- The three AIs rarely agree. For the same question on the same day, only 1.5% of cited websites were cited by all three.
- Other companies’ websites, mostly their blogs, are the biggest source everywhere. The brand being asked about gets just 2 to 4% of citations.
- ChatGPT favors docs and research, Gemini favors blog posts and YouTube, and Perplexity favors Reddit and forums.
- Listicles rule shopping questions but matter far less for B2B software.
Do ChatGPT, Gemini, and Perplexity cite the same websites?
Almost never. We matched 170 sets of answers where all three engines got the same question on the same day.
Of the 4,751 websites they cited, 88% appeared in only one engine’s answer, 11% in two, and just 1.5% in all three (leadrescue.app data, August to October 2026). Each AI is reading its own slice of the web.
That matches what we saw when we asked whether the three engines recommend the same brands. They don’t, and their sources overlap even less.
They also cite very different amounts. Perplexity links about 20 sources per answer, Gemini about 12, and ChatGPT fewer than 5.
So a spot in a ChatGPT answer is harder to win, and each one carries more weight when you do.
How did we run this study?
We sent 39 buyer-style questions for 8 brands to ChatGPT, Gemini and Perplexity between August 6 and October 4, 2026. That gave 202 answers per engine and 7,409 cited links.
Each engine ran through its official API with live web search on: OpenAI’s web search tool on its lightest setting, Gemini with Google Search grounding, and Perplexity’s Sonar model. The consumer apps may cite a little differently.
An AI classifier labeled every cited page on the day we found it, twice: by type of website (news site, forum, help docs and so on) and by page format (blog post, homepage, video and so on). We only used scans from after that labeling went live, so every link was judged the same way.
One sportswear brand produced about 23% of all links, so the share tables weight each brand equally. Most brands were scanned for only a few days, so treat this as a strong early signal rather than a final answer.
What types of websites does each AI engine cite most?
Each engine has a clear favorite beyond company websites. ChatGPT pulls 24% of its citations from help docs, research papers, government sites and Wikipedia, about six times the rate of Gemini or Perplexity.
Gemini and Perplexity go the other way. Each pulls about 9% of its citations from social and community sites like YouTube, Reddit and LinkedIn, while ChatGPT pulls under 2%.
| Type of website | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Other companies’ sites | 34% | 55% | 50% |
| News and magazines | 17% | 12% | 19% |
| Help docs | 15% | 2% | 2% |
| Research, government and Wikipedia | 9% | 2% | 3% |
| Social and community sites | 2% | 9% | 9% |
| Competitors’ sites | 6% | 4% | 6% |
| Reviews, directories and stores | 6% | 7% | 5% |
| Your own site | 4% | 2% | 2% |
| Other or unlabeled | 9% | 7% | 5% |
A citation is a link an AI engine shows as the source behind its answer. A mention is different: it’s when the engine simply names your brand.
News and magazine sites like Forbes are the one source type all three engines lean on heavily, at 12 to 19% of citations. If you can only chase one kind of coverage that works everywhere, that’s it.
Why do other companies’ blogs get cited more than your own website?
Because they answer the question being asked. A vendor blog is a company’s content marketing, such as guides, comparisons and “best tools” posts, written by a business that isn’t the one in the question.
Other companies’ websites, mostly these blogs and guides, made up 34 to 55% of citations across the three engines. The brand being asked about got 2 to 4%.
The examples look familiar. Salesforce’s guides, the ZoomInfo blog, Zapier and Shopify were all cited for questions about other companies’ products.
Reach tells the same story. At least one vendor blog or guide showed up in 84% of Perplexity answers, 75% of Gemini answers and 38% of ChatGPT answers.
The brand’s own site appeared in only 8 to 13% of answers.
This lines up with our earlier sneaker study, where Nike was named in 93% of answers but nike.com was cited only twice.
It also fits the Princeton-led team that coined “GEO”, who found that adding citations, quotes and statistics to a page can “boost visibility by up to 40% in generative engine responses” (Aggarwal et al., Princeton, 2023).
Note: this is good news for small companies. Many of the most-cited vendor blogs in our data belong to young startups like Salesforge and Warmly, not household names. AI cares whether a page answers the question, not how big the company behind it is.
Which page formats does each AI engine prefer?
Gemini prefers blog posts, ChatGPT prefers homepages and documentation, and Perplexity leans hardest on listicles and forum threads. One in three Gemini citations is a blog article, more than double ChatGPT’s 15%.
ChatGPT, meanwhile, cites documentation around eight times as often as the other two, and homepages up to three times as often.
| Page format | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Blog article | 15% | 32% | 25% |
| “Top X” listicle | 10% | 12% | 21% |
| Guide or how-to | 12% | 13% | 10% |
| Product page | 10% | 8% | 7% |
| Homepage | 10% | 3% | 6% |
| Documentation | 10% | 1% | 1% |
| Research or report | 7% | 1% | 0.7% |
| Comparison or alternatives | 4% | 4% | 5% |
| Video | 0% | 4% | 4% |
| Forum thread | 0.3% | 3% | 3% |
| Everything else | 22% | 20% | 16% |
Documentation is a product’s help center or developer guide, like HubSpot’s knowledge base or Amazon’s AWS docs. ChatGPT treats it as a trusted reference, while Gemini and Perplexity barely touch it.
Video is the starkest split. ChatGPT cited zero videos across all 202 answers, while YouTube was Gemini’s most-cited outside website and showed up for all 8 brands.
How often does each AI cite each kind of source?
Shares of citations can hide things, because Perplexity cites four times as many links as ChatGPT. So we also counted how many answers cited at least one source of each kind.
The gaps get even clearer. Perplexity pulled a forum thread into 35% of its answers, Gemini into 15%, and ChatGPT into 1 of 202.
| Share of answers citing… | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Blog article | 37% | 73% | 94% |
| Other company’s blog or guide | 38% | 75% | 84% |
| News or magazine article | 36% | 50% | 66% |
| “Top X” listicle | 27% | 45% | 64% |
| Social or community site | 3% | 46% | 54% |
| Forum thread | 0.5% | 15% | 35% |
| Video | 0% | 24% | 10% |
| Help docs | 29% | 13% | 18% |
| Research, government or Wikipedia | 25% | 16% | 18% |
| Your own site | 11% | 13% | 8% |
Two rows stand out. Gemini cited a video in 24% of its answers, more than twice Perplexity’s rate.
And ChatGPT, despite citing the fewest links, still reached for help docs and research more often than either rival.
Do “Top 10” listicles still get cited by AI?
Yes, but it depends on what’s being asked far more than on which AI is asked. A listicle is a ranked roundup like “10 best running shoes for flat feet.”
For shopping questions, listicles made up 30 to 66% of citations. For business software questions, that fell to 5 to 15%.
| Question type | ChatGPT | Gemini | Perplexity |
|---|---|---|---|
| Shopping questions (1 sportswear brand) | 62% | 30% | 66% |
| Business and software questions (7 brands) | 5% | 10% | 14% |
ChatGPT shows the biggest swing. Most of its sources for sneaker questions were listicles from sites like OutdoorGearLab and Runner’s World, but only 5% for software questions.
So if you sell something people shop around for, getting onto roundup lists is your biggest lever on every engine. If you sell B2B software, comparison sites, docs and helpful guides will do more for you.
How should you adapt your content for each AI engine?
Start with what all three share, then add one habit per engine. Helpful guides and comparisons that answer one buyer question each work everywhere, since other companies’ content like that makes up 17 to 45% of every engine’s citations.
On top of that, add docs for ChatGPT, blog posts and video for Gemini, and community presence for Perplexity.
We think of the three engines as three personalities. ChatGPT is the Librarian, Gemini the Blog Reader, and Perplexity the Crowd Listener.
| Engine | Personality | What it reaches for | What to publish |
|---|---|---|---|
| ChatGPT | The Librarian | Docs, research, Wikipedia, G2; about 5 sources per answer | Clear help docs, an FAQ, original data, and a homepage that says plainly what you do |
| Gemini | The Blog Reader | Blog posts and guides; YouTube is its top outside site | In-depth blog posts and how-to guides, plus short explainer videos |
| Perplexity | The Crowd Listener | Reddit, LinkedIn, forums, listicles; about 20 sources per answer | Honest answers in Reddit threads, LinkedIn posts, and a place on roundup lists |
Reddit was Perplexity’s most-cited outside website, ahead of LinkedIn and Forbes. That tells you where time spent on Reddit pays off.
These habits can shift as the engines update, so check your own category rather than trusting any study forever. Lead Rescue’s citation report shows which URLs each engine cited for your buyers’ questions, so you can see your own mix.
Why does ChatGPT cite fewer sources than Perplexity?
In our 2026 tests, ChatGPT averaged 4.6 sources per answer against 19.9 for Perplexity. Part of that is design, since Perplexity is built as a search engine that shows its sources up front and was handling 100 million queries a week by late 2024 (TechCrunch, 2024).
Part is our setup. We queried ChatGPT through OpenAI’s API on its lightest web search setting, and the ChatGPT app may show more sources.
Does ChatGPT ever cite YouTube or Reddit?
Rarely, based on our 2026 data. ChatGPT cited zero videos in 202 answers and cited a forum thread in just one of them.
Social and community sites made up under 2% of its citations, against about 9% for Gemini and Perplexity. Our earlier sneaker study found the same pattern, with ChatGPT leaning on editorial sites like Tom’s Guide instead of YouTube.
Is it worth writing guest posts on other companies’ blogs?
Our data suggests yes, as long as the post genuinely answers a buyer question. Other companies’ blogs and guides appeared in 75 to 84% of Gemini and Perplexity answers in 2026, far more than any brand’s own site.
A useful guest guide, or a mention in a partner’s comparison post, puts you inside the pages AI already likes to cite.
Want to know which page types AI cites for your own buyers’ questions? Lead Rescue scans ChatGPT, Gemini and Perplexity daily and shows every URL they cite, by engine. See what AI cites in your category →