Which Sources Do ChatGPT, Perplexity, and Google AI Actually Cite? We Analysed 24,000 Responses to Find Out

When an AI assistant answers a question, it is drawing on sources. On some platforms, those sources are listed explicitly. On others, they are embedded silently in the response. Either way, the sources an AI cites determine which brands, domains, and voices it treats as authoritative.

We extracted every citation from over 24,000 AI responses across ChatGPT, Perplexity, and Google AI Overview, covering 45 UK businesses across sectors ranging from logistics and insurance to investment castings and youth organisations. The dataset covers 8,700 Perplexity responses, 7,000 Google AI Overview responses, and 9,000 ChatGPT responses, each prompted by commercially oriented queries about real markets.

The finding that matters most is this: the three platforms cite almost entirely different sources. A brand optimising for one is not optimising for the others. And in most cases, no one is optimising for any of them deliberately.

The Only Domain All Three Platforms Agree On

Across the entire dataset, only one domain was cited by all three AI platforms: clutch.co. It appeared 140 times in ChatGPT responses, 300 times in Perplexity responses, and 240 times in Google AI Overview responses.

Clutch is a verified B2B review and ratings platform. It earns citations on ChatGPT because it has been indexed and linked to extensively over many years. It earns citations on Perplexity because it is a structured, authoritative source of agency data with strong on-page signals. It earns citations on Google AI Overview because it ranks well for commercial comparison queries in organic search.

The lesson from Clutch is not that every brand should be on Clutch. It is that the domains which achieve cross-platform citation authority share three properties: they have been publishing structured content for a long time, they are referenced by other authoritative sites, and they give AI systems clear, parseable information to extract. These are not new ideas. They are the fundamentals of publishing, applied to a new context.

What Google AI Overview Cites

Reddit is the most cited domain in Google AI Overview responses by a significant margin, appearing 2,400 times across the dataset. That is more than double the second-placed domain and more than the next four combined. Facebook appears 1,000 times. LinkedIn appears 680 times.

The consumer trust layer is also heavily represented: Which appears 740 times, MoneySavingExpert 280 times, Uswitch 280 times, CompareTheMarket 240 times. Amazon appears 580 times, predominantly in product and brand recommendation queries.

The pattern is consistent: Google AI Overview is built on social proof and consumer authority. It is reading what people say about brands in discussion threads, on comparison sites, in buying guides. If your brand is discussed favourably in Reddit threads, referenced on comparison sites, and mentioned in consumer guides, Google will cite those discussions when someone asks about your market. If it is not, Google will cite discussions about your competitors who are.

None of these social domains appear in Perplexity top 50 citations. Not Reddit. Not Facebook. Not LinkedIn. They are entirely separate citation ecosystems.

What Perplexity Cites

YouTube is the most cited domain on Perplexity with 780 citations. It does not appear in Google AI Overview top citations at all. After YouTube, Perplexity cites brand official websites directly and frequently. In our dataset, a social cricket entertainment brand was cited 500 times, a UK youth military organisation 460 times, a motion capture technology manufacturer 440 times, and a digital marketing agency 400 times. These are primary sources, not aggregators.

Perplexity specialist citation profile is particularly notable in technical and industrial sectors. The platform cited an industrial sensor manufacturer 340 times, an investment castings trade directory 340 times, a precision optics manufacturer 260 times, a precision measurement company 260 times, and a precision engineering firm 220 times. These are specialist trade and manufacturer domains that do not surface prominently in Google AI Overview at all. Perplexity is reading technical literature, trade publications, and manufacturer specifications that Google AI appears to discount or ignore.

Wikipedia appears exclusively in Perplexity at 280 citations, entirely absent from both Google AI Overview and ChatGPT responses. Brands with a structured, accurate Wikipedia presence hold a citation advantage on Perplexity that does not transfer to the other platforms.

The pattern: Perplexity is a research engine that follows sources. It cites what it can verify and trace. A well-structured official website, an active YouTube channel, a Wikipedia entry, and coverage in specialist trade publications are the assets that determine Perplexity citation authority. Social media presence and consumer forum discussions, the backbone of Google AI Overview citations, do not move the needle on Perplexity at all.

What ChatGPT Cites

ChatGPT cites far fewer URLs than either Perplexity or Google AI Overview. When it does include links, they are embedded inline within flowing prose rather than listed as sources, and they appear less frequently. In our dataset, the top cited domains were a major UK telecoms provider (300 citations), a digital marketing agency (220), and two further telecoms and broadband providers (200 each).

The digital agency sector is noticeably well-represented in ChatGPT inline citations, with five separate agencies appearing consistently across prompts. This reflects how the model weights industry publications and recommendation articles that feature agency rankings.

The critical point about ChatGPT citations is that they reflect what existed in the training data at the model knowledge cutoff, not what is live on the web today. Unlike Perplexity, which searches in real time, and unlike Google AI Overview, which indexes current content, ChatGPT cites what it has already learned. A brand that built strong web presence two or three years ago is better positioned than a brand that has only recently started publishing, regardless of current content quality.

This means ChatGPT citation authority decays differently from the other platforms. A brand can update its website and publish new content today, and Perplexity will reflect that within days. Google AI Overview will pick it up within weeks. ChatGPT will not reflect it until the next training cycle, which may be months or years away.

The Platform-Specific Citation Logic

After analysing the full dataset, three distinct citation logics emerge.

Google AI Overview cites what people are saying. Reddit threads, Facebook groups, LinkedIn posts, consumer comparison sites, buyer guides, and Amazon listings are all representations of collective opinion. Google AI is trained on a model of trust that is fundamentally social. For brands, this means community presence is not a nice-to-have. It is a citation prerequisite on the platform that accounts for the largest share of AI-assisted search queries in the UK.

Perplexity cites what can be verified. Primary sources, official brand websites, YouTube, Wikipedia, and structured trade directories are all manifestations of traceable authority. Perplexity is used predominantly by researchers, analysts, and technically sophisticated buyers. Its citation model reflects this: it follows the evidence rather than the crowd. For B2B brands, industrial manufacturers, and specialist service providers, this is the platform with the highest alignment between citation behaviour and actual buyer behaviour.

ChatGPT cites what it remembers. Its inline references are drawn from training data that may be one to three years old. Recency is less relevant than historical footprint. For new entrants to a market, this is a structural disadvantage that cannot be solved with content alone.

The Brands That Cross Platforms

Only a handful of domains in the dataset achieve citation presence on more than one platform. A UK youth and military organisation is cited 1,140 times by Google AI Overview and 460 times by Perplexity, making it one of the strongest cross-platform examples in the data. It earns this through government backlinks, structured youth organisation directories, extensive on-site content, and genuine community discussion across forums and social media.

A social cricket entertainment brand is cited 500 times on Perplexity and 460 times on Google AI Overview. Its content spans official venue and match information, structured location data, social media presence, and press coverage. The brand satisfies the citation requirements of both platforms simultaneously through comprehensive, well-maintained digital presence across multiple content formats.

An AI-powered manufacturing automation platform, operating on a relatively new domain, appears 180 times on Perplexity and 280 times on Google AI Overview. This suggests that well-structured, research-oriented content on a newer domain can still earn cross-platform citation authority faster than many established brand websites. The site is topic-specific, well-linked within its niche, and targets the kind of research queries that both platforms surface.

What This Means Practically

The practical implication of this analysis is uncomfortable for most brands: the activities that drive citation authority on one AI platform actively do nothing for the others. Publishing a detailed technical whitepaper will help Perplexity citations. It will have minimal impact on Google AI Overview, which needs Reddit threads and consumer reviews. Accumulating Reddit presence will help Google AI Overview. It will not move Perplexity, which does not cite Reddit at all. Neither strategy helps ChatGPT directly, where the citation pool is largely frozen in historical training data.

This is the multi-platform citation problem that very few brands are currently aware of, let alone solving for. Most are optimising for one search engine. Some are beginning to think about one AI platform. Almost none are thinking simultaneously about the distinct citation requirements of all three major AI surfaces.

The brands in this dataset that achieve the strongest cross-platform citation presence are not doing sophisticated AI optimisation. They are doing the full range of digital activities well: maintaining structured authoritative official content, being genuinely present in community discussions, producing accessible formats including video, and earning coverage in sector-appropriate publications over time. These activities happen to satisfy the citation requirements of all three platforms because they represent comprehensive digital presence, not because anyone planned it that way.

For any brand, the audit question is not which platform to optimise for. It is whether the current content and community programme covers the citation surface of each platform at all. For most, the answer reveals significant gaps. Knowing which sources each platform actually cites is the starting point. What happens next is the strategy question.

---

Further Reading

• We Ran 4,500 AI Prompts Across 349 Brands. Two-Thirds Are Invisible.
• ChatGPT and Google AI Give Completely Different Answers. Your Brand Might Only Exist on One.
• From 0% to Cited: What Brands Scoring 80%+ on AI Actually Have in Common

Track how your brand appears across ChatGPT, Perplexity, and Google AI Overview. Start free or see pricing.