What 1,454 AI Citations Across 7 Engines Show
1,454 AI Citations
What 7 AI engines cite, and where they differ
Written by Punn Kam, ex-Google AI engineer and Y Combinator repeat founder. He is the co-founder of Conduit, an AI agent platform built for hospitality, serving 300+ hospitality brands in 140+ languages. Last updated October 8, 2026.
We pulled 1,454 citations from seven AI engines answering the same 33 hospitality software questions, and the engines barely agree. 76% of the cited URLs appeared on one engine only. ChatGPT cited product and feature pages most (40% of its citations), while Claude, Copilot and Perplexity leaned on "best X" listicles (50% to 68%). YouTube and Reddit mattered only on Google's two AI surfaces.
All 33 prompts are questions a hotel, short-term rental or villa operator asks when buying software, such as "What's the best AI guest-messaging tool for hotels?" The numbers describe how AI engines pick sources for those questions, and they do not show how a hotel gets recommended to a guest searching for a place to stay.
How did we collect 1,454 AI citations from 7 engines?
The 1,454 citations come from Ahrefs Brand Radar, pulled on October 8, 2026 for the 33 custom software-buyer prompts Conduit tracks (US, created July 18 to August 24, 2026). For each prompt we took every answer and cited link from ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity, Microsoft Copilot and Claude: 240 answers, 234 of them dated October 1 to 9, 2026. Some engines returned more than one answer for a prompt, and we counted each one.
We hand-assigned each of the 388 cited domains to a category (unlisted domains default to "software vendor") and set page types with title and URL rules. Conduit publishes posts that match many of these prompts, so conduit.ai is over-represented and gets its own category.
How many sources does each AI engine cite per answer?
Perplexity cites the most sources per answer, about 10, roughly two and a half times Gemini (3.6) or Copilot (4.2). Claude gave no citations on 8 of its 33 answers, and AI Mode on 5 of 37, nearly all of them "How do..." or "How can..." process questions.
| Engine | Answers | Answers with a citation | Citations | Mean per cited answer | Unique URLs |
|---|---|---|---|---|---|
| ChatGPT | 34 | 34 | 229 | 6.7 | 191 |
| Google AI Overviews | 34 | 34 | 269 | 7.9 | 193 |
| Google AI Mode | 37 | 32 | 205 | 6.4 | 142 |
| Gemini | 33 | 33 | 119 | 3.6 | 106 |
| Perplexity | 35 | 35 | 350 | 10.0 | 281 |
| Microsoft Copilot | 34 | 33 | 137 | 4.2 | 110 |
| Claude | 33 | 25 | 145 | 5.8 | 121 |
Do ChatGPT, Google and Perplexity cite the same pages?
ChatGPT, Google and Perplexity mostly cite different pages for the same question. Of 810 unique URLs across all engines, 614 (76%) were cited by one engine only. Two URLs were cited by all seven: Conduit's 7 best AI guest communication platforms for hotels post and HotelTechReport's guest messaging category page.
Google AI Overviews and AI Mode are the closest pair, at 22% URL overlap (Jaccard) and 29% domain overlap. ChatGPT and AI Overviews share 24 URLs, a 7% overlap, and ChatGPT and Claude are the most distant pair at 5%.
Which page types does ChatGPT cite compared with other engines?
ChatGPT cites product and feature pages far more than any other engine: 40% of its citations, against 13% to 26% elsewhere. It also cites vendor homepages (12%), integration pages (9%) and help-centre articles (7%). Listicles make up 15% of ChatGPT's citations, against 50% on Perplexity, 57% on Copilot and 68% on Claude.
| Page type | ChatGPT | AI Overviews | AI Mode | Gemini | Perplexity | Copilot | Claude |
|---|---|---|---|---|---|---|---|
| Listicle ("best X", "top N") | 15% | 30% | 33% | 37% | 50% | 57% | 68% |
| Product or feature page | 40% | 25% | 18% | 26% | 17% | 13% | 13% |
| How-to, guide or article | 9% | 17% | 13% | 19% | 21% | 17% | 7% |
| Vendor homepage | 12% | 2% | 3% | 5% | 5% | 2% | 1% |
| Help centre or docs | 7% | 1% | 2% | 2% | 1% | 4% | 1% |
| Integration or marketplace page | 9% | 0% | 1% | 3% | 0% | 1% | 0% |
| Comparison or alternatives page | 7% | 4% | 7% | 8% | 5% | 7% | 10% |
| YouTube, Reddit or social post | 0% | 16% | 13% | 1% | 1% | 0% | 0% |
| Google search or shopping link | 0% | 5% | 9% | 0% | 0% | 0% | 0% |
Software vendors, Conduit included, wrote most of the listicles the engines cited: 59% of listicle citations on Copilot, 67% on Perplexity, 69% on Claude and between 84% and 91% on ChatGPT, Gemini and Google's two surfaces. "2025" or "2026" appears in the URL or title of 74% of Claude's citations and 71% of Copilot's, against 11% of ChatGPT's.
Which websites do AI engines cite for hospitality software questions?
Software vendors' own websites are the largest source on every engine, at 60% of all 1,454 citations with Conduit excluded. The second source depends on the engine. Copilot takes 31% of its citations from review aggregators and "best software" directories, Claude 23% and Perplexity 17%, while Google's two surfaces give that slot to YouTube, Reddit and social posts.
| Source type | ChatGPT | AI Overviews | AI Mode | Gemini | Perplexity | Copilot | Claude | All |
|---|---|---|---|---|---|---|---|---|
| Software vendors (not Conduit) | 67% | 53% | 52% | 76% | 66% | 41% | 63% | 60% |
| conduit.ai | 14% | 14% | 17% | 9% | 7% | 16% | 11% | 12% |
| Review aggregators and directories | 7% | 4% | 5% | 6% | 17% | 31% | 23% | 12% |
| YouTube, Reddit and social | 0% | 16% | 13% | 1% | 1% | 0% | 0% | 5% |
| Google-internal links | 0% | 5% | 9% | 0% | 0% | 0% | 0% | 2% |
| Agencies, consultants, blogs | 4% | 3% | 2% | 4% | 2% | 7% | 1% | 3% |
| Property operators | 5% | 4% | 0% | 3% | 4% | 3% | 0% | 3% |
| Trade media and publishers | 1% | 1% | 1% | 0% | 3% | 2% | 1% | 2% |
| OTAs (Airbnb, Booking.com, Expedia, Tripadvisor, Vrbo) | 2% | 0% | 0% | 1% | 0% | 0% | 0% | 0% |
OTAs barely appear because these are software questions. Cloudbeds' study of ChatGPT, Perplexity and Gemini answers about six destinations found the opposite for travel questions: 55.3% of cited sources were OTAs (Hospitality Net, June 26, 2025).
Where do YouTube and Reddit show up in AI answers?
YouTube and Reddit show up almost only on Google. Google AI Overviews took 16% of its citations from YouTube, Reddit, Facebook and LinkedIn (YouTube alone: 30 citations), and AI Mode took 13% (15 YouTube, 10 Reddit). ChatGPT, Copilot and Claude cited no YouTube or Reddit pages, and Gemini and Perplexity cited one each.
One HotelTechReport video, "Top 5 Guest Messaging Software in 2026", was cited in 8 answers on Google's two surfaces and never by ChatGPT or Claude.
How stable are these AI citation numbers?
The AI citation patterns held between pulls, but individual figures moved within a day. An earlier pull of the same report on October 8, 2026 returned 1,391 citations, mostly because ChatGPT's count was 169 before Brand Radar refreshed answers. ChatGPT's product-page share moved from 47% to 40%, and the single-engine share stayed at 76%. Treat any one figure as plus or minus a few points.
What should hotels and hospitality software teams do with this data?
Hotels and hospitality software teams should plan for each AI engine separately, because the engines in this data draw on different pages and different types of site for the same question. Four moves follow from the counts above and the Ahrefs study linked here:
- For ChatGPT, publish clear product, feature, integration and help pages, because 40% of its citations went to that kind of page.
- For Claude, Copilot and Perplexity, get onto third-party "best of" lists and directories, and date your comparison pages.
- For Google's AI surfaces, add YouTube. Google says there are "no additional requirements to appear in AI Overviews or AI Mode" beyond being indexed with a snippet (Google Search Central), so ordinary search work still applies.
- Earn mentions on other sites. In Ahrefs' study of 75,000 brands, branded web mentions correlated with AI Overview visibility at 0.664, against 0.218 for backlinks (Ahrefs, May 26, 2025). Ahrefs sells Brand Radar and notes that correlation is not causation.
What can't this AI citation data tell you?
The Brand Radar citation data cannot tell you how a hotel gets recommended to travellers, because all 33 prompts are software-buyer questions. It is one snapshot from one tool with hand-built classifications, and moving edge cases such as StayFi or RoomMaster blogs between categories shifts percentages by a few points. It is correlational, so it cannot say why an engine picked a source. For how AI agents are starting to book rooms, see Will AI agents replace Booking.com?.
FAQ
Each answer below repeats its key figure from the October 8, 2026 Brand Radar pull of 33 hospitality software prompts, so it can be quoted without the tables above. Every percentage is a share of cited links, counted the same way as in the method section.
Do ChatGPT and Google AI Overviews cite the same sources?
ChatGPT and Google AI Overviews rarely cite the same sources. In our October 8, 2026 pull of 33 hospitality software prompts, they shared 24 URLs, a 7% overlap, and 76% of all cited URLs appeared on one engine only.
Which AI engine cites the most sources per answer?
Perplexity cites the most, about 10 per cited answer in our data. Google AI Overviews averaged 7.9, ChatGPT 6.7, AI Mode 6.4, Claude 5.8, Copilot 4.2 and Gemini 3.6.
Does ChatGPT prefer listicles?
ChatGPT cites listicles less than any other engine in our data: 15% of its citations, against 50% for Perplexity, 57% for Copilot and 68% for Claude. Product and feature pages were its largest group, at 40%.
Do YouTube and Reddit help with AI search visibility?
YouTube and Reddit matter on Google's AI surfaces. YouTube, Reddit and social posts were 16% of Google AI Overviews citations and 13% of AI Mode citations in our data, while ChatGPT, Copilot and Claude cited none.
Is this data about how travellers find hotels?
This data covers software buyers. The 33 prompts are questions hotel and rental operators ask when choosing software. Cloudbeds' study of AI answers about six destinations found that 55.3% of cited sources were OTAs.
Sources and access dates: Brand Radar data pulled October 8, 2026 from 33 custom software-buyer prompts. The Google Search Central, Ahrefs and Hospitality Net pages linked above were read on October 8, 2026.
Stay in the loop
Get the latest on AI automation, product updates, and customer stories.
