When B2B buyers ask Google AI Mode what to buy or how to choose, the answer is built mostly from small agency and software vendor blogs, plus three platforms: YouTube (cited in 48% of answers), LinkedIn (44%) and Reddit (30%). Big publishers and analysts barely appear (3.1% of citations), Wikipedia never appeared, and the long tail is huge: 239 different domains were cited across 50 answers, and 200 of them only once. For a B2B founder, that means a well-structured page on your own site, a LinkedIn presence and a YouTube video can all realistically earn a citation, while a Forbes mention is not the main route in.
This is an original study. I wrote 50 questions in the words B2B buyers use, ran each one in Google AI Mode on 17 October 2026, and recorded every domain the answer cited and every brand it highlighted. The raw data is free to download: ai-citations-50-questions-2026-10.csv.
#How the study worked
- Engine: Google AI Mode (the
udm=50search mode), one fresh query per question, no follow-ups. - Date: 17 October 2026, all 50 queries in one session.
- Location: the browser was in India. Google localises AI Mode, and some answers showed rupee prices and Indian agencies, so a US or UK run will differ.
- Questions: 50, in five groups of 10: (a) choosing B2B marketing services such as agencies, ghostwriters and fractional CMOs, (b) choosing B2B SaaS tools such as CRM, sales engagement, marketing automation and AI visibility tools, (c) LinkedIn for B2B, (d) Reddit and community marketing for B2B, (e) AI search and GEO for B2B. They were phrased the way buyers type: "best X for Y", "X vs Y", "how much does X cost", "how to choose X".
- What was recorded: every external domain linked from the answer page (Google's own domains excluded, subdomains such as business.linkedin.com merged into the parent), and every company or brand the answer highlighted in bold or as a table row label.
- Completion: 50 of 50 queries returned a full answer. None failed.
The full question list is in the CSV.
#The headline results
- 48% YouTube was cited in 24 of 50 Google AI Mode answers to B2B buyer questions, more than any other domain (Shruti Kaushik study, 17 Oct 2026).
- 44% LinkedIn was cited in 22 of 50 answers (Shruti Kaushik study, 17 Oct 2026).
- 30% Reddit was cited in 15 of 50 answers (Shruti Kaushik study, 17 Oct 2026).
- 82% At least one of YouTube, LinkedIn or Reddit was cited in 41 of 50 answers (Shruti Kaushik study, 17 Oct 2026).
- 0% Wikipedia was cited in none of the 50 answers (Shruti Kaushik study, 17 Oct 2026).
- 7.1 Answers cited an average of 7.14 distinct domains, from 3 to 14 (Shruti Kaushik study, 17 Oct 2026).
- 239 Distinct domains were cited across the 50 answers, and 200 of them (84%) were cited only once (Shruti Kaushik study, 17 Oct 2026).
#Top 20 most cited domains
Share of the 50 answers that cited each domain at least once.
| Rank | Domain | Answers citing it | Share | Type |
|---|---|---|---|---|
| 1 | youtube.com | 24 | 48% | Platform (user content) |
| 2 | linkedin.com | 22 | 44% | Platform (user content) |
| 3 | reddit.com | 15 | 30% | Platform (user content) |
| 4 | heysid.com | 5 | 10% | Software vendor |
| 5 | thesmarketers.com | 4 | 8% | Agency |
| 5 | stackmatix.com | 4 | 8% | Agency |
| 5 | zoominfo.com | 4 | 8% | Software vendor |
| 5 | hubspot.com | 4 | 8% | Software vendor |
| 5 | zapier.com | 4 | 8% | Software vendor |
| 5 | growthspreeofficial.com | 4 | 8% | Agency |
| 11 | tripledart.com | 3 | 6% | Agency |
| 11 | saashero.net | 3 | 6% | Agency |
| 11 | okara.ai | 3 | 6% | Software vendor |
| 11 | gushwork.ai | 3 | 6% | Software vendor |
| 11 | apollo.io | 3 | 6% | Software vendor |
| 11 | lemlist.com | 3 | 6% | Software vendor |
| 11 | f5bot.com | 3 | 6% | Software vendor |
| 11 | ahrefs.com | 3 | 6% | Software vendor |
| 11 | semrush.com | 3 | 6% | Software vendor |
| 20 | 20 domains tied | 2 | 4% | Mixed |
The 20 domains tied at 2 answers include g2.com, forrester.com, medium.com, quora.com, clay.com, 6sense.com, demandbase.com and sproutsocial.com; the full list is in the CSV. After the three platforms, no single website appears in more than 10% of answers. The field is wide open.
#Top domains by question group
| Group | Top cited domains (answers out of 10) | Avg domains per answer | Answers naming brands |
|---|---|---|---|
| (a) Marketing services | youtube.com 4, linkedin.com 3, then thinkcapadvisors.com, saashero.net, blockclub.co, revvgrowth.com 2 each | 6.0 | 4 |
| (b) SaaS tools | youtube.com 6, zoominfo.com 4, hubspot.com 4, apollo.io 3, zapier.com 3 | 8.1 | 9 |
| (c) LinkedIn for B2B | linkedin.com 8, youtube.com 2, heysid.com 2, swydo.com 2, growthspreeofficial.com 2 | 7.1 | 4 |
| (d) Reddit and community | reddit.com 8, youtube.com 6, linkedin.com 5, f5bot.com 3, infrasity.com 2 | 6.9 | 4 |
| (e) AI search and GEO | youtube.com 6, linkedin.com 6, reddit.com 3, ahrefs.com 3, semrush.com 3 | 7.6 | 6 |
Two patterns stand out. First, the platform a question is about gets cited for it: LinkedIn in 8 of 10 LinkedIn questions, Reddit in 8 of 10 Reddit questions. Second, software questions are where big vendors win. ZoomInfo, HubSpot and Apollo each got cited on their own category questions, while service questions were answered almost entirely from small agency blogs.
#Vendors, agencies, publishers: who actually supplies the answer
I classified every cited domain into one of five types: platform (YouTube, LinkedIn, Reddit, Quora, Medium), agency or consultancy, software vendor, publisher or analyst firm, and review directory.
| Domain type | Citations | Share of all 357 citations | Answers citing at least one |
|---|---|---|---|
| Software vendor sites | 145 | 40.6% | 36 of 50 |
| Agency and consultancy sites | 131 | 36.7% | 41 of 50 |
| Platforms (YouTube, LinkedIn, Reddit, Quora, Medium) | 66 | 18.5% | 42 of 50 |
| Publishers and analysts | 11 | 3.1% | 8 of 50 |
| Review directories (G2, TrustRadius, Slofile) | 4 | 1.1% | 3 of 50 |
- 77.3% Agency and software vendor websites supplied 276 of 357 citations in Google AI Mode answers to B2B buyer questions (Shruti Kaushik study, 17 Oct 2026).
- 3.1% Publishers and analyst firms such as Forbes, Forrester and Gartner supplied 11 of 357 citations (Shruti Kaushik study, 17 Oct 2026).
- 1.1% Review directories such as G2 and TrustRadius supplied 4 of 357 citations (Shruti Kaushik study, 17 Oct 2026).
Most of the vendor and agency pages cited are blog posts that compare options, price things out, or list "best X" picks. In other words, Google AI Mode is largely answering B2B buyer questions with content that companies wrote about their own category.
#Do AI answers cite the brands they recommend?
27 of the 50 answers highlighted named companies or products. The rest, mostly "how to" questions, gave advice without naming brands.
- 74% Of 27 answers that highlighted brands, 20 also cited the website of at least one brand they named (Shruti Kaushik study, 17 Oct 2026).
- 43% Of 117 brand mentions, 50 came with a citation to that brand's own domain (Shruti Kaushik study, 17 Oct 2026).
The most highlighted brands were HubSpot (6 answers), Apollo (5), and Salesloft, ZoomInfo and Semrush (3 each). On "best LinkedIn ghostwriter for B2B founders", the three agencies named (Understory, Stop The Scroll, YellowInk Digital) all had their own sites cited. On "best GEO agency for B2B SaaS", LoudFace, Omniscient Digital and First Page Sage were named and cited. Being named and being cited travel together: the page that explains your offer is often the page the answer pulls from.
#What B2B founders should do with this
1. Publish the buyer-question page yourself. Agency and vendor blogs supplied 77% of citations. If buyers ask "how much does X cost" or "X vs Y" in your category, write the page that answers it with real numbers, a comparison table and a clear verdict. Most cited pages came from small companies, not category leaders. The founder AI visibility guide covers which pages to build first.
2. Treat LinkedIn as a source, not only a feed. LinkedIn was cited in 44% of all answers and 8 of 10 LinkedIn questions. LinkedIn posts and articles get indexed and pulled into answers. A founder who writes specific, named-number posts about their category is creating citable material every week.
3. Make one YouTube video per core buyer question. YouTube was the single most cited domain at 48%. A 5 to 10 minute walkthrough of "how to choose X" or "X pricing explained" is a direct path into the answer.
4. Be part of the Reddit conversation for Reddit-adjacent questions. Reddit appeared in 30% of answers and 8 of 10 community questions. Useful, disclosed answers in the subreddits your buyers read are what get surfaced, not promotional posts.
5. Check where you stand, then track it. Run a free AI visibility check to see whether ChatGPT, Perplexity and Google AI name your company. Read how to get your brand mentioned by ChatGPT for the content side, and use the AI referral tracker to see whether AI answers send you visits.
6. Do not wait for press. Publishers and analysts were 3.1% of citations. A Forbes feature is nice. It is not how you get into this answer set.
#Limitations
- One engine, one day, one location. Google AI Mode only, on 17 Oct 2026, from a browser in India. ChatGPT, Perplexity and Gemini cite differently, and AI Mode answers change between runs and by country.
- 50 questions is a sample. Group-level figures rest on 10 answers each. Treat them as directional.
- Domains, not pages. I recorded the domains linked from each answer, not individual URLs, so I cannot say how often a vendor's homepage, as opposed to a blog post, was the cited page. That needs a URL-level run.
- Brand capture is conservative. Brands were recorded when the answer highlighted them in bold or as a table row. Brands mentioned only in plain prose were not counted, so brand totals are a floor.
- Classification is mine. Each domain was labelled agency, vendor, platform, publisher or review site by hand. The CSV lets you re-classify.
- Logged-in browser. Queries ran in a normal Chrome profile. Personalisation in AI Mode is limited but cannot be ruled out.
#Press kit: 5 quotable findings
- "In 50 B2B buyer questions run through Google AI Mode on 17 Oct 2026, YouTube was cited in 48% of answers, LinkedIn in 44% and Reddit in 30%." (Shruti Kaushik, shrutikaushik.com)
- "Agency and software vendor websites supplied 77% of all citations in Google AI Mode answers to B2B buyer questions. Publishers and analysts supplied 3%." (Shruti Kaushik, shrutikaushik.com)
- "Wikipedia was cited in 0 of 50 Google AI Mode answers to B2B buying questions." (Shruti Kaushik, shrutikaushik.com)
- "239 different domains were cited across 50 AI answers, and 84% of them appeared only once. No single website outside YouTube, LinkedIn and Reddit appeared in more than 10% of answers." (Shruti Kaushik, shrutikaushik.com)
- "When Google AI Mode named brands, it cited at least one named brand's own website in 74% of those answers." (Shruti Kaushik, shrutikaushik.com)
Data: ai-citations-50-questions-2026-10.csv. Please link to this page when citing.
#How Shruti does this for clients
In my work with founders, I write the LinkedIn and Reddit content that this study shows Google AI Mode pulling from. On 15 Oct 2026 I checked 13 buyer queries about LinkedIn ghostwriting in Google AI Mode: the cited sources were LinkedIn posts and articles, LinkedIn profiles, small agency pages and listicles, and company homepages were rarely the source. This 50-question run tests that pattern across five B2B categories, and it holds.
For the full playbook, join my private letters.
#Methodology
I wrote 50 buyer-style questions across five B2B groups and ran each once in Google AI Mode (https://www.google.com/search?udm=50&q=...) on 17 October 2026, in a single Chrome tab from India. After each answer finished loading, I extracted every external domain linked on the answer page (Google domains excluded, subdomains merged into the parent domain) and every brand highlighted in bold or as a table row label. All 50 queries completed. Domains were hand-classified as platform, agency, software vendor, publisher or review directory. Every number on this page was computed from that single run. The raw data is at /data/ai-citations-50-questions-2026-10.csv.
#Sources
- Shruti Kaushik, "AI citations for 50 B2B buyer questions in Google AI Mode" (raw dataset), 17 Oct 2026. /data/ai-citations-50-questions-2026-10.csv
- Google, AI Mode (udm=50), queried 17 Oct 2026. https://www.google.com/search?udm=50