, as reported on its page, found branded web mentions correlated with ChatGPT visibility at 0.664 against 0.266 for Domain Rating (Spearman). It also reports ChatGPT showed the weakest correlations of the three assistants for most traditional authority signals. Ahrefs states that correlation is not causation, that its brand selection method (Domain Rating above 40, top keyword with at least 800 monthly searches) is "not a perfect calculus," and that its interpretations are possibilities (Ahrefs, 12 Dec 2025, fetched 21 Nov 2026).

Study 4: Freshness

Ahrefs analysed about 16.975 million cited URLs. Across ChatGPT, Perplexity, Gemini and Copilot, cited URLs averaged 1,064 days since publication against 1,432 for organic results, about 25.7% fresher. ChatGPT showed the strongest freshness preference: its in-text citations averaged 958 days since publication and its references 1,023. The same study says the average cited page is still about 2.9 years old, and that Google AI Overviews cited slightly older content than organic results (Ahrefs, 28 Jul 2025, fetched 21 Nov 2026). The author warns against changing publish dates without substantive edits.

Study 5: Do ChatGPT citations go to vendors or to listicles?

This is where the 2026 studies split.

What independent studies measured
Study Date Sample ChatGPT result
Onely, "best agencies for X" Oct 2026 31 prompts, 210 ChatGPT cited pages 57% vendor own sites, 22% listicles
Peec AI, software review prompts Mar 2026 13,000 listicles, 232,000 citations, six platforms Self-promotional listicles about 3.6% of ChatGPT citations, against about 10.3% in Google AI Mode and 10.4% in Perplexity
DerivateX (via Search Labs) May 2026 40 B2B SaaS categories, 233 recommendations ChatGPT cited a recommended tool's own site 11.6% of the time
BeVisibleIQ (via Search Labs) Apr 2026 75 queries, 2,391 classified citations 74.6% of ChatGPT product-query citations went to the vendor's own site

Sources: Onely, 30 Oct 2026 (updated 12 Nov 2026); Peec AI, 3 Mar 2026; Search Labs summary of DerivateX and BeVisibleIQ. All fetched 21 Nov 2026. I did not open the DerivateX or BeVisibleIQ originals, so those two rows are second-hand.

The headline numbers appear to clash, 57% and 74.6% of citations going to vendor sites against 11.6%. Search Labs itself says the two B2B studies differ in scope, one measuring citations on product queries and the other measuring citations behind named recommendations, and does not resolve the gap. The Peec 3.6% is a narrower thing again: only self-promotional listicles, where a publisher lists its own product. Onely tested one query type, "best agencies," and is itself an agency that sells GEO services, so treat its framing with care.

What these studies jointly support: for ChatGPT in 2026, vendor-owned pages are a real share of citations on product queries, which is different from Perplexity and Google where listicles dominated in Onely's sample (70% for Perplexity, 59% for AI Mode). What they do not support: a fixed percentage for any B2B category.

Study 6: An experiment with self-promotional pages

Ahrefs published 34 self-promotional pages on five domains and tracked 9,886 answers from 7 February to 31 May 2026. ChatGPT never named the author's conference for "best SEO conferences" and was the most consistent about which conference pages it cited. Copilot did name it, rising to 65% of answers by May. Citations were temporary: about a quarter of cited pages were cited once and never again, and pages were cited on about one in three eligible days between first and last citation. The author calls it a small experiment on his own brand and says it may not generalise (Ahrefs, 6 Jul 2026, fetched 21 Nov 2026). The takeaway for ChatGPT specifically: publishing pages about yourself did not get you named.

#The role of Reddit, review sites and listicles

Two things are documented. OpenAI announced on 16 May 2024 that it would access Reddit's Data API, which provides "real-time, structured, and unique content from Reddit," and bring Reddit content to ChatGPT (OpenAI, 16 May 2024, fetched 21 Nov 2026). And Profound's list of ChatGPT's most cited domains contains Reddit, Forbes, G2, TechRadar and NerdWallet alongside Wikipedia (Profound, 5 Jun 2025). So third-party review and discussion sites are cited. G2, a review platform, was 1.1% of ChatGPT citations in the Profound data.

What is disputed is how stable that is. In October 2025, G2's Kevin Indig wrote that ChatGPT was citing Reddit far less, and proposed that Google's removal of the num=100 parameter around 10 September cut the search depth that third-party data providers could reach. He labels this "I suspect" and states that no change specific to Reddit has been confirmed by OpenAI or Google. G2 has commercial ties to some data sources it cites (G2 Learn, 2 Oct 2025, fetched 21 Nov 2026). Semrush adds that it does not think that was the only cause (Semrush, 10 Nov 2025).

The defensible conclusion: appearing in third-party discussion, review and list content is consistent with Ahrefs' brand-mention correlation, but any one site's cited share moves within weeks. Do not build a plan on one platform.

#What a company can actually change

Ordered by how well the evidence supports it.

1. Let OAI-SearchBot in (documented). Check robots.txt, your CDN's bot rules and any firewall setting. Remember the separation: blocking GPTBot is about training, and blocking OAI-SearchBot removes you from search answers (OpenAI crawler documentation). Allow 24 hours after a change. If you want a page out of results while keeping crawl access, OpenAI's FAQ says to use a noindex meta tag, which the crawler must be able to fetch to read (OpenAI FAQ).

2. Be present on the pages that are cited (supported by correlation). Ahrefs' 0.664 branded-mention correlation and the Profound and Semrush domain lists both point toward third-party presence. Concretely: find the pages ChatGPT cites for your buyers' questions, and work out why you are absent. It may be a missing review profile, a listicle that omits you, or a thread where your category is discussed without you. Reddit and review sites need real participation, and no study I fetched shows manufactured posts working.

3. Publish pages that answer the question directly (plausible, not proven for ChatGPT). No study I fetched isolates "answer-first" formatting as a cause of ChatGPT citation. What the data does show is that vendor pages made up 57% of ChatGPT citations in Onely's sample and 74.6% in the BeVisibleIQ figure reported by Search Labs, so an owned page can be cited when it matches the query. Putting the answer in the first lines is cheap and helps human readers, but it is not a proven factor.

4. Publish original data (plausible, indirect). No fetched study measures original research as a citation factor. The indirect case is that numbers nobody else has give other pages something to reference, which creates the third-party mentions in step 2. This is an inference.

5. Keep important pages current (supported on average). Ahrefs found ChatGPT-cited URLs were younger than organic results by roughly 393 to 458 days on publication age. Real updates, not changed dates.

6. Measure with your own prompts, repeatedly. The Ahrefs experiment found citations were intermittent, about one in three eligible days. A single check proves little. Because citations were intermittent, run 10 to 15 buyer questions monthly and compare trends, not single runs. The free AI visibility check tests whether ChatGPT and other engines name your company for your buyer questions, and you can compare the result month to month. If you want it done by a person across your category, there is a human-run AI visibility audit.

#What does not work, or is unproven

  • llms.txt. SE Ranking analysed nearly 300,000 domains in a study dated 15 October 2026. It found no correlation between llms.txt and AI citation frequency, and its model got more accurate without the variable. 10.13% of domains had one. Correlation across domains cannot rule out an effect for a given site, but there is no evidence of benefit (SE Ranking, fetched 21 Nov 2026).
  • Ranking first in Google as a plan. With 8.0% overlap in the Ahrefs data, it will not get you cited on its own.
  • Self-promotional pages as a shortcut. Ahrefs' experiment found ChatGPT cited the pages but never named the author's brand for the target query. Peec found self-promotional listicles at about 3.6% of ChatGPT citations. Neither shows a repeatable win.
  • Guaranteed-citation services. No fetched source shows a method with a guarantee. Citations were temporary even for deliberate pages in the Ahrefs experiment. If you are weighing agencies that sell this work, GEO agencies compared sets out what 13 firms publish about their own methods.

#What this site's own ChatGPT lists show

On 20 November 2026 I recorded ChatGPT's ranked answer to "best [category] 2026" for 17 B2B software categories, one answer per category, and published them at /ai-recommends. The data file is src/content/data/ai-recommends-chatgpt.ts in this site's repository. This is a small sample and I will not over read it.

What the 17 lists contain:

  • 204 ranked vendor slots in total: 9 lists of 10 vendors, 2 lists of 12 and 6 lists of 15. ChatGPT chose the length each time. A few queries asked for a ranked list explicitly and others did not.
  • 201 distinct vendor names. Only three names appear in more than one list, Gong, Avoma and Salesloft, and all three sit in neighbouring sales software categories (revenue intelligence, conversation intelligence and sales engagement). Gong and Avoma appear in both revenue intelligence and conversation intelligence, and Salesloft in sales engagement and conversation intelligence.
  • 16 distinct first-place vendors across 17 lists. Gong was first in two lists (revenue intelligence and conversation intelligence). Every other category had a different leader.
  • 0 recorded own-site citations. None of the 204 slots records the vendor's own website as a cited source.
  • 0 recorded cited sources overall. This is a gap in my recording, not a finding. I did not capture the Sources list for these answers, so I cannot say how many named sources or which. Do not read the zeros as "ChatGPT cited nothing."

What I can conclude: ChatGPT named 10 to 15 vendors per category, and leadership was not concentrated in a few vendors. What I cannot conclude: whether the same list appears tomorrow. Semrush's 13-week data and the Ahrefs experiment both show patterns moving week to week, so one dated answer per category is a snapshot. The pages carry the date for that reason.

For the wider tool set grouped by job, see GEO tools and AEO tools.

#Methodology

I fetched each OpenAI page and each study on 21 November 2026 using a page-fetch tool that returns extracted text and summaries, not raw HTML. Where a page gave only a relative date ("Updated over a month ago"), I say so. DerivateX and BeVisibleIQ results come from Search Labs' write-up and were not checked against originals; the Onely table was checked against Onely's own post. I treated vendor-published studies (Ahrefs, Profound, Semrush, Onely, Peec, SE Ranking) as measurements with commercial interests, and labelled correlations as correlations.

The own-data section is computed from the 17 ChatGPT lists in the site's data file, recorded on 20 November 2026, one answer per category, with no repeat runs, so it cannot show variation between runs. Time of day and cited sources were not recorded.

Not known, as of 21 November 2026: how OpenAI ranks retrieved pages, which search providers supply the index, whether any on-page formatting causes citation, and how long a citation lasts after a page is published. Nothing here is a guarantee.

#Sources

Quick questions

How do I get my website cited by ChatGPT?

Let OAI-SearchBot crawl the site, because OpenAI says sites opted out of it will not be shown in ChatGPT search answers. Then make the page worth retrieving: a direct answer near the top, original data, and a presence on the third-party pages ChatGPT also cites. OpenAI documents no way to guarantee a citation, and no study reviewed here shows one.

How do I rank in ChatGPT search?

There is no published rank order to climb. OpenAI documents that search results carry inline citations and a Sources list, and that any public site can appear. In Ahrefs' study of 15,000 prompts, only 8.0% of ChatGPT in-text citations were in Google's top 10 for the original prompt, so Google rank is a weak guide.

What is the difference between being cited and being named in ChatGPT?

A citation is a link to a source page, shown inline or in the Sources list. A name is your brand appearing in the answer text. They can differ: in Ahrefs' 2026 experiment of 34 self-promotional pages, ChatGPT cited conference pages but never named the author's conference for the target query.

Does Reddit still matter for ChatGPT citations?

It mattered, but the share is unstable. Profound found Reddit was 1.8% of ChatGPT citations across 680 million citations (August 2024 to June 2025). Semrush found Reddit in close to 60% of ChatGPT responses in early August 2025, falling to about 10% by mid-September. Neither study says why, so treat any single Reddit figure as a snapshot.

Does llms.txt help ChatGPT cite my site?

No evidence says it does. SE Ranking analysed nearly 300,000 domains and found no correlation between having an llms.txt file and AI citation frequency, and its model got more accurate when the variable was removed. 10.13% of the domains had the file.

Get a free AI visibility and LinkedIn content audit

See how AI assistants describe your company today, plus a review of your LinkedIn content. No call needed, and no newsletter unless you tick the box.

Prefer to talk it through?

Book a 30-minute call

About the author

runs Wise Mango, a LinkedIn ghostwriting studio for SaaS and AI founders, and Klype (klype.ai), an AI platform that writes in a founder’s voice. She wrote The LinkedIn Blackbook and was named one of Forbes Game-Changing Leaders in 2025. Every figure on this page links to its source.