The short answer
We asked 25 B2B buyer questions in Perplexity, Google AI Mode and ChatGPT on 8 October 2026. In the 442 source links Perplexity and Google AI Mode showed, 323 (73.1%) went to software vendors and agencies. G2, Capterra, TrustRadius and Wikipedia did not appear once.
Google AI Mode leaned on YouTube and LinkedIn: each showed up on 10 of its 25 questions. Perplexity showed no YouTube at all. The source pool is a long tail. Of the 214 domains Perplexity used, 196 appeared on a single question, so no one owns these questions yet.
Headline findings
- Vendors and agencies supply most of the sources. 323 of 442 links (73.1%): 187 software vendors (42.3%), 136 agencies and consultancies (30.8%). Perplexity 198 of 251 (78.9%), Google AI Mode 125 of 191 (65.4%).
- Google AI Mode uses video and social; Perplexity mostly does not. YouTube: 20 links on 10 Google questions, none in Perplexity. LinkedIn: 10 Google questions, 5 Perplexity questions. Reddit: 4 Google questions, 1 Perplexity question.
- No G2, Capterra or Wikipedia. Zero links to G2, Capterra, TrustRadius, Gartner or Wikipedia across both engines, including the three questions about choosing an agency or partner.
- The long tail is open. Perplexity used 214 domains, 196 of them on one question only. Google AI Mode used 141, 124 on one question only. Outside LinkedIn, YouTube and Reddit, no domain appeared on more than 3 of 25 questions in either engine.
- The two engines rarely agree. For the same question they showed the same URL 16 times in total, on 11 of 25 questions. They shared at least one domain on 15 questions.
- GTMhub was not cited. No
gtmhub.ailink in either engine's sources and no mention of GTMhub as an agency in any of the 75 answers. This was the site's launch day, so this is the starting point.
Method
| Item | Detail |
|---|---|
| Questions | 25 fixed buyer questions across five service areas (listed below) |
| Engines | Perplexity (signed-in Pro, Search), Google AI Mode (signed in), ChatGPT (signed-in Plus, Instant, web search on) |
| Date and locale | 8 October 2026, English, India |
| Prompt | The question plus "Please answer in at most 150 words with up to five cited sources." Perplexity Q01 to Q04 used the plain question |
| Counted | Every URL in each answer's source panel or rendered source list, de-duplicated per answer |
| Excluded | 77 Google interface links (privacy, terms, legal notices, one AI Mode help link, one sponsored link) and one Perplexity reference drawn from the signed-in account's own context |
| Classification | Each domain labelled by GTMhub from its homepage: vendor, agency, help docs, video, LinkedIn, community, media, independent writer, academic, reference. 10 domains could not be classified |
Limits, stated once: source panels can include retrieved links beyond the inline citations, accounts were signed in, and one sample per question is directional, not share of search. ChatGPT's source buttons exposed no destination URLs, so ChatGPT appears only where the data supports it. Gemini was unavailable. Google AI Mode showed no external sources for Q06.
Results by source type
| Source type | Perplexity links | Google AI Mode links | Both engines | Share of all |
|---|---|---|---|---|
| Software vendors | 113 | 74 | 187 | 42.3% |
| Agencies and consultancies | 85 | 51 | 136 | 30.8% |
| LinkedIn posts and articles | 10 | 12 | 22 | 5.0% |
| YouTube | 0 | 20 | 20 | 4.5% |
| Platform help docs (mostly Google Ads) | 8 | 7 | 15 | 3.4% |
| News and trade media | 8 | 6 | 14 | 3.2% |
| Independent writers, newsletters, podcasts | 6 | 6 | 12 | 2.7% |
| Reddit, Quora and other communities | 3 | 8 | 11 | 2.5% |
| Academic and library | 9 | 1 | 10 | 2.3% |
| Reference and directory sites | 3 | 2 | 5 | 1.1% |
| Unclassified | 6 | 4 | 10 | 2.3% |
| Total | 251 | 191 | 442 |
LinkedIn and YouTube are counted by platform, whoever posted. Independent content as a whole (every row except vendors, agencies, help docs and unclassified) was 39 Perplexity links (15.5%) and 55 Google AI Mode links (28.8%).
Results by service area
| Area | Questions | Links | Vendors | Agencies | Independent | Vendor and agency share |
|---|---|---|---|---|---|---|
| Outbound and ABM | 4 | 73 | 40 | 19 | 10 | 80.8% |
| GTM strategy and the brain | 8 | 125 | 58 | 41 | 24 | 79.2% |
| Paid media | 5 | 96 | 27 | 41 | 16 | 70.8% |
| GTM engineering | 4 | 72 | 40 | 10 | 18 | 69.4% |
| AI search and content | 4 | 76 | 22 | 25 | 26 | 61.8% |
Three patterns stand out. Paid media is the only area where agencies outnumber vendors, and Google's own Google Ads help pages supplied 10 of its links. In GTM engineering, vendors outnumber agencies four to one. AI search looks the most independent, but 9 of its 26 independent links came from one question (Q15), where Perplexity pulled university library guides on checking AI sources.
The GTM brain questions are being answered by vendors that use the term. For "What should be in a company GTM brain?", 7 of Perplexity's 10 sources were software vendors. In a separate check the same day, ChatGPT answered it citing HubSpot, Gartner and McKinsey by name, with no link targets we could verify.
What it means for a B2B team
- Write the page the vendors are being cited for. Most sources are vendor and agency pages answering the buyer's exact question. With 196 of 214 Perplexity domains appearing only once, one clear page that answers the question in its first two lines can earn a place. Turn the question into a page brief first.
- Treat Google AI Mode and Perplexity as two audiences. They shared a URL on only 11 of 25 questions. Google AI Mode pulled YouTube on 10 questions and Reddit on 4; Perplexity pulled neither much. If Google matters most to your buyers, a video and a LinkedIn article on the same question belong in the plan.
- Do not start with G2 or Capterra for problem and method questions. Neither appeared in 442 links. Spend the effort on pages, posts and outside coverage that answer the question itself.
- Use the vendor docs that already rank. For Google Ads questions, the engines cited Google's own help pages. Link to them and add what they lack: your CRM stages, your values, your decision rules.
- Measure per engine, on a fixed set. Keep the questions and prompt wording fixed, record sign-in state, and count owned citations separately from mentions. The source-gap review sets this up and the citation playbook covers the fixes.
GTMhub repeats this sample monthly on the same 25 questions, from 9 November 2026, and updates this page with what changes.
The 25 questions
| # | Question | Area |
|---|---|---|
| Q01 | Why do our paid leads look good but pipeline does not grow? | Paid media |
| Q02 | Why do sales and marketing report different pipeline numbers? | GTM strategy |
| Q03 | How do we stop demo requests going unworked? | GTM engineering |
| Q04 | Why are our cold email reply rates falling? | Outbound and ABM |
| Q05 | Why is our B2B company missing from AI recommendations? | AI search |
| Q06 | How should a B2B team define qualified pipeline? | GTM strategy |
| Q07 | How do we connect Google Ads leads to CRM outcomes? | Paid media |
| Q08 | How should a B2B team value leads for bidding? | Paid media |
| Q09 | What should an inbound lead routing workflow check first? | GTM engineering |
| Q10 | How should we handle a demo request from an existing customer? | GTM engineering |
| Q11 | How do we choose between volume outbound and signal-led outbound? | Outbound and ABM |
| Q12 | How do we stop stale signals entering an outbound campaign? | Outbound and ABM |
| Q13 | What should be in a company GTM brain? | GTM strategy |
| Q14 | How does a GTM brain connect to a CRM? | GTM strategy |
| Q15 | How do we review which sources an AI answer cites? | AI search |
| Q16 | Should we use CRM-native automation or an orchestration tool? | GTM strategy |
| Q17 | When does Tier 1 ABM justify more research than signal-led outbound? | Outbound and ABM |
| Q18 | Should we optimise Google Ads for leads or qualified opportunities? | Paid media |
| Q19 | How should we compare enrichment tools on our own records? | GTM engineering |
| Q20 | What is the difference between a brand mention and an AI citation? | AI search |
| Q21 | What should we ask a GTM engineering agency before hiring it? | GTM strategy |
| Q22 | How should we evaluate a B2B paid media partner? | Paid media |
| Q23 | What should an AI search agency measure? | AI search |
| Q24 | Who owns the GTM brain when an agency builds it? | GTM strategy |
| Q25 | How much client time does a GTM systems engagement require? | GTM strategy |
Questions buyers ask
Which sources do AI engines cite most for B2B questions?
In our sample, software vendors and agencies: 323 of 442 source links (73.1%) across Perplexity and Google AI Mode. After them came LinkedIn (22 links), YouTube (20, all in Google AI Mode) and Google's own help pages (15).
Do G2 or Wikipedia pages get cited for B2B GTM questions?
Not in this sample. Neither engine showed a single G2, Capterra, TrustRadius, Gartner or Wikipedia link across 50 answers, including three questions about choosing an agency. Questions that name products or ask for comparisons may behave differently.
Does YouTube help a B2B company appear in AI answers?
In Google AI Mode, YouTube appeared on 10 of 25 questions with 20 links. In Perplexity it appeared on none. A video that answers the buyer's question has a place where Google AI Mode matters, but one sample does not show that posting produces citations.
How often should we check AI citations?
Monthly on a fixed question set is enough to see change without chasing noise. Keep the wording, engine settings and sign-in state the same each time, and record owned citations and brand mentions separately.
Why does ChatGPT have no source counts in this study?
ChatGPT's source buttons in the signed-in app exposed no destination URLs we could verify, so counting them would have meant guessing. We report ChatGPT only where the data holds: 25 completed answers, none naming GTMhub.
Publication and source record
Published on GTMhub: . Last reviewed: .