A ChatGPT answer cites 3.7 sources on average, the fewest of five AI engines, while a Perplexity answer cites 20, more than five times as many. On ChatGPT, a website has to be one of a handful of sources to get cited at all.
Analysis of answers to 4,984 questions · answers from ChatGPT, Gemini, Perplexity and Google's AI features, September 2026
Key Findings
- 1.ChatGPT cites 3.7 sources per answer on average, the fewest of five AI engines, while Perplexity cites 20 per answer.
- 2.Google AI Overviews cite 7.3 sources per answer on average and Google AI Mode 6.7, about twice as many as ChatGPT.
- 3.On "best" product questions, Gemini cites 6.6 sources per answer on average, more than twice the 2.8 it cites on general questions.
- 4.Perplexity cites 10 or more sources in 99.1% of its answers, against 0.4% of ChatGPT answers.
- 5.The median Gemini answer cites 2 sources and the median ChatGPT answer 3, against 19 for the median Perplexity answer.
Summary
Ask Perplexity which project management tool suits a small agency and the answer arrives with a numbered list of about 20 sources underneath: review sites, vendor pages, forum threads, comparison articles. Ask ChatGPT the same question and three or four sources appear. Both answers name products; one gives far more websites a share of the credit.
The number of sources an engine cites sets the odds for every website that wants to be one of them. On Perplexity, being among the 20 or so most useful pages on a question can be enough. On ChatGPT, a page has to be one of the few best, and each citation is harder to earn and more visible when it comes.
The gap is wide. ChatGPT cites 3.7 sources per answer on average and Perplexity 20, more than five times as many. Google's AI features sit in between, with AI Overviews at 7.3 and AI Mode at 6.7. Perplexity cites 10 or more sources on 99.1% of its answers; ChatGPT does so on 0.4%.
What we measured
On September 23, 2026 we asked five AI engines, ChatGPT, Gemini, Perplexity, Google AI Mode and Google AI Overviews, 2,984 general questions and 2,000 "best" product questions such as "what is the best robot vacuum", from a US location. ChatGPT, Gemini and Google AI Mode answered each question twice and Perplexity once. For each answer we counted the distinct web pages it cited and averaged the counts by engine and by question type.
A cited source is a link to a web page that an AI engine shows with its answer. Each page is counted once per answer, even if the answer links it several times, and answers with no sources count as zero. The median is the middle answer: half cite more sources, half cite fewer.
Perplexity cites five times as many sources as ChatGPT
The engines fall into three groups. ChatGPT (3.7) and Gemini (4.3) cite a few sources. Google AI Mode (6.7) and Google AI Overviews (7.3) cite a moderate number, about twice ChatGPT's. Perplexity cites a long list on almost every answer, 20 on average.
Source: Suff Digital analysis of answers to 4,984 questions · answers from ChatGPT, Gemini, Perplexity and Google's AI features, September 2026
A long source list turns an answer into a research page that invites the reader to check and compare. A short one asks the reader to trust the summary. That makes a ChatGPT citation scarcer and more prominent: it is one of three or four links the reader sees, not one of twenty.
Product questions get more sources
Four of the five engines cite more sources when the question is about which product to buy. Gemini changes most, from 2.8 sources on general questions to 6.6 on "best" product questions, more than twice as many. Perplexity rises from 18.4 to 22.3, Google AI Mode from 5.8 to 8 and ChatGPT from 3.4 to 4.1. Google AI Overviews is the exception, at 7.5 on general questions and 7 on product questions.
| AI engine | General questions | "Best" product questions |
|---|---|---|
| ChatGPT | 3.4 | 4.1 |
| Gemini | 2.8 | 6.6 |
| Google AI Mode | 5.8 | 8 |
| Google AI Overviews | 7.5 | 7 |
| Perplexity | 18.4 | 22.3 |
A product recommendation needs evidence from several places: reviews, specifications, prices, retailer pages. The engines gather more of it, which gives brands, retailers and review sites more openings on buying questions than on general ones.
Perplexity lists 10 or more sources on almost every answer
The median answer shows how differently the engines behave. The median Perplexity answer cites 19 sources, and 99.1% of its answers cite 10 or more. The median ChatGPT answer cites 3, and 0.4% of its answers reach 10; its longest answer cited 16 sources. Half of Gemini's answers cite two sources or fewer, because it often cites none at all.
Source: Suff Digital analysis of answers to 4,984 questions · answers from ChatGPT, Gemini, Perplexity and Google's AI features, September 2026
| AI engine | Median sources per answer | Answers with 10 or more sources | Most sources in one answer |
|---|---|---|---|
| ChatGPT | 3 | 0.4% | 16 |
| Gemini | 2 | 14.4% | 25 |
| Google AI Mode | 5 | 30.4% | 40 |
| Google AI Overviews | 7 | 17.7% | 39 |
| Perplexity | 19 | 99.1% | 45 |
Google AI Mode swings the most: its median answer cites 5 sources, but 30.4% of its answers cite 10 or more and one cited 40. It keeps short answers short and builds long source lists when a question needs them.
What this means for marketers
Treat each engine as a different target. On Perplexity, broad coverage of your topic and solid Google rankings can get you into the long source list. On ChatGPT, you need to be one of the three or four best pages on the exact question, which favors pages that answer it in the first lines, back it with specific facts, and come from a site that is clearly an authority on the subject.
Put product pages and buying guides first. Four of the five engines cite more sources on "best" questions, so comparison pages, category pages and honest buying guides have the best odds of any content you publish. Make them specific: prices, specifications, who each product suits and how it compares. As an ai seo company, Suff Digital tracks how often every major engine cites you and builds the pages that win those citations.
Embed this research
Paste this on your site to embed the charts. It links back to the source automatically.
<iframe id="sd-ai-citations-per-answer" src="https://www.suffdigital.com/embed/data-studies/ai-citations-per-answer" width="100%" height="600" style="width:100%;border:1px solid #E5E7EB;border-radius:12px" loading="lazy" title="ChatGPT Cites 3.7 Sources per Answer vs 20 for Perplexity - Suff Digital"></iframe>
<script>window.addEventListener("message",function(e){if(e&&e.data&&e.data.sdEmbed==="ai-citations-per-answer"&&e.data.height){var f=document.getElementById("sd-ai-citations-per-answer");if(f){f.style.height=e.data.height+"px";}}});</script>
<p style="font:14px/1.5 system-ui,sans-serif">Source: <a href="https://www.suffdigital.com/resources/data-studies/ai-citations-per-answer">ChatGPT Cites 3.7 Sources per Answer vs 20 for Perplexity - Suff Digital</a></p>
Cite this study
Suff Digital. (2026). ChatGPT Cites 3.7 Sources per Answer vs 20 for Perplexity. https://www.suffdigital.com/resources/data-studies/ai-citations-per-answer
ChatGPT Cites 3.7 Sources per Answer vs 20 for Perplexity - Suff Digital - https://www.suffdigital.com/resources/data-studies/ai-citations-per-answer
Frequently asked questions
Related studies
- AI SEOGoogle's Gemini Cites No Source in 61.2% of Its Answers vs 2.2% for ChatGPT
- AI SEOChatGPT and Google's AI Overviews Share Only 8.4% of the Websites They Cite
- AI SEO70% of AI Answers With Multiple Sources Cite the Same Website More Than Once
- AI SEOOnly 0.3% of the Search Results ChatGPT Retrieves Are the Pages It Ends Up Citing