We asked ChatGPT, Perplexity, Gemini, and Claude the same 47 buyer questions, with web search on, and recorded every source each one cited. Here is what 1237 citations reveal, sliced every way. The full dataset is published.
ChatGPTPerplexityGeminiClaude
40%
of citations go to independent listicles
1237
citations logged
713
distinct domains cited
5
domains cited by all 4 engines
0
answers mentioned Prefer
01
Where the citations go
Independent "best-X" listicles are the single biggest source of AI citations, far ahead of the vendors' own sites. Filter by engine to see how each one's diet differs.
Independent listicles40%
Vendors' own sites34%
Media & SaaS blogs15%
Reddit & forums4%
YouTube3%
Review sites (G2 etc.)3%
Other1%
Across all four engines, 40% of 1237 citations went to independent listicles and just 34% to the vendors' own sites.
02
Find your category
The content that gets cited changes by category. Pick yours to see which format wins, and the exact sources you should get featured in to be cited.
In SEO & AI search, 47% of citations go to independent listicles. Win that format first.
Independent listicles47%
Vendors' own sites33%
Media & SaaS blogs12%
Reddit & forums4%
YouTube3%
Review sites (G2 etc.)1%
Other1%
Sources to get featured in
Rreddit.comReddit & forums7×
Nnightwatch.ioVendors' own sites6×
Mmarketermilk.comIndependent listicles5×
Yyoutube.comYouTube5×
Rrankability.comIndependent listicles4×
Zzapier.comMedia & SaaS blogs4×
From 169 citations across 6 questions in this category.
Same questions, four very different citation habits, from how many sources they pull to whether they ever look at Reddit.
ChatGPTThe skimmergpt-4o
Answered 29 of 47 questions from memory, with no live search and no citations at all.
119citations
89domains
2.5per answer
23%vendor sites
PerplexityThe community listenersonar-pro
Cited Reddit and YouTube more than the other three engines combined (65 of them), on top of the usual listicles.
421citations
273domains
9per answer
36%vendor sites
GeminiThe wide netgemini-2.5-pro
Cited the most distinct sources (367) and leans hardest on third-party listicles (47%).
441citations
367domains
9.4per answer
34%vendor sites
ClaudeThe careful citerclaude-sonnet-4-5
Attaches a citation to nearly every specific claim, split fairly evenly between listicles and the vendors' own sites.
256citations
220domains
5.4per answer
36%vendor sites
04
The engines barely agree
Of 713 distinct domains cited across the study, only 5 were cited by all four engines. Winning a citation on one engine is almost no signal about the others.
Cited by one engine only539
Cited by two engines117
Cited by three engines52
Cited by all four engines5
The only two domains every engine agreed on: zapier.com</strong> and <strong>g2.com</strong> and <strong>thedigitalprojectmanager.com</strong> and <strong>forbes.com</strong> and <strong>getaleph.com. Everything else was fragmented, 76% of domains were cited by a single engine.
05
The most-cited sources
The domains AI search actually pulls from. Filter by engine to see who each one trusts.
Source
Type
Cited by
Citations
reddit.com
Reddit & forums
43
youtube.com
YouTube
32
zapier.com
Media & SaaS blogs
31
g2.com
Review sites (G2 etc.)
19
salesforce.com
Vendors' own sites
13
techradar.com
Media & SaaS blogs
12
monday.com
Vendors' own sites
11
forbes.com
Media & SaaS blogs
10
emailvendorselection.com
Independent listicles
9
guideflow.com
Independent listicles
9
hubspot.com
Media & SaaS blogs
8
peoplemanagingpeople.com
Independent listicles
8
technologyadvice.com
Independent listicles
8
emailtooltester.com
Independent listicles
7
marketermilk.com
Independent listicles
7
brex.com
Vendors' own sites
6
gusto.com
Vendors' own sites
6
klaviyo.com
Vendors' own sites
6
nightwatch.io
Vendors' own sites
6
thedigitalprojectmanager.com
Independent listicles
6
zoho.com
Vendors' own sites
6
efficient.app
Independent listicles
5
featurebase.app
Vendors' own sites
5
gartner.com
Review sites (G2 etc.)
5
Top 24 of 713 cited domains. Dots show which engines cited each source.
06
What AI recommends
The brands the four engines named most across all 47 questions, a share-of-voice snapshot for AI search.
HubSpot22×
Zendesk17×
Asana16×
Gusto16×
BambooHR14×
Ahrefs13×
Pipedrive13×
Rippling13×
Surfer SEO13×
Notion12×
SE Ranking12×
Semrush12×
ClickUp10×
Shopify10×
Zoho Desk10×
ActiveCampaign9×
Freshdesk9×
Google Search Console9×
07
"X vs Y" questions are won by third parties
When a buyer pits two named brands against each other, the engines stop reading the vendors' own marketing and reach for neutral comparisons.
On "X vs Y" questions
18%
of citations were the vendors' own sites
On "best tool" questions
37%
of citations were the vendors' own sites
Head-to-head answers pulled third-party listicles at 54%, roughly 2.1× less from vendor pages than "best tool" answers did.
What this means for you
You do not win AI search on your own site. You win it in the sources AI reads.
40% of citations are third-party listicles, the engines share almost no sources, and Prefer showed up in 0 of 188 answers, including our own. See exactly where you stand across every engine.
On July 10, 2026 we sent 47 buyer questions to four engines through the DataForSEO LLM API with web search on: ChatGPT (gpt-4o), Perplexity (sonar-pro), Gemini (gemini-2.5-pro), and Claude (claude-sonnet-4-5), producing 188 answers. From each we extracted every cited URL, normalized it to its root domain, and classified it into one source type. The figures are counts of those 1237 citations.
The 47 questions we asked
What are the best AI search visibility tools in 2026?
What are the best SEO tools in 2026?
What are the best AI writing tools for marketing teams?
What is the best CRM for small businesses?
What is the best project management software for remote teams in 2026?
What are the best email marketing platforms in 2026?
What is the best expense management software for startups?
What is the best help desk software for SaaS companies?
What are the best applicant tracking systems in 2026?
What are the best marketing analytics platforms in 2026?
Notion vs Coda: which is better for a startup?
HubSpot vs Salesforce: which is better for a mid-market company?
Asana vs Monday.com: which project management tool is better?
How do I improve my brand's visibility in AI search?
What are the best generative engine optimization (GEO) tools in 2026?
What are the best rank tracking tools in 2026?
What are the best content optimization tools for SEO?
What are the best social media management tools in 2026?
What are the best marketing automation platforms for B2B?
What is the best CRM for a B2B sales team?
What are the best sales engagement platforms in 2026?
Pipedrive vs HubSpot: which CRM is better for a small team?
What is the best CRM for real estate agents?
What are the best product management tools in 2026?
ClickUp vs Asana: which is better for a growing team?
What are the best time tracking tools for teams?
What are the best HR software platforms for small businesses?
What are the best accounting software tools for startups?
What is the best customer support software in 2026?
What are the best landing page builders in 2026?
What are the best payroll software platforms for small businesses?
What are the best employee onboarding software tools?
QuickBooks vs Xero: which is better for a small business?
What are the best corporate card and spend management platforms?
What are the best live chat software tools in 2026?
What are the best ecommerce platforms in 2026?
Shopify vs BigCommerce: which is better for a growing store?
What are the best product review apps for Shopify?
What are the best subscription billing platforms for ecommerce?
What are the best performance management software tools in 2026?
Rippling vs Gusto: which is better for a small business?
What are the best invoicing software tools for freelancers?
What are the best financial planning and analysis (FP&A) tools in 2026?
What are the best knowledge base software platforms?
What are the best call center software solutions in 2026?
What are the best inventory management software tools for ecommerce?
What are the best abandoned cart recovery tools in 2026?
How to read this. One run on one day, a small 47-question sample, US-default locale, using API model endpoints as proxies for the consumer apps. LLM outputs are non-deterministic, so a re-run would shift specific domains and counts. Treat the headline patterns as directional, not settled statistics, and read the full record: the playbook and the written report.