Why the same UK query returns different sources

Ask ChatGPT and Perplexity the same commercial UK question, best conveyancing solicitor in Leeds, private GP that does same-day appointments in London, and you get two answers built from mostly different pages. This is not random. It follows from how each engine retrieves. Profound’s analysis of roughly 680 million citations found only about 11% of domains cited by both engines, which is a small enough overlap that a single publishing strategy is almost guaranteed to underperform on one of them.

The gap has a mechanism, and the mechanism is stable enough to plan around even though individual rankings churn week to week. Once you can name the three levers each engine weights differently, you can stop guessing and start allocating effort where it changes citations on the engine you care about.

Lever one: how the two engines retrieve

Perplexity performs a real-time web search for every query, drawing on multiple search APIs, retrieving and reading candidate pages, then synthesising an answer with inline numbered citations. Source attribution is a core feature, not an add-on. ChatGPT blends its training data with selective web retrieval and absorbs sources more deeply into a single synthesised answer, surfacing fewer explicit citations. That difference alone reshapes the cited set. Perplexity averages about 21.9 citations per response against ChatGPT’s 10.4, so it mechanically widens the field and hands newer or smaller UK sites more room to appear.

Lever two: recency weighting

Perplexity weights recency visibly. In one 2026 analysis it cited content published within the previous 30 days at roughly an 82% rate, which means original coverage published this week can surface there quickly. ChatGPT, leaning more on training data and a set of trusted anchors, is slower to reflect brand-new pages and more likely to return an established reference source. For a UK publisher this sets a cadence. Time-sensitive, newsworthy or freshly-dated content is a Perplexity play. Durable, well-structured reference content compounds on ChatGPT.

Lever three: source-type preference

The two engines prefer different kinds of page. Across multiple 2026 studies, reference domains led by Wikipedia recur among ChatGPT’s most-cited sources, alongside established aggregators it appears to treat as reliable. Perplexity spreads its citations across a broader, more news and primary-source weighted set. The consequence for brand visibility is stark. A 2026 benchmark of 34,234 responses measured brand citation at 0.59% on ChatGPT against 13.05% on Perplexity, a 46-fold difference that reflects Perplexity’s greater willingness to cite a brand’s own pages directly rather than route through a reference intermediary.

The fourth signal: anchor text and answer position

Beyond retrieval, recency and source type, the two engines differ in how the wording around a link shapes what gets cited. Perplexity reads and quotes candidate pages closely, so the exact phrasing near a claim, the sentence that states a statistic, the heading that frames a definition, influences whether that passage is pulled into the answer. Clear, self-contained sentences that restate the question tend to be lifted verbatim. ChatGPT, synthesising more loosely across a smaller anchor set, is less sensitive to any single passage and more sensitive to whether your entity is already associated with the topic across the wider web.

Answer position matters too. On Perplexity, a source that supplies the first concrete fact in an answer often earns the first numbered citation, so leading with the specific figure a query implies is a practical edge. On ChatGPT, being named at all depends more on prior association than on passage placement, which is why brand-building on trusted third-party surfaces pays off there. The takeaway is that on-page phrasing is a Perplexity lever, while off-page association is a ChatGPT lever, and the two need different work.

Reading the three levers together

Put the levers on a single grid and the two engines separate cleanly. Perplexity scores high on recency and primary-source weight and rewards direct brand citation. ChatGPT scores high on reference-domain preference and rewards durable, structured content that maps to sources it already trusts. Neither is better in the abstract. They are different distribution channels with different admission rules, and a UK firm that measures both can decide which one matters for its buyers before it spends a content budget.

What this means for a UK content plan

The two-channel reality forces a small number of concrete choices. For Perplexity, publish original, dated, primary material and keep it current, because recency and direct citation reward first-hand UK data and timely commentary. For ChatGPT, invest in durable, clearly-structured reference pages and in the third-party surfaces engines trust, because the path to being named often runs through a reference source rather than your own page. The shared foundation, confirmed by the original GEO research, is structured content with explicit attribution, which raises citation likelihood on both. We set out the full channel-by-channel playbook in a companion post, linked in the further reading below.

A worked example makes the split concrete. Say a Manchester accountancy firm wants to appear for “R&D tax credit changes 2026”. The Perplexity play is a dated, first-party explainer that leads with the specific rule change and the date it takes effect, refreshed the week guidance shifts, so recency and passage clarity work in its favour. The ChatGPT play is different. It depends on the firm already being associated with R&D tax across trusted surfaces, a professional directory, an accountancy body listing, national press commentary, so that when ChatGPT assembles an answer it names the firm even when the cited source is a reference page. Same query, same firm, two separate bodies of work, which is exactly why a blended visibility number would hide the gap that matters.

Frequently asked questions

Do ChatGPT and Perplexity cite the same sources?

Mostly not. Profound’s 2026 analysis of about 680 million citations found only around 11% of domains were cited by both engines. They retrieve differently, Perplexity through a live search on every query and ChatGPT through a blend of training data and selective retrieval, so the same UK query returns largely different cited pages. A publishing strategy tuned for one engine will usually underperform on the other unless you plan for both.

Which engine cites more sources per answer?

Perplexity, by a wide margin. It averages about 21.9 citations per response against ChatGPT’s roughly 10.4 in 2026 analyses. Perplexity casts a broad net and reads many candidate pages before synthesising, which gives smaller and newer UK sites more chances to appear. ChatGPT surfaces fewer explicit citations because it absorbs sources more deeply into a single synthesised answer and leans on a smaller set of trusted anchors.

Why is my brand cited on Perplexity but not ChatGPT?

Because Perplexity is far more willing to cite a brand’s own pages directly. A 2026 benchmark of 34,234 responses found brand citation at 13.05% on Perplexity against 0.59% on ChatGPT, a 46-fold gap. ChatGPT more often routes through a reference domain such as Wikipedia or an established aggregator, so your brand can be the substance of an answer without your page being the cited source. Building reference-surface presence helps close that gap.

Does recency matter more on one engine?

Yes, on Perplexity. It weights recency visibly and in one 2026 analysis cited content from the previous 30 days at about an 82% rate, so fresh, dated material can surface there fast. ChatGPT reflects brand-new pages more slowly because it leans on training data and a set of durable trusted sources. Time-sensitive content is a Perplexity play, while durable reference content compounds on ChatGPT.

Should I optimise for one engine or both?

Measure both, then decide by audience. Because the engines share only about 11% of cited domains, presence on one does not carry over to the other. Identify which engine your buyers actually use, then weight effort accordingly. The shared foundation is structured content with explicit attribution, which raises citation likelihood everywhere, so start there and layer engine-specific tactics, recency for Perplexity and reference-surface authority for ChatGPT, on top.

Is this UK-specific or a general pattern?

The mechanism is general, but the cited domains are local. For UK commercial queries the reference and directory sources that ChatGPT favours are often UK-specific registers, professional bodies and national publications, while Perplexity surfaces UK news and primary sources. So the levers hold everywhere, but the exact domains you need to appear on are country-specific, which is why a UK firm should audit UK queries rather than rely on US-centric visibility studies.

By Beata Nowak, Strategy Lead, AiBoost | Published 13 July 2026 | 8 min read