What Sources Do ChatGPT, Gemini and AI Overviews Cite? 4M Citations Analyzed

What Sources Do ChatGPT, Gemini and AI Overviews Cite? 4M Citations Analyzed

What Sources Do ChatGPT, Gemini and AI Overviews Cite? 4M Citations Analyzed

Escrito por:

SEO & GEO Specialist @ aiclicks.io

Revisado por:

Pragati Gupta

Content Marketing Manager @ aiclicks.io

Última actualización:

Verificado por expertos

No headings found on page

Llegue a millones de consumidores que están usando IA para descubrir nuevos productos y marcas

Llegue a millones de consumidores que están usando IA para descubrir nuevos productos y marcas

Key Takeaways

  • 84.4% of cited URLs appear in one engine only; 3.4% are cited by ChatGPT, Gemini and AI Overviews together.

  • Sites overlap more than pages: domains cited by all three engines are 11.6% of domains but take 59.4% of citations.

  • Gemini and AI Overviews agree most: 17.6% URL overlap, against 9.4% and 8.4% for the two ChatGPT pairs.

  • Listicles take 46.3% of all citations, and 70.8% of Gemini's.

  • AI Overviews accounts for 97.2% of all YouTube citations in the data.

  • ChatGPT gives 1,658 citations per 100 answers, 5 to 7 times more than Gemini (315) or AI Overviews (244).

On buying-intent prompts, ChatGPT, Gemini and Google AI Overviews mostly cite different pages: 84.4% of the 527,656 URLs we tracked were cited by one engine only. They agree far more on sites. The 11.6% of domains that all three engines cite take 59.4% of all citations. Listicles are the most cited page type on every engine, and AI Overviews is the only one that leans on YouTube, forums and social posts.

Stacked bars: 84.4% of cited pages come from one engine, but the 11.6% of sites all three engines cite take 59.4% of citations

Figure 1. Pages, sites and citations by how many engines cite them. One engine cites 84.4% of pages; sites cited by all three engines take 59.4% of all citations.

Show the data

Measured as

Cited by 1 engine

2 engines

All 3 engines

Total

Cited pages (URLs)

84.4%

12.2%

3.4%

527,656

Cited sites (domains)

71.3%

17.1%

11.6%

127,336

Citations, by the site cited

24.4%

16.2%

59.4%

4,018,882

About this data. 833 product categories tracked in AIclicks · about 25,000 non-branded buying-intent prompts · ChatGPT, Gemini and Google AI Overviews · 535,560 answers · collected 20–29 October 2026. The full method is in Methodology and limits.

The best-known studies of AI sources put Wikipedia and Reddit at the top of ChatGPT's list (Search Engine Roundtable covered one of them). We wanted the answer for the questions that make money: "best CRM for startups", "payroll software for small businesses", "HR software pricing". So we ran the same buying-intent prompts on three engines in the same week and labelled every page they cited. Two common assumptions don't survive the data. AI search behaves like three channels, not one, and on these prompts Reddit and Wikipedia are minor sources.

How did we run the study?

We tracked 833 product categories in AIclicks, each with 30 non-branded buying-intent prompts such as "best X", "X for small businesses" and "X pricing". Every prompt ran daily on ChatGPT, Gemini and Google AI Overviews between 20 and 29 October 2026, for 6 or 8 days per category. That produced 535,560 answers and 4,018,882 citations to 527,656 unique URLs on 127,336 domains.

Each category is one AIclicks tracking project, from airlines and ISP proxies to period care brands. The prompts were written for the category and never name a brand, so every citation is a page the engine chose on its own.

A citation is one URL cited in one answer: the same URL cited in five answers counts five times. A source is one unique URL, and a domain is the registrable domain behind it, so blog.example.com counts as example.com (we used the Public Suffix List to decide). Every URL also carries one of AIclicks' 14 page-type labels, such as listicle, comparison, review, forum or homepage. AIclicks files blogs and news under one label, "Generic Article", which we call articles throughout.

The ChatGPT answers here are answers with web sources, the kind ChatGPT search shows as links. We collected all three engines as AIclicks sees them in their live interfaces. What the method can and can't tell you is at the end of the post.

Do ChatGPT, Gemini and AI Overviews cite the same sources?

Mostly not. Of 527,656 cited URLs, 84.4% were cited by one engine only, 12.2% by two and 3.4% (17,871 URLs) by all three. The shared minority carries far more weight than its size: URLs cited by two or more engines are 15.6% of sources but collect 37.6% of citations, and the all-three group alone takes 14.2%.

ChatGPT owns most of the private pages. 63.4% of all cited URLs are cited by ChatGPT and nobody else, and they collect 53.8% of all citations. AI Overviews has the next-largest set of its own (14.9% of URLs, 4.5% of citations), then Gemini (6.2% and 4.1%). Every combination that involves two or more engines punches above its size, which tells you something about the pages in it: when two engines independently pick the same page, they tend to pick it again and again.

Bar chart of the seven engine combinations: pages all three engines cite are 3.4% of sources but 14.2% of citations

Figure 2. Share of cited pages and share of citations for each combination of engines. Pages all three engines cite are 3.4% of sources and 14.2% of citations.

Show the data

Cited by

Pages

Share of pages

Citations

Share of citations

ChatGPT only

334,334

63.4%

2,161,575

53.8%

AI Overviews only

78,446

14.9%

180,990

4.5%

Gemini only

32,703

6.2%

166,506

4.1%

ChatGPT + AI Overviews

28,455

5.4%

382,236

9.5%

ChatGPT + Gemini

19,740

3.7%

381,399

9.5%

All three

17,871

3.4%

571,903

14.2%

Gemini + AI Overviews

16,107

3.1%

174,273

4.3%

Do they agree more on sites than on pages?

Yes, by a wide margin. 28.7% of cited domains are cited by two or more engines, against 15.6% of URLs, and 11.6% of domains are cited by all three, against 3.4% of URLs. Those 14,745 domains collect 59.4% of all citations (the bottom bar in Figure 1).

The engines go to the same trusted sites and pick different articles from them. A software review site might appear in all three engines' answers for "best help desk software" while each engine links a different roundup on that site.

One caveat pulls URL-level overlap down slightly. AIclicks stores some pages under two URL strings, a YouTube link with and without a language parameter for example, and we counted each string as its own URL.

Which AI engines agree with each other most?

Gemini and Google AI Overviews. Their URL overlap is 17.6%, about twice that of ChatGPT and AI Overviews (9.4%) or ChatGPT and Gemini (8.4%). The same order holds at domain level (34.3% against 19.9% and 19.2%) and inside single prompts (11.3% against 4.9% for both ChatGPT pairs).

We measured overlap as the sources both engines cite divided by the sources either engine cites, known as the Jaccard index. We fixed that measure before the data came in, and it matters which one you pick. ChatGPT cites so many URLs (400,400 of them) that its set contains 43.5% of Gemini's URLs and 32.9% of AI Overviews'. If you measure overlap as a share of the smaller engine's set instead, ChatGPT and Gemini come out first (43.5% against 39.3% for the Google pair). Jaccard doesn't reward an engine for simply citing everything, which is why we chose it.

Both Google engines draw on Google's index, and Google's own AI features documentation describes how AI Overviews links to pages from it. That is the obvious explanation for their overlap. Our data shows the overlap; it doesn't prove the cause, so read the shared index as the likely reason rather than a finding.

Bar chart: Gemini and AI Overviews share 17.6% of pages and 34.3% of sites, about twice either ChatGPT pair

Figure 3. Overlap between each pair of engines, for pages and for sites. Gemini and AI Overviews share 17.6% of pages and 34.3% of sites.

Show the data

Pair

Overlap of pages

Overlap of sites

Gemini and AI Overviews

17.6%

34.3%

ChatGPT and AI Overviews

9.4%

19.9%

ChatGPT and Gemini

8.4%

19.2%

If you win a prompt on one engine, do you win it on the others?

Rarely. On the 24,334 prompts that all three engines answered with citations, only 1.5% of the cited URLs were cited by all three, and fewer than half of the prompts (46.4%) shared even one URL across the three. Overlap within a prompt (9.3% of URLs shared by two or more engines) is lower than across the whole pool (15.6%): the engines reach the same pages, often on different prompts.

The engines differ a lot in how private their picks are. On the same prompt, 45.3% of the pages Gemini cites are also cited by ChatGPT or AI Overviews. For AI Overviews the figure is 31.6%. For ChatGPT it is 9.5%, because ChatGPT cites so many pages that most of them are its alone. Put plainly, about half of what Gemini cites for a prompt shows up in another engine's answer to that prompt; for ChatGPT, it's one page in ten.

This is the number that should change how you report. A brand that celebrates a ChatGPT citation for its top prompt has learned very little about Gemini or AI Overviews. We filter every prompt report by engine before reading it, and AIclicks' prompt tracking keeps each engine's answer for every prompt so you can do the same.

Within one prompt, 1.5% of cited pages are cited by all three engines; Gemini's pages are most often shared (45.3%)

Figure 4. Overlap inside a single prompt. 1.5% of a prompt's cited pages are cited by all three engines; 45.3% of Gemini's are also cited elsewhere, against 9.5% of ChatGPT's.

Show the data

Measure (24,334 prompts all three engines answered)

Value

Cited pages on a prompt that all three engines cite

1.5%

Prompts sharing at least one cited page across all three engines

46.4%

ChatGPT: its cited pages that another engine also cites on the same prompt

9.5%

Gemini: its cited pages that another engine also cites on the same prompt

45.3%

AI Overviews: its cited pages that another engine also cites on the same prompt

31.6%

What kinds of pages do AI engines cite most?

Listicles. "Best X" roundups take 46.3% of all citations: 70.8% of Gemini's, 42.2% of ChatGPT's and 41.8% of AI Overviews'. Articles (blogs and news, which AIclicks labels together) come second at 25.8%. On these buying-intent prompts, Wikipedia (0.05% of citations) and forums such as Reddit and Quora (0.38%) are minor sources.

Below the top two, the shares drop quickly: unclassified pages 8.1%, comparison pages 4.7%, reviews 3.6%. "Other" is AIclicks' catch-all for pages it can't classify, and its largest domain is google.com itself. Articles lean toward ChatGPT, which gives them 29.5% of its citations; AI Overviews gives them 19.0% and Gemini 11.8%.

That is a long way from the "Wikipedia and Reddit" picture. The difference comes from the question. A buying-intent prompt asks for options, and the pages that list options are roundups and reviews, not encyclopedia entries or discussion threads. Studies that mix informational and commercial questions also count citations differently from us, so their percentages and ours aren't directly comparable. What our data does show is that for "best X" questions, a brand's citation work starts with the pages that rank products.

You can label your own cited URLs by hand in a spreadsheet, page by page. AIclicks' Sources tab does it for you: every cited URL carries the same 14 page-type labels used in this study, and you can filter by engine and page type.

Grouped bars: listicles are the most cited page type on every engine, 70.8% of Gemini's citations

Figure 5. Each engine's citations by page type (8 of 14 types). Listicles lead on every engine and take 70.8% of Gemini's citations.

Show the data

Page type

ChatGPT

Gemini

AI Overviews

Listicle

42.2%

70.8%

41.8%

Articles (blogs, news)

29.5%

11.8%

19.0%

Other (unclassified)

9.0%

3.0%

8.7%

Comparison

5.2%

3.7%

2.3%

Homepage

4.1%

2.5%

7.6%

Reviews

3.5%

4.9%

2.6%

Landing page

4.0%

1.1%

1.8%

YouTube

0.0%

0.1%

9.9%

How does each engine choose its sources?

Each engine has a clear signature once you compare it with the rest of the pool. We used a preference index to show it: an engine's share of a page type divided by that type's share of all citations. A value above 1 means the engine cites the type more than average; below 1, less.

Heatmap of the preference index: AI Overviews over-indexes on YouTube, social, forums and Wikipedia; Gemini on listicles

Figure 6. Preference index by engine and page type. AI Overviews scores 9.4 on YouTube, 6.1 on social and 5.6 on forums; Gemini scores 1.53 on listicles.

Show the data

Page type

ChatGPT

Gemini

AI Overviews

Share of all citations

Listicle

0.912

1.531

0.904

46.26%

Articles (blogs, news)

1.14

0.455

0.736

25.84%

Other (unclassified)

1.11

0.37

1.074

8.09%

Comparison

1.111

0.791

0.482

4.73%

Homepage

0.972

0.587

1.776

4.25%

Reviews

0.965

1.373

0.735

3.60%

Landing page

1.19

0.339

0.533

3.35%

How-to

1.081

0.613

0.947

1.43%

YouTube

0.02

0.089

9.415

1.05%

Alternatives

1.041

0.984

0.726

0.64%

Forums

0.349

1.104

5.609

0.38%

Social

0.473

0.062

6.149

0.34%

Wikipedia

0.321

1.599

5.127

0.05%

Marketplace

0.913

0.872

1.816

0.00%

ChatGPT cites far more of everything

ChatGPT gives 1,658 citations per 100 answers, against 315 for Gemini and 244 for AI Overviews, and it supplies 75.3% of every citation in the pool. Its relative preferences are landing pages (index 1.19), articles (1.14) and comparison pages (1.11). It almost never cites YouTube (0.02).

Because ChatGPT is three quarters of the pool, the pool mix sits close to ChatGPT's own, and its index can only move between 0.02 and 1.19. Small differences in its column still mean something.

We expected ChatGPT to favour brand-owned pages such as homepages, product and pricing pages. As a share of citations, it doesn't: homepages plus landing pages are 8.1% of ChatGPT's citations against 9.3% for AI Overviews. Counted per answer, though, ChatGPT cites 134.6 of these pages per 100 answers, against 22.8 for AI Overviews and 11.5 for Gemini. That gap comes from volume, not preference, and it's the reason ChatGPT leaves the most room for your own pages. Our guide on how to appear in ChatGPT results covers what that means in practice.

ChatGPT gives 1,658 citations per 100 answers, Gemini 315, AI Overviews 244

Figure 7. Citations per 100 answers on the same prompts. ChatGPT gives 1,658, Gemini 315 and AI Overviews 244.

Show the data

Engine

Answers

Citations

Citations per 100 answers

ChatGPT

182,584

3,027,923

1,658

Gemini

182,750

576,096

315

AI Overviews

170,226

414,863

244

Gemini is the most concentrated

Seven of every ten Gemini citations go to listicles (index 1.53), and reviews are its other strong preference (1.37). Gemini's highest index technically belongs to Wikipedia (1.60), but that rests on 498 citations; the listicle preference rests on 408,004 and is the one that shapes its answers. Social pages are where Gemini almost never goes (0.06).

AI Overviews is the outlier

AI Overviews is the only engine where YouTube carries real weight: 9.9% of its citations (index 9.4), and 97.2% of all YouTube citations in the study. It cites 10,176 of the 10,317 YouTube URLs in the data. It also over-indexes on social pages (6.1), forums (5.6), Wikipedia (5.1) and homepages (1.8, or 7.6% of its citations). Reddit and Quora both sit under the forums label.

Together, YouTube, homepages, forums and social pages make up 21.7% of AI Overviews' citations. Comparison pages are what it uses least (0.48). If AI Overviews matters to you, our guides on ranking in AI Overviews and why LLMs cite Reddit go deeper on the formats it favours.

Which pages do all three engines agree on?

Mostly listicles. Half (49.3%) of the 17,871 URLs that all three engines cite are listicles, and they take 62.2% of the citations going to those shared URLs. By rate, Wikipedia pages (29.4% cited by two or more engines) and homepages (24.8%) are the most likely to be shared. Listicles come third (24.0%), but by volume they make up 46.3% of every source two or more engines share.

The other end of the chart matters as much. Unclassified pages are 25.1% of the URLs a single engine cites but only 3.4% of the URLs all three agree on. YouTube, forum and social pages are rarely shared because AI Overviews cites them almost alone.

We predicted listicles would have the highest overlap rate of any page type. They don't, so that hypothesis is rejected; Wikipedia and homepages are shared more often per page. In volume, though, nothing comes close to listicles, which is why third-party pages carry so much weight in AI answers.

Dot plot: pages all three engines cite are 49.3% listicles and 3.4% unclassified, against 27.1% and 25.1% for single-engine pages

Figure 8. Page-type mix of pages cited by one engine only vs pages cited by all three. Listicles rise from 27.1% to 49.3%; unclassified pages fall from 25.1% to 3.4%.

Show the data

Page type

Pages cited by one engine only

Pages cited by all three engines

Listicle

27.1%

49.3%

Homepage

3.9%

8.5%

Articles (blogs, news)

26.8%

27.7%

Reviews

2.6%

3.3%

Landing page

3.6%

3.0%

Comparison

4.2%

2.9%

Forums

1.2%

0.2%

YouTube

2.3%

0.1%

Other (unclassified)

25.1%

3.4%

Does the source mix change by industry?

Yes, and the listicle share moves most. On ChatGPT it ranges from 24.5% of citations in Healthcare providers to 53.4% in Fashion, beauty and consumer goods; on Gemini from 55.4% (Healthcare providers) to 82.3% (Sales and customer experience). Healthcare providers sits furthest from every engine's usual mix, and Logistics and e-commerce operations sits closest.

AI Overviews' outlier formats peak in predictable places. YouTube reaches 19.3% of its citations in Media and publishing; social pages reach 7.9% and homepages 13.5% in Travel and leisure.

Reviews did not peak where we expected. We assumed software categories would lean on review sites, but Software and developer tools ranks 17th, 9th and 10th of 23 for review share on ChatGPT, Gemini and AI Overviews. Reviews peak in Gaming and gambling (5.9% of ChatGPT's citations, 6.1% of AI Overviews') and in Marketing and advertising on Gemini (10.4%).

Citation density varies too. ChatGPT gives 2,297 citations per 100 answers in Sales and customer experience and 923 in Real estate and home services, so the same number of prompts produces very different amounts of citation surface depending on the category.

Listicle share by industry ranges from 24.5% to 53.4% on ChatGPT and from 55.4% to 82.3% on Gemini

Figure 9. Listicles' share of each engine's citations: all 23 industries, and the lowest and highest industry. Healthcare providers is the lowest on every engine.

Show the data

Engine

Lowest industry

All 23 industries

Highest industry

ChatGPT

Healthcare providers: 24.5%

42.2%

Fashion, beauty and consumer goods: 53.4%

Gemini

Healthcare providers: 55.4%

70.8%

Sales and customer experience: 82.3%

AI Overviews

Healthcare providers: 26.6%

41.8%

Sales and customer experience: 61.2%

Which of our hypotheses held up?

We wrote four hypotheses and their pass tests before pulling any data. Two held and two didn't.

Hypothesis

Verdict

Deciding numbers

Gemini and AI Overviews overlap with each other more than either does with ChatGPT

Confirmed

URL overlap: Gemini and AI Overviews 17.6%, ChatGPT and AI Overviews 9.4%, ChatGPT and Gemini 8.4%

Listicles are the page type with the highest cross-engine overlap

Rejected

Listicles rank 3rd of 14 (24.0% of their URLs cited by 2+ engines), behind Wikipedia (29.4%) and homepages (24.8%)

AI Overviews cites forums and video more than ChatGPT and Gemini

Confirmed

AI Overviews preference index: forums 5.6, YouTube 9.4

ChatGPT cites brand-owned pages more than the Google engines

Rejected

Homepage and landing-page share: AI Overviews 9.3%, ChatGPT 8.1%, Gemini 3.6%

What does this mean for your brand?

Track each engine on its own. With 84.4% of cited URLs unique to one engine, a ChatGPT win says little about Gemini or AI Overviews. Done by hand, that means running your ten most important prompts in each engine every week and logging the cited URLs in a sheet, one tab per engine. AIclicks runs the same prompts on each engine and reports them separately; the ChatGPT, Gemini and AI Overviews trackers show what that looks like.

Target the shared sites, not only the shared pages. The 11.6% of domains all three engines cite collect 59.4% of citations, so a placement on one of those sites is the closest thing to cross-engine coverage the data shows. Manually, list the domains cited across your prompts and mark the ones two or more engines use. In AIclicks, the Sources tab's domain view shows each domain's share of all citations, and Get Mentioned lists the cited pages that leave your brand out. There's also a free AI citation tracker on our tools page.

Get into the roundups, starting with Gemini. Listicles are 70.8% of Gemini's citations and half of everything all three engines agree on. Search "best [your category]" and list the roundups that rank, then work out which ones already get cited. Filtering the Sources tab to page type Listicle and Mentioned: No gives you that outreach list directly, and the Outreach Agent can work through it.

For AI Overviews, cover video and community threads. AI Overviews is the only engine where YouTube, forums and social pages matter, at 21.7% of its citations together with homepages. Search YouTube and Reddit for the buying questions in your category and see who answers them. Engage in AIclicks surfaces the discussion threads that already appear in your citation data, and our guide to using Reddit for AI citations covers how to take part without sounding like an ad.

On ChatGPT, your own content has the most room. It cites 5 to 7 times more sources per answer and leans toward comparison pages, landing pages and articles. Writer 2.0 drafts comparison, listicle and how-to formats from your tracked prompts if you'd rather not start from a blank page.

Check your industry before copying anyone's playbook. In healthcare, listicles carry far less weight and unclassified pages far more; in sales, customer experience and B2B software, listicles dominate even more than average. For the bigger picture on getting cited at all, start with how to get cited by AI.

Methodology and limits


Item

Value

Categories

833 product categories, one AIclicks tracking project each

Prompts

30 per category, about 25,000 in total; all commercial and non-branded

Engines

ChatGPT, Gemini and Google AI Overviews, collected by AIclicks from the live interfaces

Run length

6 or 8 days per category, from the day tracking started to the end of the plan

Window

20–29 October 2026

Answers

535,560 (ChatGPT 182,584 · Gemini 182,750 · AI Overviews 170,226)

Citations

4,018,882

Cited URLs and domains

527,656 URLs on 127,336 domains

Collection. We pulled each category's data with four calls (answer counts, plus one citations call per engine), run slowly in a single lane so the service was never strained. 3,366 calls were made; 34 failed on the first try and all were redone. No category failed and no data was lost.

Citation counts. AIclicks sometimes counts one URL more than once in a single answer. We capped each URL's count at the number of answers it could have appeared in: the prompts that cited it multiplied by the days the category ran. The cap removed 6.9% of raw citations: 17.8% of Gemini's raw count, 5.4% of ChatGPT's and 0.01% of AI Overviews'.

Definitions.

Term

Meaning

Citation

One URL cited in one answer

Source

One unique cited URL

Domain

The registrable domain of a URL, using the Public Suffix List

Overlap

A source or domain cited by two or more engines

Pair overlap (Jaccard)

Sources both engines cite ÷ sources either engine cites

Preference index

An engine's share of a page type ÷ that type's share of all citations

Citations per 100 answers

Citations ÷ answers × 100, used because categories ran 6 or 8 days and AI Overviews answered fewer prompts

Hypotheses. Every hypothesis had a pass test written before the data was pulled. We report each one as confirmed or rejected against that test, including the two that failed.

Limits. Every prompt was a buying-intent question, so the results describe "best X" questions, not informational ones. The page types are AIclicks' own format labels: "Other" holds 8.1% of citations and 22.2% of URLs, and "Landing page" describes a format rather than ownership, so it includes pages such as sec.gov. The data covers one collection week, which makes it a snapshot rather than a trend. ChatGPT supplies 75% of all citations, so any whole-pool figure leans toward ChatGPT's own mix. AI Overviews answered fewer prompts because Google doesn't show an AI Overview for every query. Three small categories had little or no data on at least one engine; we kept them, and they are too small to move any figure. AIclicks' engine labels aren't confirmed as the exact model versions behind each interface.

Cite this study. AIclicks (2026). What Sources Do ChatGPT, Gemini and AI Overviews Cite? 535,560 AI answers across 833 product categories, 20–29 October 2026. aiclicks.io/blog/chatgpt-gemini-ai-overviews-sources-study

Frequently asked questions

Where does ChatGPT get its information?

ChatGPT answers from its training data and, when it searches, from web pages it retrieves and cites. On buying-intent prompts, the pages it cites are mostly listicles (42.2%) and blog or news articles (29.5%), and it cites far more of them per answer than Gemini or AI Overviews. More in our guide on appearing in ChatGPT results.

What sources does Google AI Overviews use?

In our data, listicles (41.8% of its citations), articles (19.0%), YouTube (9.9%) and homepages (7.6%). It is the only one of the three engines where YouTube, forums and social posts carry real weight. More in how to rank in AI Overviews.

Does Gemini cite the same sources as Google AI Overviews?

More than either does with ChatGPT, but mostly not: they share 17.6% of cited URLs and 34.3% of cited domains. You can follow both in the Gemini tracker.

Do ChatGPT and Google's AI engines cite Reddit?

On buying-intent prompts, rarely. Forums such as Reddit and Quora take 0.38% of all citations: 2.15% of AI Overviews' citations and 0.13% of ChatGPT's. More in why LLMs cite Reddit.

Why do AI engines cite listicles so often?

Buying-intent prompts ask for options, and roundups list them. Listicles take 46.3% of all citations and half of the URLs that all three engines agree on. More in why AI cites third-party sources.

If ChatGPT cites my page, will Gemini cite it too?

Not necessarily. 84.4% of cited URLs appear in one engine only, and within the same prompt just 1.5% of cited URLs are shared by all three engines. Prompt tracking per engine shows where you stand on each.

How many sources does ChatGPT cite per answer?

On our buying-intent prompts, about 16.6 citations per answer on average, against 3.2 for Gemini and 2.4 for AI Overviews. The ChatGPT tracker shows which of them mention your brand.

Is this a trend over time?

No. It is a snapshot of one collection week (20–29 October 2026). For the wider context, see our AI SEO statistics.

What to do next

The fastest way to use this study is to run it on your own category. Pick your ten most valuable prompts, check which pages each engine cites for them, and sort those pages into the ones you can earn a place on and the ones you can't. AIclicks does that for every prompt you track, per engine, with the cited pages already labelled. See how AIclicks tracks your prompts.

This post is the first of three built on the same data. The other two look at what decides which sites and pages get cited:

SEO & GEO Specialist @ aiclicks.io

SEO & GEO Specialist @ aiclicks.io

Maksim Baradulin is an SEO and GEO specialist with more than 1.5 years of experience, working with some of Lithuania’s biggest companies. At AIclicks, Maksim takes on high-priority GEO work for the platform’s largest clients. Maksim specializes in agentic SEO and GEO, and builds automations that handle repeat search and AI visibility work. The goal is brands that rank in Google and get cited in answers from ChatGPT and other LLMs.

Maksim Baradulin is an SEO and GEO specialist with more than 1.5 years of experience, working with some of Lithuania’s biggest companies. At AIclicks, Maksim takes on high-priority GEO work for the platform’s largest clients. Maksim specializes in agentic SEO and GEO, and builds automations that handle repeat search and AI visibility work. The goal is brands that rank in Google and get cited in answers from ChatGPT and other LLMs.

¿Te gustó lo que leíste? Obtén tu guía definitiva de SEO con IA aquí:

¿Te gustó lo que leíste? Obtén tu guía definitiva de SEO con IA aquí:

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

Usa AIclicks para optimizar para SEO de IA mediante el seguimiento, análisis y mejora de tus menciones en las respuestas de IA.

¿Te queda alguna duda?
Reserva una llamada aquí: